1PERLGUTS(1) Perl Programmers Reference Guide PERLGUTS(1)
2
6 perlguts - Introduction to the Perl API
7
9 This document attempts to describe how to use the Perl API, as well as
10 to provide some info on the basic workings of the Perl core. It is far
11 from complete and probably contains many errors. Please refer any
12 questions or comments to the author below.
13
15 Datatypes
16 Perl has three typedefs that handle Perl's three main data types:
17 SV Scalar Value
19 AV Array Value
20 HV Hash Value
21 Each typedef has specific routines that manipulate the various data
23 types.
24 What is an "IV"?
26 Perl uses a special typedef IV which is a simple signed integer type
27 that is guaranteed to be large enough to hold a pointer (as well as an
28 integer). Additionally, there is the UV, which is simply an unsigned
29 IV.
30 Perl also uses several special typedefs to declare variables to hold
32 integers of (at least) a given size. Use I8, I16, I32, and I64 to
33 declare a signed integer variable which has at least as many bits as
34 the number in its name. These all evaluate to the native C type that
35 is closest to the given number of bits, but no smaller than that
36 number. For example, on many platforms, a "short" is 16 bits long, and
37 if so, I16 will evaluate to a "short". But on platforms where a
38 "short" isn't exactly 16 bits, Perl will use the smallest type that
39 contains 16 bits or more.
40 U8, U16, U32, and U64 are to declare the corresponding unsigned integer
42 types.
43 If the platform doesn't support 64-bit integers, both I64 and U64 will
45 be undefined. Use IV and UV to declare the largest practicable, and
46 ""WIDEST_UTYPE" in perlapi" for the absolute maximum unsigned, but
47 which may not be usable in all circumstances.
48 A numeric constant can be specified with ""INT16_C"" in perlapi,
50 ""UINTMAX_C"" in perlapi, and similar.
51 Working with SVs
53 An SV can be created and loaded with one command. There are five types
54 of values that can be loaded: an integer value (IV), an unsigned
55 integer value (UV), a double (NV), a string (PV), and another scalar
56 (SV). ("PV" stands for "Pointer Value". You might think that it is
57 misnamed because it is described as pointing only to strings. However,
58 it is possible to have it point to other things. For example, it could
59 point to an array of UVs. But, using it for non-strings requires care,
60 as the underlying assumption of much of the internals is that PVs are
61 just for strings. Often, for example, a trailing "NUL" is tacked on
62 automatically. The non-string use is documented only in this
63 paragraph.)
64 The seven routines are:
66 SV* newSViv(IV);
68 SV* newSVuv(UV);
69 SV* newSVnv(double);
70 SV* newSVpv(const char*, STRLEN);
71 SV* newSVpvn(const char*, STRLEN);
72 SV* newSVpvf(const char*, ...);
73 SV* newSVsv(SV*);
74 "STRLEN" is an integer type ("Size_t", usually defined as "size_t" in
76 config.h) guaranteed to be large enough to represent the size of any
77 string that perl can handle.
78 In the unlikely case of a SV requiring more complex initialization, you
80 can create an empty SV with newSV(len). If "len" is 0 an empty SV of
81 type NULL is returned, else an SV of type PV is returned with len + 1
82 (for the "NUL") bytes of storage allocated, accessible via SvPVX. In
83 both cases the SV has the undef value.
84 SV *sv = newSV(0); /* no storage allocated */
86 SV *sv = newSV(10); /* 10 (+1) bytes of uninitialised storage
87 * allocated */
88 To change the value of an already-existing SV, there are eight
90 routines:
91 void sv_setiv(SV*, IV);
93 void sv_setuv(SV*, UV);
94 void sv_setnv(SV*, double);
95 void sv_setpv(SV*, const char*);
96 void sv_setpvn(SV*, const char*, STRLEN)
97 void sv_setpvf(SV*, const char*, ...);
98 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *,
99 SV **, Size_t, bool *);
100 void sv_setsv(SV*, SV*);
101 Notice that you can choose to specify the length of the string to be
103 assigned by using "sv_setpvn", "newSVpvn", or "newSVpv", or you may
104 allow Perl to calculate the length by using "sv_setpv" or by specifying
105 0 as the second argument to "newSVpv". Be warned, though, that Perl
106 will determine the string's length by using "strlen", which depends on
107 the string terminating with a "NUL" character, and not otherwise
108 containing NULs.
109 The arguments of "sv_setpvf" are processed like "sprintf", and the
111 formatted output becomes the value.
112 "sv_vsetpvfn" is an analogue of "vsprintf", but it allows you to
114 specify either a pointer to a variable argument list or the address and
115 length of an array of SVs. The last argument points to a boolean; on
116 return, if that boolean is true, then locale-specific information has
117 been used to format the string, and the string's contents are therefore
118 untrustworthy (see perlsec). This pointer may be NULL if that
119 information is not important. Note that this function requires you to
120 specify the length of the format.
121 The "sv_set*()" functions are not generic enough to operate on values
123 that have "magic". See "Magic Virtual Tables" later in this document.
124 All SVs that contain strings should be terminated with a "NUL"
126 character. If it is not "NUL"-terminated there is a risk of core dumps
127 and corruptions from code which passes the string to C functions or
128 system calls which expect a "NUL"-terminated string. Perl's own
129 functions typically add a trailing "NUL" for this reason.
130 Nevertheless, you should be very careful when you pass a string stored
131 in an SV to a C function or system call.
132 To access the actual value that an SV points to, Perl's API exposes
134 several macros that coerce the actual scalar type into an IV, UV,
135 double, or string:
136 • SvIV(SV*) ("IV") and SvUV(SV*) ("UV")
138 • SvNV(SV*) ("double")
140 • Strings are a bit complicated:
142 • Byte string: "SvPVbyte(SV*, STRLEN len)" or SvPVbyte_nolen(SV*)
144 If the Perl string is "\xff\xff", then this returns a 2-byte
146 "char*".
147 This is suitable for Perl strings that represent bytes.
149 • UTF-8 string: "SvPVutf8(SV*, STRLEN len)" or
151 SvPVutf8_nolen(SV*)
152 If the Perl string is "\xff\xff", then this returns a 4-byte
154 "char*".
155 This is suitable for Perl strings that represent characters.
157 CAVEAT: That "char*" will be encoded via Perl's internal UTF-8
159 variant, which means that if the SV contains non-Unicode code
160 points (e.g., 0x110000), then the result may contain extensions
161 over valid UTF-8. See "is_strict_utf8_string" in perlapi for
162 some methods Perl gives you to check the UTF-8 validity of
163 these macros' returns.
164 • You can also use "SvPV(SV*, STRLEN len)" or SvPV_nolen(SV*) to
166 fetch the SV's raw internal buffer. This is tricky, though; if
167 your Perl string is "\xff\xff", then depending on the SV's
168 internal encoding you might get back a 2-byte OR a 4-byte
169 "char*". Moreover, if it's the 4-byte string, that could come
170 from either Perl "\xff\xff" stored UTF-8 encoded, or Perl
171 "\xc3\xbf\xc3\xbf" stored as raw octets. To differentiate
172 between these you MUST look up the SV's UTF8 bit (cf. "SvUTF8")
173 to know whether the source Perl string is 2 characters
174 ("SvUTF8" would be on) or 4 characters ("SvUTF8" would be off).
175 IMPORTANT: Use of "SvPV", "SvPV_nolen", or similarly-named
177 macros without looking up the SV's UTF8 bit is almost certainly
178 a bug if non-ASCII input is allowed.
179 When the UTF8 bit is on, the same CAVEAT about UTF-8 validity
181 applies here as for "SvPVutf8".
182 (See "How do I pass a Perl string to a C library?" for more
184 details.)
185 In "SvPVbyte", "SvPVutf8", and "SvPV", the length of the "char*"
187 returned is placed into the variable "len" (these are macros, so
188 you do not use &len). If you do not care what the length of the
189 data is, use "SvPVbyte_nolen", "SvPVutf8_nolen", or "SvPV_nolen"
190 instead. The global variable "PL_na" can also be given to
191 "SvPVbyte"/"SvPVutf8"/"SvPV" in this case. But that can be quite
192 inefficient because "PL_na" must be accessed in thread-local
193 storage in threaded Perl. In any case, remember that Perl allows
194 arbitrary strings of data that may both contain NULs and might not
195 be terminated by a "NUL".
196 Also remember that C doesn't allow you to safely say
198 "foo(SvPVbyte(s, len), len);". It might work with your compiler,
199 but it won't work for everyone. Break this sort of statement up
200 into separate assignments:
201 SV *s;
203 STRLEN len;
204 char *ptr;
205 ptr = SvPVbyte(s, len);
206 foo(ptr, len);
207 If you want to know if the scalar value is TRUE, you can use:
209 SvTRUE(SV*)
211 Although Perl will automatically grow strings for you, if you need to
213 force Perl to allocate more memory for your SV, you can use the macro
214 SvGROW(SV*, STRLEN newlen)
216 which will determine if more memory needs to be allocated. If so, it
218 will call the function "sv_grow". Note that "SvGROW" can only
219 increase, not decrease, the allocated memory of an SV and that it does
220 not automatically add space for the trailing "NUL" byte (perl's own
221 string functions typically do "SvGROW(sv, len + 1)").
222 If you want to write to an existing SV's buffer and set its value to a
224 string, use SvPVbyte_force() or one of its variants to force the SV to
225 be a PV. This will remove any of various types of non-stringness from
226 the SV while preserving the content of the SV in the PV. This can be
227 used, for example, to append data from an API function to a buffer
228 without extra copying:
229 (void)SvPVbyte_force(sv, len);
231 s = SvGROW(sv, len + needlen + 1);
232 /* something that modifies up to needlen bytes at s+len, but
233 modifies newlen bytes
234 eg. newlen = read(fd, s + len, needlen);
235 ignoring errors for these examples
236 */
237 s[len + newlen] = '\0';
238 SvCUR_set(sv, len + newlen);
239 SvUTF8_off(sv);
240 SvSETMAGIC(sv);
241 If you already have the data in memory or if you want to keep your code
243 simple, you can use one of the sv_cat*() variants, such as sv_catpvn().
244 If you want to insert anywhere in the string you can use sv_insert() or
245 sv_insert_flags().
246 If you don't need the existing content of the SV, you can avoid some
248 copying with:
249 SvPVCLEAR(sv);
251 s = SvGROW(sv, needlen + 1);
252 /* something that modifies up to needlen bytes at s, but modifies
253 newlen bytes
254 eg. newlen = read(fd, s, needlen);
255 */
256 s[newlen] = '\0';
257 SvCUR_set(sv, newlen);
258 SvPOK_only(sv); /* also clears SVf_UTF8 */
259 SvSETMAGIC(sv);
260 Again, if you already have the data in memory or want to avoid the
262 complexity of the above, you can use sv_setpvn().
263 If you have a buffer allocated with Newx() and want to set that as the
265 SV's value, you can use sv_usepvn_flags(). That has some requirements
266 if you want to avoid perl re-allocating the buffer to fit the trailing
267 NUL:
268 Newx(buf, somesize+1, char);
270 /* ... fill in buf ... */
271 buf[somesize] = '\0';
272 sv_usepvn_flags(sv, buf, somesize, SV_SMAGIC | SV_HAS_TRAILING_NUL);
273 /* buf now belongs to perl, don't release it */
274 If you have an SV and want to know what kind of data Perl thinks is
276 stored in it, you can use the following macros to check the type of SV
277 you have.
278 SvIOK(SV*)
280 SvNOK(SV*)
281 SvPOK(SV*)
282 Be aware that retrieving the numeric value of an SV can set IOK or NOK
284 on that SV, even when the SV started as a string. Prior to Perl 5.36.0
285 retrieving the string value of an integer could set POK, but this can
286 no longer occur. From 5.36.0 this can be used to distinguish the
287 original representation of an SV and is intended to make life simpler
288 for serializers:
289 /* references handled elsewhere */
291 if (SvIsBOOL(sv)) {
292 /* originally boolean */
293 ...
294 }
295 else if (SvPOK(sv)) {
296 /* originally a string */
297 ...
298 }
299 else if (SvNIOK(sv)) {
300 /* originally numeric */
301 ...
302 }
303 else {
304 /* something special or undef */
305 }
306 You can get and set the current length of the string stored in an SV
308 with the following macros:
309 SvCUR(SV*)
311 SvCUR_set(SV*, I32 val)
312 You can also get a pointer to the end of the string stored in the SV
314 with the macro:
315 SvEND(SV*)
317 But note that these last three macros are valid only if SvPOK() is
319 true.
320 If you want to append something to the end of string stored in an
322 "SV*", you can use the following functions:
323 void sv_catpv(SV*, const char*);
325 void sv_catpvn(SV*, const char*, STRLEN);
326 void sv_catpvf(SV*, const char*, ...);
327 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **,
328 I32, bool);
329 void sv_catsv(SV*, SV*);
330 The first function calculates the length of the string to be appended
332 by using "strlen". In the second, you specify the length of the string
333 yourself. The third function processes its arguments like "sprintf"
334 and appends the formatted output. The fourth function works like
335 "vsprintf". You can specify the address and length of an array of SVs
336 instead of the va_list argument. The fifth function extends the string
337 stored in the first SV with the string stored in the second SV. It
338 also forces the second SV to be interpreted as a string.
339 The "sv_cat*()" functions are not generic enough to operate on values
341 that have "magic". See "Magic Virtual Tables" later in this document.
342 If you know the name of a scalar variable, you can get a pointer to its
344 SV by using the following:
345 SV* get_sv("package::varname", 0);
347 This returns NULL if the variable does not exist.
349 If you want to know if this variable (or any other SV) is actually
351 "defined", you can call:
352 SvOK(SV*)
354 The scalar "undef" value is stored in an SV instance called
356 "PL_sv_undef".
357 Its address can be used whenever an "SV*" is needed. Make sure that
359 you don't try to compare a random sv with &PL_sv_undef. For example
360 when interfacing Perl code, it'll work correctly for:
361 foo(undef);
363 But won't work when called as:
365 $x = undef;
367 foo($x);
368 So to repeat always use SvOK() to check whether an sv is defined.
370 Also you have to be careful when using &PL_sv_undef as a value in AVs
372 or HVs (see "AVs, HVs and undefined values").
373 There are also the two values "PL_sv_yes" and "PL_sv_no", which contain
375 boolean TRUE and FALSE values, respectively. Like "PL_sv_undef", their
376 addresses can be used whenever an "SV*" is needed.
377 Do not be fooled into thinking that "(SV *) 0" is the same as
379 &PL_sv_undef. Take this code:
380 SV* sv = (SV*) 0;
382 if (I-am-to-return-a-real-value) {
383 sv = sv_2mortal(newSViv(42));
384 }
385 sv_setsv(ST(0), sv);
386 This code tries to return a new SV (which contains the value 42) if it
388 should return a real value, or undef otherwise. Instead it has
389 returned a NULL pointer which, somewhere down the line, will cause a
390 segmentation violation, bus error, or just weird results. Change the
391 zero to &PL_sv_undef in the first line and all will be well.
392 To free an SV that you've created, call SvREFCNT_dec(SV*). Normally
394 this call is not necessary (see "Reference Counts and Mortality").
395 Offsets
397 Perl provides the function "sv_chop" to efficiently remove characters
398 from the beginning of a string; you give it an SV and a pointer to
399 somewhere inside the PV, and it discards everything before the pointer.
400 The efficiency comes by means of a little hack: instead of actually
401 removing the characters, "sv_chop" sets the flag "OOK" (offset OK) to
402 signal to other functions that the offset hack is in effect, and it
403 moves the PV pointer (called "SvPVX") forward by the number of bytes
404 chopped off, and adjusts "SvCUR" and "SvLEN" accordingly. (A portion
405 of the space between the old and new PV pointers is used to store the
406 count of chopped bytes.)
407 Hence, at this point, the start of the buffer that we allocated lives
409 at "SvPVX(sv) - SvIV(sv)" in memory and the PV pointer is pointing into
410 the middle of this allocated storage.
411 This is best demonstrated by example. Normally copy-on-write will
413 prevent the substitution from operator from using this hack, but if you
414 can craft a string for which copy-on-write is not possible, you can see
415 it in play. In the current implementation, the final byte of a string
416 buffer is used as a copy-on-write reference count. If the buffer is
417 not big enough, then copy-on-write is skipped. First have a look at an
418 empty string:
419 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a .= ""; Dump $a'
421 SV = PV(0x7ffb7c008a70) at 0x7ffb7c030390
422 REFCNT = 1
423 FLAGS = (POK,pPOK)
424 PV = 0x7ffb7bc05b50 ""\0
425 CUR = 0
426 LEN = 10
427 Notice here the LEN is 10. (It may differ on your platform.) Extend
429 the length of the string to one less than 10, and do a substitution:
430 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a.="123456789"; $a=~s/.//; \
432 Dump($a)'
433 SV = PV(0x7ffa04008a70) at 0x7ffa04030390
434 REFCNT = 1
435 FLAGS = (POK,OOK,pPOK)
436 OFFSET = 1
437 PV = 0x7ffa03c05b61 ( "\1" . ) "23456789"\0
438 CUR = 8
439 LEN = 9
440 Here the number of bytes chopped off (1) is shown next as the OFFSET.
442 The portion of the string between the "real" and the "fake" beginnings
443 is shown in parentheses, and the values of "SvCUR" and "SvLEN" reflect
444 the fake beginning, not the real one. (The first character of the
445 string buffer happens to have changed to "\1" here, not "1", because
446 the current implementation stores the offset count in the string
447 buffer. This is subject to change.)
448 Something similar to the offset hack is performed on AVs to enable
450 efficient shifting and splicing off the beginning of the array; while
451 "AvARRAY" points to the first element in the array that is visible from
452 Perl, "AvALLOC" points to the real start of the C array. These are
453 usually the same, but a "shift" operation can be carried out by
454 increasing "AvARRAY" by one and decreasing "AvFILL" and "AvMAX".
455 Again, the location of the real start of the C array only comes into
456 play when freeing the array. See "av_shift" in av.c.
457 What's Really Stored in an SV?
459 Recall that the usual method of determining the type of scalar you have
460 is to use "Sv*OK" macros. Because a scalar can be both a number and a
461 string, usually these macros will always return TRUE and calling the
462 "Sv*V" macros will do the appropriate conversion of string to
463 integer/double or integer/double to string.
464 If you really need to know if you have an integer, double, or string
466 pointer in an SV, you can use the following three macros instead:
467 SvIOKp(SV*)
469 SvNOKp(SV*)
470 SvPOKp(SV*)
471 These will tell you if you truly have an integer, double, or string
473 pointer stored in your SV. The "p" stands for private.
474 There are various ways in which the private and public flags may
476 differ. For example, in perl 5.16 and earlier a tied SV may have a
477 valid underlying value in the IV slot (so SvIOKp is true), but the data
478 should be accessed via the FETCH routine rather than directly, so SvIOK
479 is false. (In perl 5.18 onwards, tied scalars use the flags the same
480 way as untied scalars.) Another is when numeric conversion has
481 occurred and precision has been lost: only the private flag is set on
482 'lossy' values. So when an NV is converted to an IV with loss, SvIOKp,
483 SvNOKp and SvNOK will be set, while SvIOK wont be.
484 In general, though, it's best to use the "Sv*V" macros.
486 Working with AVs
488 There are two main, longstanding ways to create and load an AV. The
489 first method creates an empty AV:
490 AV* newAV();
492 The second method both creates the AV and initially populates it with
494 SVs:
495 AV* av_make(SSize_t num, SV **ptr);
497 The second argument points to an array containing "num" "SV*"'s. Once
499 the AV has been created, the SVs can be destroyed, if so desired.
500 Perl v5.36 added two new ways to create an AV and allocate a SV** array
502 without populating it. These are more efficient than a newAV() followed
503 by an av_extend().
504 /* Creates but does not initialize (Zero) the SV** array */
506 AV *av = newAV_alloc_x(1);
507 /* Creates and does initialize (Zero) the SV** array */
508 AV *av = newAV_alloc_xz(1);
509 The numerical argument refers to the number of array elements to
511 allocate, not an array index, and must be >0. The first form must only
512 ever be used when all elements will be initialized before any read
513 occurs. Reading a non-initialized SV* - i.e. treating a random memory
514 address as a SV* - is a serious bug.
515 Once the AV has been created, the following operations are possible on
517 it:
518 void av_push(AV*, SV*);
520 SV* av_pop(AV*);
521 SV* av_shift(AV*);
522 void av_unshift(AV*, SSize_t num);
523 These should be familiar operations, with the exception of
525 "av_unshift". This routine adds "num" elements at the front of the
526 array with the "undef" value. You must then use "av_store" (described
527 below) to assign values to these new elements.
528 Here are some other functions:
530 SSize_t av_top_index(AV*);
532 SV** av_fetch(AV*, SSize_t key, I32 lval);
533 SV** av_store(AV*, SSize_t key, SV* val);
534 The "av_top_index" function returns the highest index value in an array
536 (just like $#array in Perl). If the array is empty, -1 is returned.
537 The "av_fetch" function returns the value at index "key", but if "lval"
538 is non-zero, then "av_fetch" will store an undef value at that index.
539 The "av_store" function stores the value "val" at index "key", and does
540 not increment the reference count of "val". Thus the caller is
541 responsible for taking care of that, and if "av_store" returns NULL,
542 the caller will have to decrement the reference count to avoid a memory
543 leak. Note that "av_fetch" and "av_store" both return "SV**"'s, not
544 "SV*"'s as their return value.
545 A few more:
547 void av_clear(AV*);
549 void av_undef(AV*);
550 void av_extend(AV*, SSize_t key);
551 The "av_clear" function deletes all the elements in the AV* array, but
553 does not actually delete the array itself. The "av_undef" function
554 will delete all the elements in the array plus the array itself. The
555 "av_extend" function extends the array so that it contains at least
556 "key+1" elements. If "key+1" is less than the currently allocated
557 length of the array, then nothing is done.
558 If you know the name of an array variable, you can get a pointer to its
560 AV by using the following:
561 AV* get_av("package::varname", 0);
563 This returns NULL if the variable does not exist.
565 See "Understanding the Magic of Tied Hashes and Arrays" for more
567 information on how to use the array access functions on tied arrays.
568 More efficient working with new or vanilla AVs
570 Perl v5.36 and v5.38 introduced streamlined, inlined versions of some
572 functions:
573 • "av_store_simple"
575 • "av_fetch_simple"
577 • "av_push_simple"
579 These are drop-in replacements, but can only be used on straightforward
581 AVs that meet the following criteria:
582 • are not magical
584 • are not readonly
586 • are "real" (refcounted) AVs
588 • have an av_top_index value > -2
590 AVs created using newAV(), "av_make", "newAV_alloc_x", and
592 "newAV_alloc_xz" are all compatible at the time of creation. It is only
593 if they are declared readonly or unreal, have magic attached, or are
594 otherwise configured unusually that they will stop being compatible.
595 Note that some interpreter functions may attach magic to an AV as part
597 of normal operations. It is therefore safest, unless you are sure of
598 the lifecycle of an AV, to only use these new functions close to the
599 point of AV creation.
600 Working with HVs
602 To create an HV, you use the following routine:
603 HV* newHV();
605 Once the HV has been created, the following operations are possible on
607 it:
608 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
610 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
611 The "klen" parameter is the length of the key being passed in (Note
613 that you cannot pass 0 in as a value of "klen" to tell Perl to measure
614 the length of the key). The "val" argument contains the SV pointer to
615 the scalar being stored, and "hash" is the precomputed hash value (zero
616 if you want "hv_store" to calculate it for you). The "lval" parameter
617 indicates whether this fetch is actually a part of a store operation,
618 in which case a new undefined value will be added to the HV with the
619 supplied key and "hv_fetch" will return as if the value had already
620 existed.
621 Remember that "hv_store" and "hv_fetch" return "SV**"'s and not just
623 "SV*". To access the scalar value, you must first dereference the
624 return value. However, you should check to make sure that the return
625 value is not NULL before dereferencing it.
626 The first of these two functions checks if a hash table entry exists,
628 and the second deletes it.
629 bool hv_exists(HV*, const char* key, U32 klen);
631 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
632 If "flags" does not include the "G_DISCARD" flag then "hv_delete" will
634 create and return a mortal copy of the deleted value.
635 And more miscellaneous functions:
637 void hv_clear(HV*);
639 void hv_undef(HV*);
640 Like their AV counterparts, "hv_clear" deletes all the entries in the
642 hash table but does not actually delete the hash table. The "hv_undef"
643 deletes both the entries and the hash table itself.
644 Perl keeps the actual data in a linked list of structures with a
646 typedef of HE. These contain the actual key and value pointers (plus
647 extra administrative overhead). The key is a string pointer; the value
648 is an "SV*". However, once you have an "HE*", to get the actual key
649 and value, use the routines specified below.
650 I32 hv_iterinit(HV*);
652 /* Prepares starting point to traverse hash table */
653 HE* hv_iternext(HV*);
654 /* Get the next entry, and return a pointer to a
655 structure that has both the key and value */
656 char* hv_iterkey(HE* entry, I32* retlen);
657 /* Get the key from an HE structure and also return
658 the length of the key string */
659 SV* hv_iterval(HV*, HE* entry);
660 /* Return an SV pointer to the value of the HE
661 structure */
662 SV* hv_iternextsv(HV*, char** key, I32* retlen);
663 /* This convenience routine combines hv_iternext,
664 hv_iterkey, and hv_iterval. The key and retlen
665 arguments are return values for the key and its
666 length. The value is returned in the SV* argument */
667 If you know the name of a hash variable, you can get a pointer to its
669 HV by using the following:
670 HV* get_hv("package::varname", 0);
672 This returns NULL if the variable does not exist.
674 The hash algorithm is defined in the "PERL_HASH" macro:
676 PERL_HASH(hash, key, klen)
678 The exact implementation of this macro varies by architecture and
680 version of perl, and the return value may change per invocation, so the
681 value is only valid for the duration of a single perl process.
682 See "Understanding the Magic of Tied Hashes and Arrays" for more
684 information on how to use the hash access functions on tied hashes.
685 Hash API Extensions
687 Beginning with version 5.004, the following functions are also
688 supported:
689 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
691 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
692 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
694 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
695 SV* hv_iterkeysv (HE* entry);
697 Note that these functions take "SV*" keys, which simplifies writing of
699 extension code that deals with hash structures. These functions also
700 allow passing of "SV*" keys to "tie" functions without forcing you to
701 stringify the keys (unlike the previous set of functions).
702 They also return and accept whole hash entries ("HE*"), making their
704 use more efficient (since the hash number for a particular string
705 doesn't have to be recomputed every time). See perlapi for detailed
706 descriptions.
707 The following macros must always be used to access the contents of hash
709 entries. Note that the arguments to these macros must be simple
710 variables, since they may get evaluated more than once. See perlapi
711 for detailed descriptions of these macros.
712 HePV(HE* he, STRLEN len)
714 HeVAL(HE* he)
715 HeHASH(HE* he)
716 HeSVKEY(HE* he)
717 HeSVKEY_force(HE* he)
718 HeSVKEY_set(HE* he, SV* sv)
719 These two lower level macros are defined, but must only be used when
721 dealing with keys that are not "SV*"s:
722 HeKEY(HE* he)
724 HeKLEN(HE* he)
725 Note that both "hv_store" and "hv_store_ent" do not increment the
727 reference count of the stored "val", which is the caller's
728 responsibility. If these functions return a NULL value, the caller
729 will usually have to decrement the reference count of "val" to avoid a
730 memory leak.
731 AVs, HVs and undefined values
733 Sometimes you have to store undefined values in AVs or HVs. Although
734 this may be a rare case, it can be tricky. That's because you're used
735 to using &PL_sv_undef if you need an undefined SV.
736 For example, intuition tells you that this XS code:
738 AV *av = newAV();
740 av_store( av, 0, &PL_sv_undef );
741 is equivalent to this Perl code:
743 my @av;
745 $av[0] = undef;
746 Unfortunately, this isn't true. In perl 5.18 and earlier, AVs use
748 &PL_sv_undef as a marker for indicating that an array element has not
749 yet been initialized. Thus, "exists $av[0]" would be true for the
750 above Perl code, but false for the array generated by the XS code. In
751 perl 5.20, storing &PL_sv_undef will create a read-only element,
752 because the scalar &PL_sv_undef itself is stored, not a copy.
753 Similar problems can occur when storing &PL_sv_undef in HVs:
755 hv_store( hv, "key", 3, &PL_sv_undef, 0 );
757 This will indeed make the value "undef", but if you try to modify the
759 value of "key", you'll get the following error:
760 Modification of non-creatable hash value attempted
762 In perl 5.8.0, &PL_sv_undef was also used to mark placeholders in
764 restricted hashes. This caused such hash entries not to appear when
765 iterating over the hash or when checking for the keys with the
766 "hv_exists" function.
767 You can run into similar problems when you store &PL_sv_yes or
769 &PL_sv_no into AVs or HVs. Trying to modify such elements will give
770 you the following error:
771 Modification of a read-only value attempted
773 To make a long story short, you can use the special variables
775 &PL_sv_undef, &PL_sv_yes and &PL_sv_no with AVs and HVs, but you have
776 to make sure you know what you're doing.
777 Generally, if you want to store an undefined value in an AV or HV, you
779 should not use &PL_sv_undef, but rather create a new undefined value
780 using the "newSV" function, for example:
781 av_store( av, 42, newSV(0) );
783 hv_store( hv, "foo", 3, newSV(0), 0 );
784 References
786 References are a special type of scalar that point to other data types
787 (including other references).
788 To create a reference, use either of the following functions:
790 SV* newRV_inc((SV*) thing);
792 SV* newRV_noinc((SV*) thing);
793 The "thing" argument can be any of an "SV*", "AV*", or "HV*". The
795 functions are identical except that "newRV_inc" increments the
796 reference count of the "thing", while "newRV_noinc" does not. For
797 historical reasons, "newRV" is a synonym for "newRV_inc".
798 Once you have a reference, you can use the following macro to
800 dereference the reference:
801 SvRV(SV*)
803 then call the appropriate routines, casting the returned "SV*" to
805 either an "AV*" or "HV*", if required.
806 To determine if an SV is a reference, you can use the following macro:
808 SvROK(SV*)
810 To discover what type of value the reference refers to, use the
812 following macro and then check the return value.
813 SvTYPE(SvRV(SV*))
815 The most useful types that will be returned are:
817 SVt_PVAV Array
819 SVt_PVHV Hash
820 SVt_PVCV Code
821 SVt_PVGV Glob (possibly a file handle)
822 Any numerical value returned which is less than SVt_PVAV will be a
824 scalar of some form.
825 See "svtype" in perlapi for more details.
827 Blessed References and Class Objects
829 References are also used to support object-oriented programming. In
830 perl's OO lexicon, an object is simply a reference that has been
831 blessed into a package (or class). Once blessed, the programmer may
832 now use the reference to access the various methods in the class.
833 A reference can be blessed into a package with the following function:
835 SV* sv_bless(SV* sv, HV* stash);
837 The "sv" argument must be a reference value. The "stash" argument
839 specifies which class the reference will belong to. See "Stashes and
840 Globs" for information on converting class names into stashes.
841 /* Still under construction */
843 The following function upgrades rv to reference if not already one.
845 Creates a new SV for rv to point to. If "classname" is non-null, the
846 SV is blessed into the specified class. SV is returned.
847 SV* newSVrv(SV* rv, const char* classname);
849 The following three functions copy integer, unsigned integer or double
851 into an SV whose reference is "rv". SV is blessed if "classname" is
852 non-null.
853 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
855 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
856 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
857 The following function copies the pointer value (the address, not the
859 string!) into an SV whose reference is rv. SV is blessed if
860 "classname" is non-null.
861 SV* sv_setref_pv(SV* rv, const char* classname, void* pv);
863 The following function copies a string into an SV whose reference is
865 "rv". Set length to 0 to let Perl calculate the string length. SV is
866 blessed if "classname" is non-null.
867 SV* sv_setref_pvn(SV* rv, const char* classname, char* pv,
869 STRLEN length);
870 The following function tests whether the SV is blessed into the
872 specified class. It does not check inheritance relationships.
873 int sv_isa(SV* sv, const char* name);
875 The following function tests whether the SV is a reference to a blessed
877 object.
878 int sv_isobject(SV* sv);
880 The following function tests whether the SV is derived from the
882 specified class. SV can be either a reference to a blessed object or a
883 string containing a class name. This is the function implementing the
884 "UNIVERSAL::isa" functionality.
885 bool sv_derived_from(SV* sv, const char* name);
887 To check if you've got an object derived from a specific class you have
889 to write:
890 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
892 Creating New Variables
894 To create a new Perl variable with an undef value which can be accessed
895 from your Perl script, use the following routines, depending on the
896 variable type.
897 SV* get_sv("package::varname", GV_ADD);
899 AV* get_av("package::varname", GV_ADD);
900 HV* get_hv("package::varname", GV_ADD);
901 Notice the use of GV_ADD as the second parameter. The new variable can
903 now be set, using the routines appropriate to the data type.
904 There are additional macros whose values may be bitwise OR'ed with the
906 "GV_ADD" argument to enable certain extra features. Those bits are:
907 GV_ADDMULTI
909 Marks the variable as multiply defined, thus preventing the:
910 Name <varname> used only once: possible typo
912 warning.
914 GV_ADDWARN
916 Issues the warning:
917 Had to create <varname> unexpectedly
919 if the variable did not exist before the function was called.
921 If you do not specify a package name, the variable is created in the
923 current package.
924 Reference Counts and Mortality
926 Perl uses a reference count-driven garbage collection mechanism. SVs,
927 AVs, or HVs (xV for short in the following) start their life with a
928 reference count of 1. If the reference count of an xV ever drops to 0,
929 then it will be destroyed and its memory made available for reuse. At
930 the most basic internal level, reference counts can be manipulated with
931 the following macros:
932 int SvREFCNT(SV* sv);
934 SV* SvREFCNT_inc(SV* sv);
935 void SvREFCNT_dec(SV* sv);
936 (There are also suffixed versions of the increment and decrement
938 macros, for situations where the full generality of these basic macros
939 can be exchanged for some performance.)
940 However, the way a programmer should think about references is not so
942 much in terms of the bare reference count, but in terms of ownership of
943 references. A reference to an xV can be owned by any of a variety of
944 entities: another xV, the Perl interpreter, an XS data structure, a
945 piece of running code, or a dynamic scope. An xV generally does not
946 know what entities own the references to it; it only knows how many
947 references there are, which is the reference count.
948 To correctly maintain reference counts, it is essential to keep track
950 of what references the XS code is manipulating. The programmer should
951 always know where a reference has come from and who owns it, and be
952 aware of any creation or destruction of references, and any transfers
953 of ownership. Because ownership isn't represented explicitly in the xV
954 data structures, only the reference count need be actually maintained
955 by the code, and that means that this understanding of ownership is not
956 actually evident in the code. For example, transferring ownership of a
957 reference from one owner to another doesn't change the reference count
958 at all, so may be achieved with no actual code. (The transferring code
959 doesn't touch the referenced object, but does need to ensure that the
960 former owner knows that it no longer owns the reference, and that the
961 new owner knows that it now does.)
962 An xV that is visible at the Perl level should not become unreferenced
964 and thus be destroyed. Normally, an object will only become
965 unreferenced when it is no longer visible, often by the same means that
966 makes it invisible. For example, a Perl reference value (RV) owns a
967 reference to its referent, so if the RV is overwritten that reference
968 gets destroyed, and the no-longer-reachable referent may be destroyed
969 as a result.
970 Many functions have some kind of reference manipulation as part of
972 their purpose. Sometimes this is documented in terms of ownership of
973 references, and sometimes it is (less helpfully) documented in terms of
974 changes to reference counts. For example, the newRV_inc() function is
975 documented to create a new RV (with reference count 1) and increment
976 the reference count of the referent that was supplied by the caller.
977 This is best understood as creating a new reference to the referent,
978 which is owned by the created RV, and returning to the caller ownership
979 of the sole reference to the RV. The newRV_noinc() function instead
980 does not increment the reference count of the referent, but the RV
981 nevertheless ends up owning a reference to the referent. It is
982 therefore implied that the caller of newRV_noinc() is relinquishing a
983 reference to the referent, making this conceptually a more complicated
984 operation even though it does less to the data structures.
985 For example, imagine you want to return a reference from an XSUB
987 function. Inside the XSUB routine, you create an SV which initially
988 has just a single reference, owned by the XSUB routine. This reference
989 needs to be disposed of before the routine is complete, otherwise it
990 will leak, preventing the SV from ever being destroyed. So to create
991 an RV referencing the SV, it is most convenient to pass the SV to
992 newRV_noinc(), which consumes that reference. Now the XSUB routine no
993 longer owns a reference to the SV, but does own a reference to the RV,
994 which in turn owns a reference to the SV. The ownership of the
995 reference to the RV is then transferred by the process of returning the
996 RV from the XSUB.
997 There are some convenience functions available that can help with the
999 destruction of xVs. These functions introduce the concept of
1000 "mortality". Much documentation speaks of an xV itself being mortal,
1001 but this is misleading. It is really a reference to an xV that is
1002 mortal, and it is possible for there to be more than one mortal
1003 reference to a single xV. For a reference to be mortal means that it
1004 is owned by the temps stack, one of perl's many internal stacks, which
1005 will destroy that reference "a short time later". Usually the "short
1006 time later" is the end of the current Perl statement. However, it gets
1007 more complicated around dynamic scopes: there can be multiple sets of
1008 mortal references hanging around at the same time, with different death
1009 dates. Internally, the actual determinant for when mortal xV
1010 references are destroyed depends on two macros, SAVETMPS and FREETMPS.
1011 See perlcall and perlxs and "Temporaries Stack" below for more details
1012 on these macros.
1013 Mortal references are mainly used for xVs that are placed on perl's
1015 main stack. The stack is problematic for reference tracking, because
1016 it contains a lot of xV references, but doesn't own those references:
1017 they are not counted. Currently, there are many bugs resulting from
1018 xVs being destroyed while referenced by the stack, because the stack's
1019 uncounted references aren't enough to keep the xVs alive. So when
1020 putting an (uncounted) reference on the stack, it is vitally important
1021 to ensure that there will be a counted reference to the same xV that
1022 will last at least as long as the uncounted reference. But it's also
1023 important that that counted reference be cleaned up at an appropriate
1024 time, and not unduly prolong the xV's life. For there to be a mortal
1025 reference is often the best way to satisfy this requirement, especially
1026 if the xV was created especially to be put on the stack and would
1027 otherwise be unreferenced.
1028 To create a mortal reference, use the functions:
1030 SV* sv_newmortal()
1032 SV* sv_mortalcopy(SV*)
1033 SV* sv_2mortal(SV*)
1034 sv_newmortal() creates an SV (with the undefined value) whose sole
1036 reference is mortal. sv_mortalcopy() creates an xV whose value is a
1037 copy of a supplied xV and whose sole reference is mortal. sv_2mortal()
1038 mortalises an existing xV reference: it transfers ownership of a
1039 reference from the caller to the temps stack. Because "sv_newmortal"
1040 gives the new SV no value, it must normally be given one via
1041 "sv_setpv", "sv_setiv", etc. :
1042 SV *tmp = sv_newmortal();
1044 sv_setiv(tmp, an_integer);
1045 As that is multiple C statements it is quite common so see this idiom
1047 instead:
1048 SV *tmp = sv_2mortal(newSViv(an_integer));
1050 The mortal routines are not just for SVs; AVs and HVs can be made
1052 mortal by passing their address (type-casted to "SV*") to the
1053 "sv_2mortal" or "sv_mortalcopy" routines.
1054 Stashes and Globs
1056 A stash is a hash that contains all variables that are defined within a
1057 package. Each key of the stash is a symbol name (shared by all the
1058 different types of objects that have the same name), and each value in
1059 the hash table is a GV (Glob Value). This GV in turn contains
1060 references to the various objects of that name, including (but not
1061 limited to) the following:
1062 Scalar Value
1064 Array Value
1065 Hash Value
1066 I/O Handle
1067 Format
1068 Subroutine
1069 There is a single stash called "PL_defstash" that holds the items that
1071 exist in the "main" package. To get at the items in other packages,
1072 append the string "::" to the package name. The items in the "Foo"
1073 package are in the stash "Foo::" in PL_defstash. The items in the
1074 "Bar::Baz" package are in the stash "Baz::" in "Bar::"'s stash.
1075 To get the stash pointer for a particular package, use the function:
1077 HV* gv_stashpv(const char* name, I32 flags)
1079 HV* gv_stashsv(SV*, I32 flags)
1080 The first function takes a literal string, the second uses the string
1082 stored in the SV. Remember that a stash is just a hash table, so you
1083 get back an "HV*". The "flags" flag will create a new package if it is
1084 set to GV_ADD.
1085 The name that "gv_stash*v" wants is the name of the package whose
1087 symbol table you want. The default package is called "main". If you
1088 have multiply nested packages, pass their names to "gv_stash*v",
1089 separated by "::" as in the Perl language itself.
1090 Alternately, if you have an SV that is a blessed reference, you can
1092 find out the stash pointer by using:
1093 HV* SvSTASH(SvRV(SV*));
1095 then use the following to get the package name itself:
1097 char* HvNAME(HV* stash);
1099 If you need to bless or re-bless an object you can use the following
1101 function:
1102 SV* sv_bless(SV*, HV* stash)
1104 where the first argument, an "SV*", must be a reference, and the second
1106 argument is a stash. The returned "SV*" can now be used in the same
1107 way as any other SV.
1108 For more information on references and blessings, consult perlref.
1110 I/O Handles
1112 Like AVs and HVs, IO objects are another type of non-scalar SV which
1113 may contain input and output PerlIO objects or a "DIR *" from
1114 opendir().
1115 You can create a new IO object:
1117 IO* newIO();
1119 Unlike other SVs, a new IO object is automatically blessed into the
1121 IO::File class.
1122 The IO object contains an input and output PerlIO handle:
1124 PerlIO *IoIFP(IO *io);
1126 PerlIO *IoOFP(IO *io);
1127 Typically if the IO object has been opened on a file, the input handle
1129 is always present, but the output handle is only present if the file is
1130 open for output. For a file, if both are present they will be the same
1131 PerlIO object.
1132 Distinct input and output PerlIO objects are created for sockets and
1134 character devices.
1135 The IO object also contains other data associated with Perl I/O
1137 handles:
1138 IV IoLINES(io); /* $. */
1140 IV IoPAGE(io); /* $% */
1141 IV IoPAGE_LEN(io); /* $= */
1142 IV IoLINES_LEFT(io); /* $- */
1143 char *IoTOP_NAME(io); /* $^ */
1144 GV *IoTOP_GV(io); /* $^ */
1145 char *IoFMT_NAME(io); /* $~ */
1146 GV *IoFMT_GV(io); /* $~ */
1147 char *IoBOTTOM_NAME(io);
1148 GV *IoBOTTOM_GV(io);
1149 char IoTYPE(io);
1150 U8 IoFLAGS(io);
1151 =for apidoc_sections $io_scn, $formats_section
1153 =for apidoc_section $reports
1154 =for apidoc Amh|IV|IoLINES|IO *io
1155 =for apidoc Amh|IV|IoPAGE|IO *io
1156 =for apidoc Amh|IV|IoPAGE_LEN|IO *io
1157 =for apidoc Amh|IV|IoLINES_LEFT|IO *io
1158 =for apidoc Amh|char *|IoTOP_NAME|IO *io
1159 =for apidoc Amh|GV *|IoTOP_GV|IO *io
1160 =for apidoc Amh|char *|IoFMT_NAME|IO *io
1161 =for apidoc Amh|GV *|IoFMT_GV|IO *io
1162 =for apidoc Amh|char *|IoBOTTOM_NAME|IO *io
1163 =for apidoc Amh|GV *|IoBOTTOM_GV|IO *io
1164 =for apidoc_section $io
1165 =for apidoc Amh|char|IoTYPE|IO *io
1166 =for apidoc Amh|U8|IoFLAGS|IO *io
1167 Most of these are involved with formats.
1169 IoFLAGs() may contain a combination of flags, the most interesting of
1171 which are "IOf_FLUSH" ($|) for autoflush and "IOf_UNTAINT", settable
1172 with IO::Handle's untaint() method.
1173 The IO object may also contains a directory handle:
1175 DIR *IoDIRP(io);
1177 suitable for use with PerlDir_read() etc.
1179 All of these accessors macros are lvalues, there are no distinct _set()
1181 macros to modify the members of the IO object.
1182 Double-Typed SVs
1184 Scalar variables normally contain only one type of value, an integer,
1185 double, pointer, or reference. Perl will automatically convert the
1186 actual scalar data from the stored type into the requested type.
1187 Some scalar variables contain more than one type of scalar data. For
1189 example, the variable $! contains either the numeric value of "errno"
1190 or its string equivalent from either "strerror" or "sys_errlist[]".
1191 To force multiple data values into an SV, you must do two things: use
1193 the "sv_set*v" routines to add the additional scalar type, then set a
1194 flag so that Perl will believe it contains more than one type of data.
1195 The four macros to set the flags are:
1196 SvIOK_on
1198 SvNOK_on
1199 SvPOK_on
1200 SvROK_on
1201 The particular macro you must use depends on which "sv_set*v" routine
1203 you called first. This is because every "sv_set*v" routine turns on
1204 only the bit for the particular type of data being set, and turns off
1205 all the rest.
1206 For example, to create a new Perl variable called "dberror" that
1208 contains both the numeric and descriptive string error values, you
1209 could use the following code:
1210 extern int dberror;
1212 extern char *dberror_list;
1213 SV* sv = get_sv("dberror", GV_ADD);
1215 sv_setiv(sv, (IV) dberror);
1216 sv_setpv(sv, dberror_list[dberror]);
1217 SvIOK_on(sv);
1218 If the order of "sv_setiv" and "sv_setpv" had been reversed, then the
1220 macro "SvPOK_on" would need to be called instead of "SvIOK_on".
1221 Read-Only Values
1223 In Perl 5.16 and earlier, copy-on-write (see the next section) shared a
1224 flag bit with read-only scalars. So the only way to test whether
1225 "sv_setsv", etc., will raise a "Modification of a read-only value"
1226 error in those versions is:
1227 SvREADONLY(sv) && !SvIsCOW(sv)
1229 Under Perl 5.18 and later, SvREADONLY only applies to read-only
1231 variables, and, under 5.20, copy-on-write scalars can also be read-
1232 only, so the above check is incorrect. You just want:
1233 SvREADONLY(sv)
1235 If you need to do this check often, define your own macro like this:
1237 #if PERL_VERSION >= 18
1239 # define SvTRULYREADONLY(sv) SvREADONLY(sv)
1240 #else
1241 # define SvTRULYREADONLY(sv) (SvREADONLY(sv) && !SvIsCOW(sv))
1242 #endif
1243 Copy on Write
1245 Perl implements a copy-on-write (COW) mechanism for scalars, in which
1246 string copies are not immediately made when requested, but are deferred
1247 until made necessary by one or the other scalar changing. This is
1248 mostly transparent, but one must take care not to modify string buffers
1249 that are shared by multiple SVs.
1250 You can test whether an SV is using copy-on-write with SvIsCOW(sv).
1252 You can force an SV to make its own copy of its string buffer by
1254 calling sv_force_normal(sv) or SvPV_force_nolen(sv).
1255 If you want to make the SV drop its string buffer, use
1257 "sv_force_normal_flags(sv, SV_COW_DROP_PV)" or simply "sv_setsv(sv,
1258 NULL)".
1259 All of these functions will croak on read-only scalars (see the
1261 previous section for more on those).
1262 To test that your code is behaving correctly and not modifying COW
1264 buffers, on systems that support mmap(2) (i.e., Unix) you can configure
1265 perl with "-Accflags=-DPERL_DEBUG_READONLY_COW" and it will turn buffer
1266 violations into crashes. You will find it to be marvellously slow, so
1267 you may want to skip perl's own tests.
1268 Magic Variables
1270 [This section still under construction. Ignore everything here. Post
1271 no bills. Everything not permitted is forbidden.]
1272 Any SV may be magical, that is, it has special features that a normal
1274 SV does not have. These features are stored in the SV structure in a
1275 linked list of "struct magic"'s, typedef'ed to "MAGIC".
1276 struct magic {
1278 MAGIC* mg_moremagic;
1279 MGVTBL* mg_virtual;
1280 U16 mg_private;
1281 char mg_type;
1282 U8 mg_flags;
1283 I32 mg_len;
1284 SV* mg_obj;
1285 char* mg_ptr;
1286 };
1287 Note this is current as of patchlevel 0, and could change at any time.
1289 Assigning Magic
1291 Perl adds magic to an SV using the sv_magic function:
1292 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
1294 The "sv" argument is a pointer to the SV that is to acquire a new
1296 magical feature.
1297 If "sv" is not already magical, Perl uses the "SvUPGRADE" macro to
1299 convert "sv" to type "SVt_PVMG". Perl then continues by adding new
1300 magic to the beginning of the linked list of magical features. Any
1301 prior entry of the same type of magic is deleted. Note that this can
1302 be overridden, and multiple instances of the same type of magic can be
1303 associated with an SV.
1304 The "name" and "namlen" arguments are used to associate a string with
1306 the magic, typically the name of a variable. "namlen" is stored in the
1307 "mg_len" field and if "name" is non-null then either a "savepvn" copy
1308 of "name" or "name" itself is stored in the "mg_ptr" field, depending
1309 on whether "namlen" is greater than zero or equal to zero respectively.
1310 As a special case, if "(name && namlen == HEf_SVKEY)" then "name" is
1311 assumed to contain an "SV*" and is stored as-is with its REFCNT
1312 incremented.
1313 The sv_magic function uses "how" to determine which, if any, predefined
1315 "Magic Virtual Table" should be assigned to the "mg_virtual" field.
1316 See the "Magic Virtual Tables" section below. The "how" argument is
1317 also stored in the "mg_type" field. The value of "how" should be
1318 chosen from the set of macros "PERL_MAGIC_foo" found in perl.h. Note
1319 that before these macros were added, Perl internals used to directly
1320 use character literals, so you may occasionally come across old code or
1321 documentation referring to 'U' magic rather than "PERL_MAGIC_uvar" for
1322 example.
1323 The "obj" argument is stored in the "mg_obj" field of the "MAGIC"
1325 structure. If it is not the same as the "sv" argument, the reference
1326 count of the "obj" object is incremented. If it is the same, or if the
1327 "how" argument is "PERL_MAGIC_arylen", "PERL_MAGIC_regdatum",
1328 "PERL_MAGIC_regdata", or if it is a NULL pointer, then "obj" is merely
1329 stored, without the reference count being incremented.
1330 See also "sv_magicext" in perlapi for a more flexible way to add magic
1332 to an SV.
1333 There is also a function to add magic to an "HV":
1335 void hv_magic(HV *hv, GV *gv, int how);
1337 This simply calls "sv_magic" and coerces the "gv" argument into an
1339 "SV".
1340 To remove the magic from an SV, call the function sv_unmagic:
1342 int sv_unmagic(SV *sv, int type);
1344 The "type" argument should be equal to the "how" value when the "SV"
1346 was initially made magical.
1347 However, note that "sv_unmagic" removes all magic of a certain "type"
1349 from the "SV". If you want to remove only certain magic of a "type"
1350 based on the magic virtual table, use "sv_unmagicext" instead:
1351 int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);
1353 Magic Virtual Tables
1355 The "mg_virtual" field in the "MAGIC" structure is a pointer to an
1356 "MGVTBL", which is a structure of function pointers and stands for
1357 "Magic Virtual Table" to handle the various operations that might be
1358 applied to that variable.
1359 The "MGVTBL" has five (or sometimes eight) pointers to the following
1361 routine types:
1362 int (*svt_get) (pTHX_ SV* sv, MAGIC* mg);
1364 int (*svt_set) (pTHX_ SV* sv, MAGIC* mg);
1365 U32 (*svt_len) (pTHX_ SV* sv, MAGIC* mg);
1366 int (*svt_clear)(pTHX_ SV* sv, MAGIC* mg);
1367 int (*svt_free) (pTHX_ SV* sv, MAGIC* mg);
1368 int (*svt_copy) (pTHX_ SV *sv, MAGIC* mg, SV *nsv,
1370 const char *name, I32 namlen);
1371 int (*svt_dup) (pTHX_ MAGIC *mg, CLONE_PARAMS *param);
1372 int (*svt_local)(pTHX_ SV *nsv, MAGIC *mg);
1373 This MGVTBL structure is set at compile-time in perl.h and there are
1375 currently 32 types. These different structures contain pointers to
1376 various routines that perform additional actions depending on which
1377 function is being called.
1378 Function pointer Action taken
1380 ---------------- ------------
1381 svt_get Do something before the value of the SV is
1382 retrieved.
1383 svt_set Do something after the SV is assigned a value.
1384 svt_len Report on the SV's length.
1385 svt_clear Clear something the SV represents.
1386 svt_free Free any extra storage associated with the SV.
1387 svt_copy copy tied variable magic to a tied element
1389 svt_dup duplicate a magic structure during thread cloning
1390 svt_local copy magic to local value during 'local'
1391 For instance, the MGVTBL structure called "vtbl_sv" (which corresponds
1393 to an "mg_type" of "PERL_MAGIC_sv") contains:
1394 { magic_get, magic_set, magic_len, 0, 0 }
1396 Thus, when an SV is determined to be magical and of type
1398 "PERL_MAGIC_sv", if a get operation is being performed, the routine
1399 "magic_get" is called. All the various routines for the various
1400 magical types begin with "magic_". NOTE: the magic routines are not
1401 considered part of the Perl API, and may not be exported by the Perl
1402 library.
1403 The last three slots are a recent addition, and for source code
1405 compatibility they are only checked for if one of the three flags
1406 "MGf_COPY", "MGf_DUP", or "MGf_LOCAL" is set in mg_flags. This means
1407 that most code can continue declaring a vtable as a 5-element value.
1408 These three are currently used exclusively by the threading code, and
1409 are highly subject to change.
1410 The current kinds of Magic Virtual Tables are:
1412 mg_type
1414 (old-style char and macro) MGVTBL Type of magic
1415 -------------------------- ------ -------------
1416 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
1417 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
1418 % PERL_MAGIC_rhash (none) Extra data for restricted
1419 hashes
1420 * PERL_MAGIC_debugvar vtbl_debugvar $DB::single, signal, trace
1421 vars
1422 . PERL_MAGIC_pos vtbl_pos pos() lvalue
1423 : PERL_MAGIC_symtab (none) Extra data for symbol
1424 tables
1425 < PERL_MAGIC_backref vtbl_backref For weak ref data
1426 @ PERL_MAGIC_arylen_p (none) To move arylen out of XPVAV
1427 B PERL_MAGIC_bm vtbl_regexp Boyer-Moore
1428 (fast string search)
1429 c PERL_MAGIC_overload_table vtbl_ovrld Holds overload table
1430 (AMT) on stash
1431 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
1432 (@+ and @- vars)
1433 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
1434 element
1435 E PERL_MAGIC_env vtbl_env %ENV hash
1436 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
1437 f PERL_MAGIC_fm vtbl_regexp Formline
1438 ('compiled' format)
1439 g PERL_MAGIC_regex_global vtbl_mglob m//g target
1440 H PERL_MAGIC_hints vtbl_hints %^H hash
1441 h PERL_MAGIC_hintselem vtbl_hintselem %^H hash element
1442 I PERL_MAGIC_isa vtbl_isa @ISA array
1443 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
1444 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
1445 L PERL_MAGIC_dbfile (none) Debugger %_<filename
1446 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename
1447 element
1448 N PERL_MAGIC_shared (none) Shared between threads
1449 n PERL_MAGIC_shared_scalar (none) Shared between threads
1450 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale transformation
1451 P PERL_MAGIC_tied vtbl_pack Tied array or hash
1452 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
1453 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
1454 r PERL_MAGIC_qr vtbl_regexp Precompiled qr// regex
1455 S PERL_MAGIC_sig vtbl_sig %SIG hash
1456 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
1457 t PERL_MAGIC_taint vtbl_taint Taintedness
1458 U PERL_MAGIC_uvar vtbl_uvar Available for use by
1459 extensions
1460 u PERL_MAGIC_uvar_elem (none) Reserved for use by
1461 extensions
1462 V PERL_MAGIC_vstring (none) SV was vstring literal
1463 v PERL_MAGIC_vec vtbl_vec vec() lvalue
1464 w PERL_MAGIC_utf8 vtbl_utf8 Cached UTF-8 information
1465 X PERL_MAGIC_destruct vtbl_destruct destruct callback
1466 x PERL_MAGIC_substr vtbl_substr substr() lvalue
1467 Y PERL_MAGIC_nonelem vtbl_nonelem Array element that does not
1468 exist
1469 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
1470 variable / smart parameter
1471 vivification
1472 Z PERL_MAGIC_hook vtbl_hook %{^HOOK} hash
1473 z PERL_MAGIC_hookelem vtbl_hookelem %{^HOOK} hash element
1474 \ PERL_MAGIC_lvref vtbl_lvref Lvalue reference
1475 constructor
1476 ] PERL_MAGIC_checkcall vtbl_checkcall Inlining/mutation of call
1477 to this CV
1478 ^ PERL_MAGIC_extvalue (none) Value magic available for
1479 use by extensions
1480 ~ PERL_MAGIC_ext (none) Variable magic available
1481 for use by extensions
1482 When an uppercase and lowercase letter both exist in the table, then
1484 the uppercase letter is typically used to represent some kind of
1485 composite type (a list or a hash), and the lowercase letter is used to
1486 represent an element of that composite type. Some internals code makes
1487 use of this case relationship. However, 'v' and 'V' (vec and v-string)
1488 are in no way related.
1489 The "PERL_MAGIC_ext", "PERL_MAGIC_extvalue" and "PERL_MAGIC_uvar" magic
1491 types are defined specifically for use by extensions and will not be
1492 used by perl itself. Extensions can use "PERL_MAGIC_ext" or
1493 "PERL_MAGIC_extvalue" magic to 'attach' private information to
1494 variables (typically objects). This is especially useful because there
1495 is no way for normal perl code to corrupt this private information
1496 (unlike using extra elements of a hash object). "PERL_MAGIC_extvalue"
1497 is value magic (unlike "PERL_MAGIC_ext" and "PERL_MAGIC_uvar") meaning
1498 that on localization the new value will not be magical.
1499 Similarly, "PERL_MAGIC_uvar" magic can be used much like tie() to call
1501 a C function any time a scalar's value is used or changed. The
1502 "MAGIC"'s "mg_ptr" field points to a "ufuncs" structure:
1503 struct ufuncs {
1505 I32 (*uf_val)(pTHX_ IV, SV*);
1506 I32 (*uf_set)(pTHX_ IV, SV*);
1507 IV uf_index;
1508 };
1509 When the SV is read from or written to, the "uf_val" or "uf_set"
1511 function will be called with "uf_index" as the first arg and a pointer
1512 to the SV as the second. A simple example of how to add
1513 "PERL_MAGIC_uvar" magic is shown below. Note that the ufuncs structure
1514 is copied by sv_magic, so you can safely allocate it on the stack.
1515 void
1517 Umagic(sv)
1518 SV *sv;
1519 PREINIT:
1520 struct ufuncs uf;
1521 CODE:
1522 uf.uf_val = &my_get_fn;
1523 uf.uf_set = &my_set_fn;
1524 uf.uf_index = 0;
1525 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
1526 Attaching "PERL_MAGIC_uvar" to arrays is permissible but has no effect.
1528 For hashes there is a specialized hook that gives control over hash
1530 keys (but not values). This hook calls "PERL_MAGIC_uvar" 'get' magic
1531 if the "set" function in the "ufuncs" structure is NULL. The hook is
1532 activated whenever the hash is accessed with a key specified as an "SV"
1533 through the functions "hv_store_ent", "hv_fetch_ent", "hv_delete_ent",
1534 and "hv_exists_ent". Accessing the key as a string through the
1535 functions without the "..._ent" suffix circumvents the hook. See
1536 "GUTS" in Hash::Util::FieldHash for a detailed description.
1537 Note that because multiple extensions may be using "PERL_MAGIC_ext" or
1539 "PERL_MAGIC_uvar" magic, it is important for extensions to take extra
1540 care to avoid conflict. Typically only using the magic on objects
1541 blessed into the same class as the extension is sufficient. For
1542 "PERL_MAGIC_ext" magic, it is usually a good idea to define an
1543 "MGVTBL", even if all its fields will be 0, so that individual "MAGIC"
1544 pointers can be identified as a particular kind of magic using their
1545 magic virtual table. "mg_findext" provides an easy way to do that:
1546 STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };
1548 MAGIC *mg;
1550 if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
1551 /* this is really ours, not another module's PERL_MAGIC_ext */
1552 my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
1553 ...
1554 }
1555 Also note that the "sv_set*()" and "sv_cat*()" functions described
1557 earlier do not invoke 'set' magic on their targets. This must be done
1558 by the user either by calling the SvSETMAGIC() macro after calling
1559 these functions, or by using one of the "sv_set*_mg()" or
1560 "sv_cat*_mg()" functions. Similarly, generic C code must call the
1561 SvGETMAGIC() macro to invoke any 'get' magic if they use an SV obtained
1562 from external sources in functions that don't handle magic. See
1563 perlapi for a description of these functions. For example, calls to
1564 the "sv_cat*()" functions typically need to be followed by
1565 SvSETMAGIC(), but they don't need a prior SvGETMAGIC() since their
1566 implementation handles 'get' magic.
1567 Finding Magic
1569 MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that
1570 * type */
1571 This routine returns a pointer to a "MAGIC" structure stored in the SV.
1573 If the SV does not have that magical feature, "NULL" is returned. If
1574 the SV has multiple instances of that magical feature, the first one
1575 will be returned. "mg_findext" can be used to find a "MAGIC" structure
1576 of an SV based on both its magic type and its magic virtual table:
1577 MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);
1579 Also, if the SV passed to "mg_find" or "mg_findext" is not of type
1581 SVt_PVMG, Perl may core dump.
1582 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
1584 This routine checks to see what types of magic "sv" has. If the
1586 mg_type field is an uppercase letter, then the mg_obj is copied to
1587 "nsv", but the mg_type field is changed to be the lowercase letter.
1588 Understanding the Magic of Tied Hashes and Arrays
1590 Tied hashes and arrays are magical beasts of the "PERL_MAGIC_tied"
1591 magic type.
1592 WARNING: As of the 5.004 release, proper usage of the array and hash
1594 access functions requires understanding a few caveats. Some of these
1595 caveats are actually considered bugs in the API, to be fixed in later
1596 releases, and are bracketed with [MAYCHANGE] below. If you find
1597 yourself actually applying such information in this section, be aware
1598 that the behavior may change in the future, umm, without warning.
1599 The perl tie function associates a variable with an object that
1601 implements the various GET, SET, etc methods. To perform the
1602 equivalent of the perl tie function from an XSUB, you must mimic this
1603 behaviour. The code below carries out the necessary steps -- firstly
1604 it creates a new hash, and then creates a second hash which it blesses
1605 into the class which will implement the tie methods. Lastly it ties
1606 the two hashes together, and returns a reference to the new tied hash.
1607 Note that the code below does NOT call the TIEHASH method in the MyTie
1608 class - see "Calling Perl Routines from within C Programs" for details
1609 on how to do this.
1610 SV*
1612 mytie()
1613 PREINIT:
1614 HV *hash;
1615 HV *stash;
1616 SV *tie;
1617 CODE:
1618 hash = newHV();
1619 tie = newRV_noinc((SV*)newHV());
1620 stash = gv_stashpv("MyTie", GV_ADD);
1621 sv_bless(tie, stash);
1622 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1623 RETVAL = newRV_noinc(hash);
1624 OUTPUT:
1625 RETVAL
1626 The "av_store" function, when given a tied array argument, merely
1628 copies the magic of the array onto the value to be "stored", using
1629 "mg_copy". It may also return NULL, indicating that the value did not
1630 actually need to be stored in the array. [MAYCHANGE] After a call to
1631 "av_store" on a tied array, the caller will usually need to call
1632 mg_set(val) to actually invoke the perl level "STORE" method on the
1633 TIEARRAY object. If "av_store" did return NULL, a call to
1634 SvREFCNT_dec(val) will also be usually necessary to avoid a memory
1635 leak. [/MAYCHANGE]
1636 The previous paragraph is applicable verbatim to tied hash access using
1638 the "hv_store" and "hv_store_ent" functions as well.
1639 "av_fetch" and the corresponding hash functions "hv_fetch" and
1641 "hv_fetch_ent" actually return an undefined mortal value whose magic
1642 has been initialized using "mg_copy". Note the value so returned does
1643 not need to be deallocated, as it is already mortal. [MAYCHANGE] But
1644 you will need to call mg_get() on the returned value in order to
1645 actually invoke the perl level "FETCH" method on the underlying TIE
1646 object. Similarly, you may also call mg_set() on the return value
1647 after possibly assigning a suitable value to it using "sv_setsv",
1648 which will invoke the "STORE" method on the TIE object. [/MAYCHANGE]
1649 [MAYCHANGE] In other words, the array or hash fetch/store functions
1651 don't really fetch and store actual values in the case of tied arrays
1652 and hashes. They merely call "mg_copy" to attach magic to the values
1653 that were meant to be "stored" or "fetched". Later calls to "mg_get"
1654 and "mg_set" actually do the job of invoking the TIE methods on the
1655 underlying objects. Thus the magic mechanism currently implements a
1656 kind of lazy access to arrays and hashes.
1657 Currently (as of perl version 5.004), use of the hash and array access
1659 functions requires the user to be aware of whether they are operating
1660 on "normal" hashes and arrays, or on their tied variants. The API may
1661 be changed to provide more transparent access to both tied and normal
1662 data types in future versions. [/MAYCHANGE]
1663 You would do well to understand that the TIEARRAY and TIEHASH
1665 interfaces are mere sugar to invoke some perl method calls while using
1666 the uniform hash and array syntax. The use of this sugar imposes some
1667 overhead (typically about two to four extra opcodes per FETCH/STORE
1668 operation, in addition to the creation of all the mortal variables
1669 required to invoke the methods). This overhead will be comparatively
1670 small if the TIE methods are themselves substantial, but if they are
1671 only a few statements long, the overhead will not be insignificant.
1672 Localizing changes
1674 Perl has a very handy construction
1675 {
1677 local $var = 2;
1678 ...
1679 }
1680 This construction is approximately equivalent to
1682 {
1684 my $oldvar = $var;
1685 $var = 2;
1686 ...
1687 $var = $oldvar;
1688 }
1689 The biggest difference is that the first construction would reinstate
1691 the initial value of $var, irrespective of how control exits the block:
1692 "goto", "return", "die"/"eval", etc. It is a little bit more efficient
1693 as well.
1694 There is a way to achieve a similar task from C via Perl API: create a
1696 pseudo-block, and arrange for some changes to be automatically undone
1697 at the end of it, either explicit, or via a non-local exit (via die()).
1698 A block-like construct is created by a pair of "ENTER"/"LEAVE" macros
1699 (see "Returning a Scalar" in perlcall). Such a construct may be
1700 created specially for some important localized task, or an existing one
1701 (like boundaries of enclosing Perl subroutine/block, or an existing
1702 pair for freeing TMPs) may be used. (In the second case the overhead
1703 of additional localization must be almost negligible.) Note that any
1704 XSUB is automatically enclosed in an "ENTER"/"LEAVE" pair.
1705 Inside such a pseudo-block the following service is available:
1707 "SAVEINT(int i)"
1709 "SAVEIV(IV i)"
1710 "SAVEI32(I32 i)"
1711 "SAVELONG(long i)"
1712 "SAVEI8(I8 i)"
1713 "SAVEI16(I16 i)"
1714 "SAVEBOOL(int i)"
1715 "SAVESTRLEN(STRLEN i)"
1716 These macros arrange things to restore the value of integer
1717 variable "i" at the end of the enclosing pseudo-block.
1718 SAVESPTR(s)
1720 SAVEPPTR(p)
1721 These macros arrange things to restore the value of pointers "s"
1722 and "p". "s" must be a pointer of a type which survives conversion
1723 to "SV*" and back, "p" should be able to survive conversion to
1724 "char*" and back.
1725 "SAVERCPV(char **ppv)"
1727 This macro arranges to restore the value of a "char *" variable
1728 which was allocated with a call to rcpv_new() to its previous state
1729 when the current pseudo block is completed. The pointer stored in
1730 *ppv at the time of the call will be refcount incremented and
1731 stored on the save stack. Later when the current pseudo-block is
1732 completed the value stored in *ppv will be refcount decremented,
1733 and the previous value restored from the savestack which will also
1734 be refcount decremented.
1735 This is the "RCPV" equivalent of SAVEGENERICSV().
1737 "SAVEGENERICSV(SV **psv)"
1739 This macro arranges to restore the value of a "SV *" variable to
1740 its previous state when the current pseudo block is completed. The
1741 pointer stored in *psv at the time of the call will be refcount
1742 incremented and stored on the save stack. Later when the current
1743 pseudo-block is completed the value stored in *ppv will be refcount
1744 decremented, and the previous value restored from the savestack
1745 which will also be refcount decremented. This the C equivalent of
1746 "local $sv".
1747 "SAVEFREESV(SV *sv)"
1749 The refcount of "sv" will be decremented at the end of pseudo-
1750 block. This is similar to "sv_2mortal" in that it is also a
1751 mechanism for doing a delayed "SvREFCNT_dec". However, while
1752 "sv_2mortal" extends the lifetime of "sv" until the beginning of
1753 the next statement, "SAVEFREESV" extends it until the end of the
1754 enclosing scope. These lifetimes can be wildly different.
1755 Also compare "SAVEMORTALIZESV".
1757 "SAVEMORTALIZESV(SV *sv)"
1759 Just like "SAVEFREESV", but mortalizes "sv" at the end of the
1760 current scope instead of decrementing its reference count. This
1761 usually has the effect of keeping "sv" alive until the statement
1762 that called the currently live scope has finished executing.
1763 "SAVEFREEOP(OP *op)"
1765 The "OP *" is op_free()ed at the end of pseudo-block.
1766 SAVEFREEPV(p)
1768 The chunk of memory which is pointed to by "p" is Safefree()ed at
1769 the end of the current pseudo-block.
1770 "SAVEFREERCPV(char *pv)"
1772 Ensures that a "char *" which was created by a call to rcpv_new()
1773 is rcpv_free()ed at the end of the current pseudo-block.
1774 This is the RCPV equivalent of SAVEFREESV().
1776 "SAVECLEARSV(SV *sv)"
1778 Clears a slot in the current scratchpad which corresponds to "sv"
1779 at the end of pseudo-block.
1780 "SAVEDELETE(HV *hv, char *key, I32 length)"
1782 The key "key" of "hv" is deleted at the end of pseudo-block. The
1783 string pointed to by "key" is Safefree()ed. If one has a key in
1784 short-lived storage, the corresponding string may be reallocated
1785 like this:
1786 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
1788 "SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)"
1790 At the end of pseudo-block the function "f" is called with the only
1791 argument "p" which may be NULL.
1792 "SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)"
1794 At the end of pseudo-block the function "f" is called with the
1795 implicit context argument (if any), and "p" which may be NULL.
1796 Note the end of the current pseudo-block may occur much later than
1798 the end of the current statement. You may wish to look at the
1799 MORTALDESTRUCTOR_X() macro instead.
1800 "MORTALSVFUNC_X(SVFUNC_t f, SV *sv)"
1802 At the end of the current statement the function "f" is called with
1803 the implicit context argument (if any), and "sv" which may be NULL.
1804 Be aware that the parameter argument to the destructor function
1806 differs from the related SAVEDESTRUCTOR_X() in that it MUST be
1807 either NULL or an "SV*".
1808 Note the end of the current statement may occur much before the the
1810 end of the current pseudo-block. You may wish to look at the
1811 SAVEDESTRUCTOR_X() macro instead.
1812 "MORTALDESTRUCTOR_SV(SV *coderef, SV *args)"
1814 At the end of the current statement the Perl function contained in
1815 "coderef" is called with the arguments provided (if any) in "args".
1816 See the documentation for mortal_destructor_sv() for details on the
1817 "args" parameter is handled.
1818 Note the end of the current statement may occur much before the the
1820 end of the current pseudo-block. If you wish to call a perl
1821 function at the end of the current pseudo block you should use the
1822 SAVEDESTRUCTOR_X() API instead, which will require you create a C
1823 wrapper to call the Perl function.
1824 SAVESTACK_POS()
1826 The current offset on the Perl internal stack (cf. "SP") is
1827 restored at the end of pseudo-block.
1828 The following API list contains functions, thus one needs to provide
1830 pointers to the modifiable data explicitly (either C pointers, or
1831 Perlish "GV *"s). Where the above macros take "int", a similar
1832 function takes "int *".
1833 Other macros above have functions implementing them, but its probably
1835 best to just use the macro, and not those or the ones below.
1836 "SV* save_scalar(GV *gv)"
1838 Equivalent to Perl code "local $gv".
1839 "AV* save_ary(GV *gv)"
1841 "HV* save_hash(GV *gv)"
1842 Similar to "save_scalar", but localize @gv and %gv.
1843 "void save_item(SV *item)"
1845 Duplicates the current value of "SV". On the exit from the current
1846 "ENTER"/"LEAVE" pseudo-block the value of "SV" will be restored
1847 using the stored value. It doesn't handle magic. Use
1848 "save_scalar" if magic is affected.
1849 "SV* save_svref(SV **sptr)"
1851 Similar to "save_scalar", but will reinstate an "SV *".
1852 "void save_aptr(AV **aptr)"
1854 "void save_hptr(HV **hptr)"
1855 Similar to "save_svref", but localize "AV *" and "HV *".
1856 The "Alias" module implements localization of the basic types within
1858 the caller's scope. People who are interested in how to localize
1859 things in the containing scope should take a look there too.
1860
1862 XSUBs and the Argument Stack
1863 The XSUB mechanism is a simple way for Perl programs to access C
1864 subroutines. An XSUB routine will have a stack that contains the
1865 arguments from the Perl program, and a way to map from the Perl data
1866 structures to a C equivalent.
1867 The stack arguments are accessible through the ST(n) macro, which
1869 returns the "n"'th stack argument. Argument 0 is the first argument
1870 passed in the Perl subroutine call. These arguments are "SV*", and can
1871 be used anywhere an "SV*" is used.
1872 Most of the time, output from the C routine can be handled through use
1874 of the RETVAL and OUTPUT directives. However, there are some cases
1875 where the argument stack is not already long enough to handle all the
1876 return values. An example is the POSIX tzname() call, which takes no
1877 arguments, but returns two, the local time zone's standard and summer
1878 time abbreviations.
1879 To handle this situation, the PPCODE directive is used and the stack is
1881 extended using the macro:
1882 EXTEND(SP, num);
1884 where "SP" is the macro that represents the local copy of the stack
1886 pointer, and "num" is the number of elements the stack should be
1887 extended by.
1888 Now that there is room on the stack, values can be pushed on it using
1890 "PUSHs" macro. The pushed values will often need to be "mortal" (See
1891 "Reference Counts and Mortality"):
1892 PUSHs(sv_2mortal(newSViv(an_integer)))
1894 PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
1895 PUSHs(sv_2mortal(newSVnv(a_double)))
1896 PUSHs(sv_2mortal(newSVpv("Some String",0)))
1897 /* Although the last example is better written as the more
1898 * efficient: */
1899 PUSHs(newSVpvs_flags("Some String", SVs_TEMP))
1900 And now the Perl program calling "tzname", the two values will be
1902 assigned as in:
1903 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1905 An alternate (and possibly simpler) method to pushing values on the
1907 stack is to use the macro:
1908 XPUSHs(SV*)
1910 This macro automatically adjusts the stack for you, if needed. Thus,
1912 you do not need to call "EXTEND" to extend the stack.
1913 Despite their suggestions in earlier versions of this document the
1915 macros "(X)PUSH[iunp]" are not suited to XSUBs which return multiple
1916 results. For that, either stick to the "(X)PUSHs" macros shown above,
1917 or use the new "m(X)PUSH[iunp]" macros instead; see "Putting a C value
1918 on Perl stack".
1919 For more information, consult perlxs and perlxstut.
1921 Autoloading with XSUBs
1923 If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts
1924 the fully-qualified name of the autoloaded subroutine in the $AUTOLOAD
1925 variable of the XSUB's package.
1926 But it also puts the same information in certain fields of the XSUB
1928 itself:
1929 HV *stash = CvSTASH(cv);
1931 const char *subname = SvPVX(cv);
1932 STRLEN name_length = SvCUR(cv); /* in bytes */
1933 U32 is_utf8 = SvUTF8(cv);
1934 SvPVX(cv) contains just the sub name itself, not including the package.
1936 For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
1937 CvSTASH(cv) returns NULL during a method call on a nonexistent package.
1938 Note: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
1940 XS AUTOLOAD subs at all. Perl 5.8.0 introduced the use of fields in
1941 the XSUB itself. Perl 5.16.0 restored the setting of $AUTOLOAD. If
1942 you need to support 5.8-5.14, use the XSUB's fields.
1943 Calling Perl Routines from within C Programs
1945 There are four routines that can be used to call a Perl subroutine from
1946 within a C program. These four are:
1947 I32 call_sv(SV*, I32);
1949 I32 call_pv(const char*, I32);
1950 I32 call_method(const char*, I32);
1951 I32 call_argv(const char*, I32, char**);
1952 The routine most often used is "call_sv". The "SV*" argument contains
1954 either the name of the Perl subroutine to be called, or a reference to
1955 the subroutine. The second argument consists of flags that control the
1956 context in which the subroutine is called, whether or not the
1957 subroutine is being passed arguments, how errors should be trapped, and
1958 how to treat return values.
1959 All four routines return the number of arguments that the subroutine
1961 returned on the Perl stack.
1962 These routines used to be called "perl_call_sv", etc., before Perl
1964 v5.6.0, but those names are now deprecated; macros of the same name are
1965 provided for compatibility.
1966 When using any of these routines (except "call_argv"), the programmer
1968 must manipulate the Perl stack. These include the following macros and
1969 functions:
1970 dSP
1972 SP
1973 PUSHMARK()
1974 PUTBACK
1975 SPAGAIN
1976 ENTER
1977 SAVETMPS
1978 FREETMPS
1979 LEAVE
1980 XPUSH*()
1981 POP*()
1982 For a detailed description of calling conventions from C to Perl,
1984 consult perlcall.
1985 Putting a C value on Perl stack
1987 A lot of opcodes (this is an elementary operation in the internal perl
1988 stack machine) put an SV* on the stack. However, as an optimization
1989 the corresponding SV is (usually) not recreated each time. The opcodes
1990 reuse specially assigned SVs (targets) which are (as a corollary) not
1991 constantly freed/created.
1992 Each of the targets is created only once (but see "Scratchpads and
1994 recursion" below), and when an opcode needs to put an integer, a
1995 double, or a string on the stack, it just sets the corresponding parts
1996 of its target and puts the target on stack.
1997 The macro to put this target on stack is "PUSHTARG", and it is directly
1999 used in some opcodes, as well as indirectly in zillions of others,
2000 which use it via "(X)PUSH[iunp]".
2001 Because the target is reused, you must be careful when pushing multiple
2003 values on the stack. The following code will not do what you think:
2004 XPUSHi(10);
2006 XPUSHi(20);
2007 This translates as "set "TARG" to 10, push a pointer to "TARG" onto the
2009 stack; set "TARG" to 20, push a pointer to "TARG" onto the stack". At
2010 the end of the operation, the stack does not contain the values 10 and
2011 20, but actually contains two pointers to "TARG", which we have set to
2012 20.
2013 If you need to push multiple different values then you should either
2015 use the "(X)PUSHs" macros, or else use the new "m(X)PUSH[iunp]" macros,
2016 none of which make use of "TARG". The "(X)PUSHs" macros simply push an
2017 SV* on the stack, which, as noted under "XSUBs and the Argument Stack",
2018 will often need to be "mortal". The new "m(X)PUSH[iunp]" macros make
2019 this a little easier to achieve by creating a new mortal for you (via
2020 "(X)PUSHmortal"), pushing that onto the stack (extending it if
2021 necessary in the case of the "mXPUSH[iunp]" macros), and then setting
2022 its value. Thus, instead of writing this to "fix" the example above:
2023 XPUSHs(sv_2mortal(newSViv(10)))
2025 XPUSHs(sv_2mortal(newSViv(20)))
2026 you can simply write:
2028 mXPUSHi(10)
2030 mXPUSHi(20)
2031 On a related note, if you do use "(X)PUSH[iunp]", then you're going to
2033 need a "dTARG" in your variable declarations so that the "*PUSH*"
2034 macros can make use of the local variable "TARG". See also "dTARGET"
2035 and "dXSTARG".
2036 Scratchpads
2038 The question remains on when the SVs which are targets for opcodes are
2039 created. The answer is that they are created when the current unit--a
2040 subroutine or a file (for opcodes for statements outside of
2041 subroutines)--is compiled. During this time a special anonymous Perl
2042 array is created, which is called a scratchpad for the current unit.
2043 A scratchpad keeps SVs which are lexicals for the current unit and are
2045 targets for opcodes. A previous version of this document stated that
2046 one can deduce that an SV lives on a scratchpad by looking on its
2047 flags: lexicals have "SVs_PADMY" set, and targets have "SVs_PADTMP"
2048 set. But this has never been fully true. "SVs_PADMY" could be set on
2049 a variable that no longer resides in any pad. While targets do have
2050 "SVs_PADTMP" set, it can also be set on variables that have never
2051 resided in a pad, but nonetheless act like targets. As of perl 5.21.5,
2052 the "SVs_PADMY" flag is no longer used and is defined as 0. SvPADMY()
2053 now returns true for anything without "SVs_PADTMP".
2054 The correspondence between OPs and targets is not 1-to-1. Different
2056 OPs in the compile tree of the unit can use the same target, if this
2057 would not conflict with the expected life of the temporary.
2058 Scratchpads and recursion
2060 In fact it is not 100% true that a compiled unit contains a pointer to
2061 the scratchpad AV. In fact it contains a pointer to an AV of
2062 (initially) one element, and this element is the scratchpad AV. Why do
2063 we need an extra level of indirection?
2064 The answer is recursion, and maybe threads. Both these can create
2066 several execution pointers going into the same subroutine. For the
2067 subroutine-child not write over the temporaries for the subroutine-
2068 parent (lifespan of which covers the call to the child), the parent and
2069 the child should have different scratchpads. (And the lexicals should
2070 be separate anyway!)
2071 So each subroutine is born with an array of scratchpads (of length 1).
2073 On each entry to the subroutine it is checked that the current depth of
2074 the recursion is not more than the length of this array, and if it is,
2075 new scratchpad is created and pushed into the array.
2076 The targets on this scratchpad are "undef"s, but they are already
2078 marked with correct flags.
2079
2081 Allocation
2082 All memory meant to be used with the Perl API functions should be
2083 manipulated using the macros described in this section. The macros
2084 provide the necessary transparency between differences in the actual
2085 malloc implementation that is used within perl.
2086 The following three macros are used to initially allocate memory :
2088 Newx(pointer, number, type);
2090 Newxc(pointer, number, type, cast);
2091 Newxz(pointer, number, type);
2092 The first argument "pointer" should be the name of a variable that will
2094 point to the newly allocated memory.
2095 The second and third arguments "number" and "type" specify how many of
2097 the specified type of data structure should be allocated. The argument
2098 "type" is passed to "sizeof". The final argument to "Newxc", "cast",
2099 should be used if the "pointer" argument is different from the "type"
2100 argument.
2101 Unlike the "Newx" and "Newxc" macros, the "Newxz" macro calls "memzero"
2103 to zero out all the newly allocated memory.
2104 Reallocation
2106 Renew(pointer, number, type);
2107 Renewc(pointer, number, type, cast);
2108 Safefree(pointer)
2109 These three macros are used to change a memory buffer size or to free a
2111 piece of memory no longer needed. The arguments to "Renew" and
2112 "Renewc" match those of "New" and "Newc" with the exception of not
2113 needing the "magic cookie" argument.
2114 Moving
2116 Move(source, dest, number, type);
2117 Copy(source, dest, number, type);
2118 Zero(dest, number, type);
2119 These three macros are used to move, copy, or zero out previously
2121 allocated memory. The "source" and "dest" arguments point to the
2122 source and destination starting points. Perl will move, copy, or zero
2123 out "number" instances of the size of the "type" data structure (using
2124 the "sizeof" function).
2125
2127 The most recent development releases of Perl have been experimenting
2128 with removing Perl's dependency on the "normal" standard I/O suite and
2129 allowing other stdio implementations to be used. This involves
2130 creating a new abstraction layer that then calls whichever
2131 implementation of stdio Perl was compiled with. All XSUBs should now
2132 use the functions in the PerlIO abstraction layer and not make any
2133 assumptions about what kind of stdio is being used.
2134 For a complete description of the PerlIO abstraction, consult perlapio.
2136
2138 Code tree
2139 Here we describe the internal form your code is converted to by Perl.
2140 Start with a simple example:
2141 $a = $b + $c;
2143 This is converted to a tree similar to this one:
2145 assign-to
2147 / \
2148 + $a
2149 / \
2150 $b $c
2151 (but slightly more complicated). This tree reflects the way Perl
2153 parsed your code, but has nothing to do with the execution order.
2154 There is an additional "thread" going through the nodes of the tree
2155 which shows the order of execution of the nodes. In our simplified
2156 example above it looks like:
2157 $b ---> $c ---> + ---> $a ---> assign-to
2159 But with the actual compile tree for "$a = $b + $c" it is different:
2161 some nodes optimized away. As a corollary, though the actual tree
2162 contains more nodes than our simplified example, the execution order is
2163 the same as in our example.
2164 Examining the tree
2166 If you have your perl compiled for debugging (usually done with
2167 "-DDEBUGGING" on the "Configure" command line), you may examine the
2168 compiled tree by specifying "-Dx" on the Perl command line. The output
2169 takes several lines per node, and for "$b+$c" it looks like this:
2170 5 TYPE = add ===> 6
2172 TARG = 1
2173 FLAGS = (SCALAR,KIDS)
2174 {
2175 TYPE = null ===> (4)
2176 (was rv2sv)
2177 FLAGS = (SCALAR,KIDS)
2178 {
2179 3 TYPE = gvsv ===> 4
2180 FLAGS = (SCALAR)
2181 GV = main::b
2182 }
2183 }
2184 {
2185 TYPE = null ===> (5)
2186 (was rv2sv)
2187 FLAGS = (SCALAR,KIDS)
2188 {
2189 4 TYPE = gvsv ===> 5
2190 FLAGS = (SCALAR)
2191 GV = main::c
2192 }
2193 }
2194 This tree has 5 nodes (one per "TYPE" specifier), only 3 of them are
2196 not optimized away (one per number in the left column). The immediate
2197 children of the given node correspond to "{}" pairs on the same level
2198 of indentation, thus this listing corresponds to the tree:
2199 add
2201 / \
2202 null null
2203 | |
2204 gvsv gvsv
2205 The execution order is indicated by "===>" marks, thus it is "3 4 5 6"
2207 (node 6 is not included into above listing), i.e., "gvsv gvsv add
2208 whatever".
2209 Each of these nodes represents an op, a fundamental operation inside
2211 the Perl core. The code which implements each operation can be found
2212 in the pp*.c files; the function which implements the op with type
2213 "gvsv" is "pp_gvsv", and so on. As the tree above shows, different ops
2214 have different numbers of children: "add" is a binary operator, as one
2215 would expect, and so has two children. To accommodate the various
2216 different numbers of children, there are various types of op data
2217 structure, and they link together in different ways.
2218 The simplest type of op structure is "OP": this has no children. Unary
2220 operators, "UNOP"s, have one child, and this is pointed to by the
2221 "op_first" field. Binary operators ("BINOP"s) have not only an
2222 "op_first" field but also an "op_last" field. The most complex type of
2223 op is a "LISTOP", which has any number of children. In this case, the
2224 first child is pointed to by "op_first" and the last child by
2225 "op_last". The children in between can be found by iteratively
2226 following the "OpSIBLING" pointer from the first child to the last (but
2227 see below).
2228 There are also some other op types: a "PMOP" holds a regular
2230 expression, and has no children, and a "LOOP" may or may not have
2231 children. If the "op_children" field is non-zero, it behaves like a
2232 "LISTOP". To complicate matters, if a "UNOP" is actually a "null" op
2233 after optimization (see "Compile pass 2: context propagation") it will
2234 still have children in accordance with its former type.
2235 Finally, there is a "LOGOP", or logic op. Like a "LISTOP", this has one
2237 or more children, but it doesn't have an "op_last" field: so you have
2238 to follow "op_first" and then the "OpSIBLING" chain itself to find the
2239 last child. Instead it has an "op_other" field, which is comparable to
2240 the "op_next" field described below, and represents an alternate
2241 execution path. Operators like "and", "or" and "?" are "LOGOP"s. Note
2242 that in general, "op_other" may not point to any of the direct children
2243 of the "LOGOP".
2244 Starting in version 5.21.2, perls built with the experimental define
2246 "-DPERL_OP_PARENT" add an extra boolean flag for each op, "op_moresib".
2247 When not set, this indicates that this is the last op in an "OpSIBLING"
2248 chain. This frees up the "op_sibling" field on the last sibling to
2249 point back to the parent op. Under this build, that field is also
2250 renamed "op_sibparent" to reflect its joint role. The macro
2251 OpSIBLING(o) wraps this special behaviour, and always returns NULL on
2252 the last sibling. With this build the op_parent(o) function can be
2253 used to find the parent of any op. Thus for forward compatibility, you
2254 should always use the OpSIBLING(o) macro rather than accessing
2255 "op_sibling" directly.
2256 Another way to examine the tree is to use a compiler back-end module,
2258 such as B::Concise.
2259 Compile pass 1: check routines
2261 The tree is created by the compiler while yacc code feeds it the
2262 constructions it recognizes. Since yacc works bottom-up, so does the
2263 first pass of perl compilation.
2264 What makes this pass interesting for perl developers is that some
2266 optimization may be performed on this pass. This is optimization by
2267 so-called "check routines". The correspondence between node names and
2268 corresponding check routines is described in opcode.pl (do not forget
2269 to run "make regen_headers" if you modify this file).
2270 A check routine is called when the node is fully constructed except for
2272 the execution-order thread. Since at this time there are no back-links
2273 to the currently constructed node, one can do most any operation to the
2274 top-level node, including freeing it and/or creating new nodes
2275 above/below it.
2276 The check routine returns the node which should be inserted into the
2278 tree (if the top-level node was not modified, check routine returns its
2279 argument).
2280 By convention, check routines have names "ck_*". They are usually
2282 called from "new*OP" subroutines (or "convert") (which in turn are
2283 called from perly.y).
2284 Compile pass 1a: constant folding
2286 Immediately after the check routine is called the returned node is
2287 checked for being compile-time executable. If it is (the value is
2288 judged to be constant) it is immediately executed, and a constant node
2289 with the "return value" of the corresponding subtree is substituted
2290 instead. The subtree is deleted.
2291 If constant folding was not performed, the execution-order thread is
2293 created.
2294 Compile pass 2: context propagation
2296 When a context for a part of compile tree is known, it is propagated
2297 down through the tree. At this time the context can have 5 values
2298 (instead of 2 for runtime context): void, boolean, scalar, list, and
2299 lvalue. In contrast with the pass 1 this pass is processed from top to
2300 bottom: a node's context determines the context for its children.
2301 Additional context-dependent optimizations are performed at this time.
2303 Since at this moment the compile tree contains back-references (via
2304 "thread" pointers), nodes cannot be free()d now. To allow optimized-
2305 away nodes at this stage, such nodes are null()ified instead of
2306 free()ing (i.e. their type is changed to OP_NULL).
2307 Compile pass 3: peephole optimization
2309 After the compile tree for a subroutine (or for an "eval" or a file) is
2310 created, an additional pass over the code is performed. This pass is
2311 neither top-down or bottom-up, but in the execution order (with
2312 additional complications for conditionals). Optimizations performed at
2313 this stage are subject to the same restrictions as in the pass 2.
2314 Peephole optimizations are done by calling the function pointed to by
2316 the global variable "PL_peepp". By default, "PL_peepp" just calls the
2317 function pointed to by the global variable "PL_rpeepp". By default,
2318 that performs some basic op fixups and optimisations along the
2319 execution-order op chain, and recursively calls "PL_rpeepp" for each
2320 side chain of ops (resulting from conditionals). Extensions may
2321 provide additional optimisations or fixups, hooking into either the
2322 per-subroutine or recursive stage, like this:
2323 static peep_t prev_peepp;
2325 static void my_peep(pTHX_ OP *o)
2326 {
2327 /* custom per-subroutine optimisation goes here */
2328 prev_peepp(aTHX_ o);
2329 /* custom per-subroutine optimisation may also go here */
2330 }
2331 BOOT:
2332 prev_peepp = PL_peepp;
2333 PL_peepp = my_peep;
2334 static peep_t prev_rpeepp;
2336 static void my_rpeep(pTHX_ OP *first)
2337 {
2338 OP *o = first, *t = first;
2339 for(; o = o->op_next, t = t->op_next) {
2340 /* custom per-op optimisation goes here */
2341 o = o->op_next;
2342 if (!o || o == t) break;
2343 /* custom per-op optimisation goes AND here */
2344 }
2345 prev_rpeepp(aTHX_ orig_o);
2346 }
2347 BOOT:
2348 prev_rpeepp = PL_rpeepp;
2349 PL_rpeepp = my_rpeep;
2350 Pluggable runops
2352 The compile tree is executed in a runops function. There are two
2353 runops functions, in run.c and in dump.c. "Perl_runops_debug" is used
2354 with DEBUGGING and "Perl_runops_standard" is used otherwise. For fine
2355 control over the execution of the compile tree it is possible to
2356 provide your own runops function.
2357 It's probably best to copy one of the existing runops functions and
2359 change it to suit your needs. Then, in the BOOT section of your XS
2360 file, add the line:
2361 PL_runops = my_runops;
2363 This function should be as efficient as possible to keep your programs
2365 running as fast as possible.
2366 Compile-time scope hooks
2368 As of perl 5.14 it is possible to hook into the compile-time lexical
2369 scope mechanism using "Perl_blockhook_register". This is used like
2370 this:
2371 STATIC void my_start_hook(pTHX_ int full);
2373 STATIC BHK my_hooks;
2374 BOOT:
2376 BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
2377 Perl_blockhook_register(aTHX_ &my_hooks);
2378 This will arrange to have "my_start_hook" called at the start of
2380 compiling every lexical scope. The available hooks are:
2381 "void bhk_start(pTHX_ int full)"
2383 This is called just after starting a new lexical scope. Note that
2384 Perl code like
2385 if ($x) { ... }
2387 creates two scopes: the first starts at the "(" and has "full ==
2389 1", the second starts at the "{" and has "full == 0". Both end at
2390 the "}", so calls to "start" and "pre"/"post_end" will match.
2391 Anything pushed onto the save stack by this hook will be popped
2392 just before the scope ends (between the "pre_" and "post_end"
2393 hooks, in fact).
2394 "void bhk_pre_end(pTHX_ OP **o)"
2396 This is called at the end of a lexical scope, just before unwinding
2397 the stack. o is the root of the optree representing the scope; it
2398 is a double pointer so you can replace the OP if you need to.
2399 "void bhk_post_end(pTHX_ OP **o)"
2401 This is called at the end of a lexical scope, just after unwinding
2402 the stack. o is as above. Note that it is possible for calls to
2403 "pre_" and "post_end" to nest, if there is something on the save
2404 stack that calls string eval.
2405 "void bhk_eval(pTHX_ OP *const o)"
2407 This is called just before starting to compile an "eval STRING",
2408 "do FILE", "require" or "use", after the eval has been set up. o
2409 is the OP that requested the eval, and will normally be an
2410 "OP_ENTEREVAL", "OP_DOFILE" or "OP_REQUIRE".
2411 Once you have your hook functions, you need a "BHK" structure to put
2413 them in. It's best to allocate it statically, since there is no way to
2414 free it once it's registered. The function pointers should be inserted
2415 into this structure using the "BhkENTRY_set" macro, which will also set
2416 flags indicating which entries are valid. If you do need to allocate
2417 your "BHK" dynamically for some reason, be sure to zero it before you
2418 start.
2419 Once registered, there is no mechanism to switch these hooks off, so if
2421 that is necessary you will need to do this yourself. An entry in "%^H"
2422 is probably the best way, so the effect is lexically scoped; however it
2423 is also possible to use the "BhkDISABLE" and "BhkENABLE" macros to
2424 temporarily switch entries on and off. You should also be aware that
2425 generally speaking at least one scope will have opened before your
2426 extension is loaded, so you will see some "pre"/"post_end" pairs that
2427 didn't have a matching "start".
2428
2430 To aid debugging, the source file dump.c contains a number of functions
2431 which produce formatted output of internal data structures.
2432 The most commonly used of these functions is "Perl_sv_dump"; it's used
2434 for dumping SVs, AVs, HVs, and CVs. The "Devel::Peek" module calls
2435 "sv_dump" to produce debugging output from Perl-space, so users of that
2436 module should already be familiar with its format.
2437 "Perl_op_dump" can be used to dump an "OP" structure or any of its
2439 derivatives, and produces output similar to "perl -Dx"; in fact,
2440 "Perl_dump_eval" will dump the main root of the code being evaluated,
2441 exactly like "-Dx".
2442 Other useful functions are "Perl_dump_sub", which turns a "GV" into an
2444 op tree, "Perl_dump_packsubs" which calls "Perl_dump_sub" on all the
2445 subroutines in a package like so: (Thankfully, these are all xsubs, so
2446 there is no op tree)
2447 (gdb) print Perl_dump_packsubs(PL_defstash)
2449 SUB attributes::bootstrap = (xsub 0x811fedc 0)
2451 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
2453 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
2455 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
2457 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
2459 and "Perl_dump_all", which dumps all the subroutines in the stash and
2461 the op tree of the main root.
2462
2464 Background and MULTIPLICITY
2465 The Perl interpreter can be regarded as a closed box: it has an API for
2466 feeding it code or otherwise making it do things, but it also has
2467 functions for its own use. This smells a lot like an object, and there
2468 is a way for you to build Perl so that you can have multiple
2469 interpreters, with one interpreter represented either as a C structure,
2470 or inside a thread-specific structure. These structures contain all
2471 the context, the state of that interpreter.
2472 The macro that controls the major Perl build flavor is MULTIPLICITY.
2474 The MULTIPLICITY build has a C structure that packages all the
2475 interpreter state, which is being passed to various perl functions as a
2476 "hidden" first argument. MULTIPLICITY makes multi-threaded perls
2477 possible (with the ithreads threading model, related to the macro
2478 USE_ITHREADS.)
2479 PERL_IMPLICIT_CONTEXT is a legacy synonym for MULTIPLICITY.
2481 To see whether you have non-const data you can use a BSD (or GNU)
2483 compatible "nm":
2484 nm libperl.a | grep -v ' [TURtr] '
2486 If this displays any "D" or "d" symbols (or possibly "C" or "c"), you
2488 have non-const data. The symbols the "grep" removed are as follows:
2489 "Tt" are text, or code, the "Rr" are read-only (const) data, and the
2490 "U" is <undefined>, external symbols referred to.
2491 The test t/porting/libperl.t does this kind of symbol sanity checking
2493 on "libperl.a".
2494 All this obviously requires a way for the Perl internal functions to be
2496 either subroutines taking some kind of structure as the first argument,
2497 or subroutines taking nothing as the first argument. To enable these
2498 two very different ways of building the interpreter, the Perl source
2499 (as it does in so many other situations) makes heavy use of macros and
2500 subroutine naming conventions.
2501 First problem: deciding which functions will be public API functions
2503 and which will be private. All functions whose names begin "S_" are
2504 private (think "S" for "secret" or "static"). All other functions
2505 begin with "Perl_", but just because a function begins with "Perl_"
2506 does not mean it is part of the API. (See "Internal Functions".) The
2507 easiest way to be sure a function is part of the API is to find its
2508 entry in perlapi. If it exists in perlapi, it's part of the API. If
2509 it doesn't, and you think it should be (i.e., you need it for your
2510 extension), submit an issue at <https://github.com/Perl/perl5/issues>
2511 explaining why you think it should be.
2512 Second problem: there must be a syntax so that the same subroutine
2514 declarations and calls can pass a structure as their first argument, or
2515 pass nothing. To solve this, the subroutines are named and declared in
2516 a particular way. Here's a typical start of a static function used
2517 within the Perl guts:
2518 STATIC void
2520 S_incline(pTHX_ char *s)
2521 STATIC becomes "static" in C, and may be #define'd to nothing in some
2523 configurations in the future.
2524 A public function (i.e. part of the internal API, but not necessarily
2526 sanctioned for use in extensions) begins like this:
2527 void
2529 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
2530 "pTHX_" is one of a number of macros (in perl.h) that hide the details
2532 of the interpreter's context. THX stands for "thread", "this", or
2533 "thingy", as the case may be. (And no, George Lucas is not involved.
2534 :-) The first character could be 'p' for a prototype, 'a' for argument,
2535 or 'd' for declaration, so we have "pTHX", "aTHX" and "dTHX", and their
2536 variants.
2537 When Perl is built without options that set MULTIPLICITY, there is no
2539 first argument containing the interpreter's context. The trailing
2540 underscore in the pTHX_ macro indicates that the macro expansion needs
2541 a comma after the context argument because other arguments follow it.
2542 If MULTIPLICITY is not defined, pTHX_ will be ignored, and the
2543 subroutine is not prototyped to take the extra argument. The form of
2544 the macro without the trailing underscore is used when there are no
2545 additional explicit arguments.
2546 When a core function calls another, it must pass the context. This is
2548 normally hidden via macros. Consider "sv_setiv". It expands into
2549 something like this:
2550 #ifdef MULTIPLICITY
2552 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
2553 /* can't do this for vararg functions, see below */
2554 #else
2555 #define sv_setiv Perl_sv_setiv
2556 #endif
2557 This works well, and means that XS authors can gleefully write:
2559 sv_setiv(foo, bar);
2561 and still have it work under all the modes Perl could have been
2563 compiled with.
2564 This doesn't work so cleanly for varargs functions, though, as macros
2566 imply that the number of arguments is known in advance. Instead we
2567 either need to spell them out fully, passing "aTHX_" as the first
2568 argument (the Perl core tends to do this with functions like
2569 Perl_warner), or use a context-free version.
2570 The context-free version of Perl_warner is called
2572 Perl_warner_nocontext, and does not take the extra argument. Instead
2573 it does "dTHX;" to get the context from thread-local storage. We
2574 "#define warner Perl_warner_nocontext" so that extensions get source
2575 compatibility at the expense of performance. (Passing an arg is
2576 cheaper than grabbing it from thread-local storage.)
2577 You can ignore [pad]THXx when browsing the Perl headers/sources. Those
2579 are strictly for use within the core. Extensions and embedders need
2580 only be aware of [pad]THX.
2581 So what happened to dTHR?
2583 "dTHR" was introduced in perl 5.005 to support the older thread model.
2584 The older thread model now uses the "THX" mechanism to pass context
2585 pointers around, so "dTHR" is not useful any more. Perl 5.6.0 and
2586 later still have it for backward source compatibility, but it is
2587 defined to be a no-op.
2588 How do I use all this in extensions?
2590 When Perl is built with MULTIPLICITY, extensions that call any
2591 functions in the Perl API will need to pass the initial context
2592 argument somehow. The kicker is that you will need to write it in such
2593 a way that the extension still compiles when Perl hasn't been built
2594 with MULTIPLICITY enabled.
2595 There are three ways to do this. First, the easy but inefficient way,
2597 which is also the default, in order to maintain source compatibility
2598 with extensions: whenever XSUB.h is #included, it redefines the aTHX
2599 and aTHX_ macros to call a function that will return the context.
2600 Thus, something like:
2601 sv_setiv(sv, num);
2603 in your extension will translate to this when MULTIPLICITY is in
2605 effect:
2606 Perl_sv_setiv(Perl_get_context(), sv, num);
2608 or to this otherwise:
2610 Perl_sv_setiv(sv, num);
2612 You don't have to do anything new in your extension to get this; since
2614 the Perl library provides Perl_get_context(), it will all just work.
2615 The second, more efficient way is to use the following template for
2617 your Foo.xs:
2618 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2620 #include "EXTERN.h"
2621 #include "perl.h"
2622 #include "XSUB.h"
2623 STATIC void my_private_function(int arg1, int arg2);
2625 STATIC void
2627 my_private_function(int arg1, int arg2)
2628 {
2629 dTHX; /* fetch context */
2630 ... call many Perl API functions ...
2631 }
2632 [... etc ...]
2634 MODULE = Foo PACKAGE = Foo
2636 /* typical XSUB */
2638 void
2640 my_xsub(arg)
2641 int arg
2642 CODE:
2643 my_private_function(arg, 10);
2644 Note that the only two changes from the normal way of writing an
2646 extension is the addition of a "#define PERL_NO_GET_CONTEXT" before
2647 including the Perl headers, followed by a "dTHX;" declaration at the
2648 start of every function that will call the Perl API. (You'll know
2649 which functions need this, because the C compiler will complain that
2650 there's an undeclared identifier in those functions.) No changes are
2651 needed for the XSUBs themselves, because the XS() macro is correctly
2652 defined to pass in the implicit context if needed.
2653 The third, even more efficient way is to ape how it is done within the
2655 Perl guts:
2656 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2658 #include "EXTERN.h"
2659 #include "perl.h"
2660 #include "XSUB.h"
2661 /* pTHX_ only needed for functions that call Perl API */
2663 STATIC void my_private_function(pTHX_ int arg1, int arg2);
2664 STATIC void
2666 my_private_function(pTHX_ int arg1, int arg2)
2667 {
2668 /* dTHX; not needed here, because THX is an argument */
2669 ... call Perl API functions ...
2670 }
2671 [... etc ...]
2673 MODULE = Foo PACKAGE = Foo
2675 /* typical XSUB */
2677 void
2679 my_xsub(arg)
2680 int arg
2681 CODE:
2682 my_private_function(aTHX_ arg, 10);
2683 This implementation never has to fetch the context using a function
2685 call, since it is always passed as an extra argument. Depending on
2686 your needs for simplicity or efficiency, you may mix the previous two
2687 approaches freely.
2688 Never add a comma after "pTHX" yourself--always use the form of the
2690 macro with the underscore for functions that take explicit arguments,
2691 or the form without the argument for functions with no explicit
2692 arguments.
2693 Should I do anything special if I call perl from multiple threads?
2695 If you create interpreters in one thread and then proceed to call them
2696 in another, you need to make sure perl's own Thread Local Storage (TLS)
2697 slot is initialized correctly in each of those threads.
2698 The "perl_alloc" and "perl_clone" API functions will automatically set
2700 the TLS slot to the interpreter they created, so that there is no need
2701 to do anything special if the interpreter is always accessed in the
2702 same thread that created it, and that thread did not create or call any
2703 other interpreters afterwards. If that is not the case, you have to
2704 set the TLS slot of the thread before calling any functions in the Perl
2705 API on that particular interpreter. This is done by calling the
2706 "PERL_SET_CONTEXT" macro in that thread as the first thing you do:
2707 /* do this before doing anything else with some_perl */
2709 PERL_SET_CONTEXT(some_perl);
2710 ... other Perl API calls on some_perl go here ...
2712 (You can always get the current context via "PERL_GET_CONTEXT".)
2714 Future Plans and PERL_IMPLICIT_SYS
2716 Just as MULTIPLICITY provides a way to bundle up everything that the
2717 interpreter knows about itself and pass it around, so too are there
2718 plans to allow the interpreter to bundle up everything it knows about
2719 the environment it's running on. This is enabled with the
2720 PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS on
2721 Windows.
2722 This allows the ability to provide an extra pointer (called the "host"
2724 environment) for all the system calls. This makes it possible for all
2725 the system stuff to maintain their own state, broken down into seven C
2726 structures. These are thin wrappers around the usual system calls (see
2727 win32/perllib.c) for the default perl executable, but for a more
2728 ambitious host (like the one that would do fork() emulation) all the
2729 extra work needed to pretend that different interpreters are actually
2730 different "processes", would be done here.
2731 The Perl engine/interpreter and the host are orthogonal entities.
2733 There could be one or more interpreters in a process, and one or more
2734 "hosts", with free association between them.
2735
2737 All of Perl's internal functions which will be exposed to the outside
2738 world are prefixed by "Perl_" so that they will not conflict with XS
2739 functions or functions used in a program in which Perl is embedded.
2740 Similarly, all global variables begin with "PL_". (By convention,
2741 static functions start with "S_".)
2742 Inside the Perl core ("PERL_CORE" defined), you can get at the
2744 functions either with or without the "Perl_" prefix, thanks to a bunch
2745 of defines that live in embed.h. Note that extension code should not
2746 set "PERL_CORE"; this exposes the full perl internals, and is likely to
2747 cause breakage of the XS in each new perl release.
2748 The file embed.h is generated automatically from embed.pl and
2750 embed.fnc. embed.pl also creates the prototyping header files for the
2751 internal functions, generates the documentation and a lot of other bits
2752 and pieces. It's important that when you add a new function to the
2753 core or change an existing one, you change the data in the table in
2754 embed.fnc as well. Here's a sample entry from that table:
2755 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2757 The first column is a set of flags, the second column the return type,
2759 the third column the name. Columns after that are the arguments. The
2760 flags are documented at the top of embed.fnc.
2761 If you edit embed.pl or embed.fnc, you will need to run "make
2763 regen_headers" to force a rebuild of embed.h and other auto-generated
2764 files.
2765 Formatted Printing of IVs, UVs, and NVs
2767 If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2768 formatting codes like %d, %ld, %f, you should use the following macros
2769 for portability
2770 IVdf IV in decimal
2772 UVuf UV in decimal
2773 UVof UV in octal
2774 UVxf UV in hexadecimal
2775 NVef NV %e-like
2776 NVff NV %f-like
2777 NVgf NV %g-like
2778 These will take care of 64-bit integers and long doubles. For example:
2780 printf("IV is %" IVdf "\n", iv);
2782 The "IVdf" will expand to whatever is the correct format for the IVs.
2784 Note that the spaces are required around the format in case the code is
2785 compiled with C++, to maintain compliance with its standard.
2786 Note that there are different "long doubles": Perl will use whatever
2788 the compiler has.
2789 If you are printing addresses of pointers, use %p or UVxf combined with
2791 PTR2UV().
2792 Formatted Printing of SVs
2794 The contents of SVs may be printed using the "SVf" format, like so:
2795 Perl_croak(aTHX_ "This croaked because: %" SVf "\n", SVfARG(err_msg))
2797 where "err_msg" is an SV.
2799 Not all scalar types are printable. Simple values certainly are: one
2801 of IV, UV, NV, or PV. Also, if the SV is a reference to some value,
2802 either it will be dereferenced and the value printed, or information
2803 about the type of that value and its address are displayed. The
2804 results of printing any other type of SV are undefined and likely to
2805 lead to an interpreter crash. NVs are printed using a %g-ish format.
2806 Note that the spaces are required around the "SVf" in case the code is
2808 compiled with C++, to maintain compliance with its standard.
2809 Note that any filehandle being printed to under UTF-8 must be expecting
2811 UTF-8 in order to get good results and avoid Wide-character warnings.
2812 One way to do this for typical filehandles is to invoke perl with the
2813 "-C" parameter. (See "-C [number/list]" in perlrun.
2814 You can use this to concatenate two scalars:
2816 SV *var1 = get_sv("var1", GV_ADD);
2818 SV *var2 = get_sv("var2", GV_ADD);
2819 SV *var3 = newSVpvf("var1=%" SVf " and var2=%" SVf,
2820 SVfARG(var1), SVfARG(var2));
2821 "SVf_QUOTEDPREFIX" is similar to "SVf" except that it restricts the
2823 number of the characters printed, showing at most the first
2824 "PERL_QUOTEDPREFIX_LEN" characters of the argument, and rendering it
2825 with double quotes and with the contents escaped using double quoted
2826 string escaping rules. If the string is longer than this then ellipses
2827 "..." will be appended after the trailing quote. This is intended for
2828 error messages where the string is assumed to be a class name.
2829 "HvNAMEf" and "HvNAMEf_QUOTEDPREFIX" are similar to "SVf" except they
2831 extract the string, length and utf8 flags from the argument using the
2832 HvNAME(), HvNAMELEN(), HvNAMEUTF8() macros. This is intended for
2833 stringifying a class name directly from an stash HV.
2834 Formatted Printing of Strings
2836 If you just want the bytes printed in a 7bit NUL-terminated string, you
2837 can just use %s (assuming they are all really only 7bit). But if there
2838 is a possibility the value will be encoded as UTF-8 or contains bytes
2839 above 0x7F (and therefore 8bit), you should instead use the "UTF8f"
2840 format. And as its parameter, use the UTF8fARG() macro:
2841 chr * msg;
2843 /* U+2018: \xE2\x80\x98 LEFT SINGLE QUOTATION MARK
2845 U+2019: \xE2\x80\x99 RIGHT SINGLE QUOTATION MARK */
2846 if (can_utf8)
2847 msg = "\xE2\x80\x98Uses fancy quotes\xE2\x80\x99";
2848 else
2849 msg = "'Uses simple quotes'";
2850 Perl_croak(aTHX_ "The message is: %" UTF8f "\n",
2852 UTF8fARG(can_utf8, strlen(msg), msg));
2853 The first parameter to "UTF8fARG" is a boolean: 1 if the string is in
2855 UTF-8; 0 if string is in native byte encoding (Latin1). The second
2856 parameter is the number of bytes in the string to print. And the third
2857 and final parameter is a pointer to the first byte in the string.
2858 Note that any filehandle being printed to under UTF-8 must be expecting
2860 UTF-8 in order to get good results and avoid Wide-character warnings.
2861 One way to do this for typical filehandles is to invoke perl with the
2862 "-C" parameter. (See "-C [number/list]" in perlrun.
2863 Formatted Printing of "Size_t" and "SSize_t"
2865 The most general way to do this is to cast them to a UV or IV, and
2866 print as in the previous section.
2867 But if you're using PerlIO_printf(), it's less typing and visual
2869 clutter to use the %z length modifier (for siZe):
2870 PerlIO_printf("STRLEN is %zu\n", len);
2872 This modifier is not portable, so its use should be restricted to
2874 PerlIO_printf().
2875 Formatted Printing of "Ptrdiff_t", "intmax_t", "short" and other special
2877 sizes
2878 There are modifiers for these special situations if you are using
2879 PerlIO_printf(). See "size" in perlfunc.
2880 Pointer-To-Integer and Integer-To-Pointer
2882 Because pointer size does not necessarily equal integer size, use the
2883 follow macros to do it right.
2884 PTR2UV(pointer)
2886 PTR2IV(pointer)
2887 PTR2NV(pointer)
2888 INT2PTR(pointertotype, integer)
2889 For example:
2891 IV iv = ...;
2893 SV *sv = INT2PTR(SV*, iv);
2894 and
2896 AV *av = ...;
2898 UV uv = PTR2UV(av);
2899 There are also
2901 PTR2nat(pointer) /* pointer to integer of PTRSIZE */
2903 PTR2ul(pointer) /* pointer to unsigned long */
2904 And "PTRV" which gives the native type for an integer the same size as
2906 pointers, such as "unsigned" or "unsigned long".
2907 Exception Handling
2909 There are a couple of macros to do very basic exception handling in XS
2910 modules. You have to define "NO_XSLOCKS" before including XSUB.h to be
2911 able to use these macros:
2912 #define NO_XSLOCKS
2914 #include "XSUB.h"
2915 You can use these macros if you call code that may croak, but you need
2917 to do some cleanup before giving control back to Perl. For example:
2918 dXCPT; /* set up necessary variables */
2920 XCPT_TRY_START {
2922 code_that_may_croak();
2923 } XCPT_TRY_END
2924 XCPT_CATCH
2926 {
2927 /* do cleanup here */
2928 XCPT_RETHROW;
2929 }
2930 Note that you always have to rethrow an exception that has been caught.
2932 Using these macros, it is not possible to just catch the exception and
2933 ignore it. If you have to ignore the exception, you have to use the
2934 "call_*" function.
2935 The advantage of using the above macros is that you don't have to setup
2937 an extra function for "call_*", and that using these macros is faster
2938 than using "call_*".
2939 Source Documentation
2941 There's an effort going on to document the internal functions and
2942 automatically produce reference manuals from them -- perlapi is one
2943 such manual which details all the functions which are available to XS
2944 writers. perlintern is the autogenerated manual for the functions
2945 which are not part of the API and are supposedly for internal use only.
2946 Source documentation is created by putting POD comments into the C
2948 source, like this:
2949 /*
2951 =for apidoc sv_setiv
2952 Copies an integer into the given SV. Does not handle 'set' magic. See
2954 L<perlapi/sv_setiv_mg>.
2955 =cut
2957 */
2958 Please try and supply some documentation if you add functions to the
2960 Perl core.
2961 Backwards compatibility
2963 The Perl API changes over time. New functions are added or the
2964 interfaces of existing functions are changed. The "Devel::PPPort"
2965 module tries to provide compatibility code for some of these changes,
2966 so XS writers don't have to code it themselves when supporting multiple
2967 versions of Perl.
2968 "Devel::PPPort" generates a C header file ppport.h that can also be run
2970 as a Perl script. To generate ppport.h, run:
2971 perl -MDevel::PPPort -eDevel::PPPort::WriteFile
2973 Besides checking existing XS code, the script can also be used to
2975 retrieve compatibility information for various API calls using the
2976 "--api-info" command line switch. For example:
2977 % perl ppport.h --api-info=sv_magicext
2979 For details, see "perldoc ppport.h".
2981
2983 Perl 5.6.0 introduced Unicode support. It's important for porters and
2984 XS writers to understand this support and make sure that the code they
2985 write does not corrupt Unicode data.
2986 What is Unicode, anyway?
2988 In the olden, less enlightened times, we all used to use ASCII. Most
2989 of us did, anyway. The big problem with ASCII is that it's American.
2990 Well, no, that's not actually the problem; the problem is that it's not
2991 particularly useful for people who don't use the Roman alphabet. What
2992 used to happen was that particular languages would stick their own
2993 alphabet in the upper range of the sequence, between 128 and 255. Of
2994 course, we then ended up with plenty of variants that weren't quite
2995 ASCII, and the whole point of it being a standard was lost.
2996 Worse still, if you've got a language like Chinese or Japanese that has
2998 hundreds or thousands of characters, then you really can't fit them
2999 into a mere 256, so they had to forget about ASCII altogether, and
3000 build their own systems using pairs of numbers to refer to one
3001 character.
3002 To fix this, some people formed Unicode, Inc. and produced a new
3004 character set containing all the characters you can possibly think of
3005 and more. There are several ways of representing these characters, and
3006 the one Perl uses is called UTF-8. UTF-8 uses a variable number of
3007 bytes to represent a character. You can learn more about Unicode and
3008 Perl's Unicode model in perlunicode.
3009 (On EBCDIC platforms, Perl uses instead UTF-EBCDIC, which is a form of
3011 UTF-8 adapted for EBCDIC platforms. Below, we just talk about UTF-8.
3012 UTF-EBCDIC is like UTF-8, but the details are different. The macros
3013 hide the differences from you, just remember that the particular
3014 numbers and bit patterns presented below will differ in UTF-EBCDIC.)
3015 How can I recognise a UTF-8 string?
3017 You can't. This is because UTF-8 data is stored in bytes just like
3018 non-UTF-8 data. The Unicode character 200, (0xC8 for you hex types)
3019 capital E with a grave accent, is represented by the two bytes
3020 "v196.172". Unfortunately, the non-Unicode string "chr(196).chr(172)"
3021 has that byte sequence as well. So you can't tell just by looking --
3022 this is what makes Unicode input an interesting problem.
3023 In general, you either have to know what you're dealing with, or you
3025 have to guess. The API function "is_utf8_string" can help; it'll tell
3026 you if a string contains only valid UTF-8 characters, and the chances
3027 of a non-UTF-8 string looking like valid UTF-8 become very small very
3028 quickly with increasing string length. On a character-by-character
3029 basis, "isUTF8_CHAR" will tell you whether the current character in a
3030 string is valid UTF-8.
3031 How does UTF-8 represent Unicode characters?
3033 As mentioned above, UTF-8 uses a variable number of bytes to store a
3034 character. Characters with values 0...127 are stored in one byte, just
3035 like good ol' ASCII. Character 128 is stored as "v194.128"; this
3036 continues up to character 191, which is "v194.191". Now we've run out
3037 of bits (191 is binary 10111111) so we move on; character 192 is
3038 "v195.128". And so it goes on, moving to three bytes at character
3039 2048. "Unicode Encodings" in perlunicode has pictures of how this
3040 works.
3041 Assuming you know you're dealing with a UTF-8 string, you can find out
3043 how long the first character in it is with the "UTF8SKIP" macro:
3044 char *utf = "\305\233\340\240\201";
3046 I32 len;
3047 len = UTF8SKIP(utf); /* len is 2 here */
3049 utf += len;
3050 len = UTF8SKIP(utf); /* len is 3 here */
3051 Another way to skip over characters in a UTF-8 string is to use
3053 "utf8_hop", which takes a string and a number of characters to skip
3054 over. You're on your own about bounds checking, though, so don't use
3055 it lightly.
3056 All bytes in a multi-byte UTF-8 character will have the high bit set,
3058 so you can test if you need to do something special with this character
3059 like this (the UTF8_IS_INVARIANT() is a macro that tests whether the
3060 byte is encoded as a single byte even in UTF-8):
3061 U8 *utf; /* Initialize this to point to the beginning of the
3063 sequence to convert */
3064 U8 *utf_end; /* Initialize this to 1 beyond the end of the sequence
3065 pointed to by 'utf' */
3066 UV uv; /* Returned code point; note: a UV, not a U8, not a
3067 char */
3068 STRLEN len; /* Returned length of character in bytes */
3069 if (!UTF8_IS_INVARIANT(*utf))
3071 /* Must treat this as UTF-8 */
3072 uv = utf8_to_uvchr_buf(utf, utf_end, &len);
3073 else
3074 /* OK to treat this character as a byte */
3075 uv = *utf;
3076 You can also see in that example that we use "utf8_to_uvchr_buf" to get
3078 the value of the character; the inverse function "uvchr_to_utf8" is
3079 available for putting a UV into UTF-8:
3080 if (!UVCHR_IS_INVARIANT(uv))
3082 /* Must treat this as UTF8 */
3083 utf8 = uvchr_to_utf8(utf8, uv);
3084 else
3085 /* OK to treat this character as a byte */
3086 *utf8++ = uv;
3087 You must convert characters to UVs using the above functions if you're
3089 ever in a situation where you have to match UTF-8 and non-UTF-8
3090 characters. You may not skip over UTF-8 characters in this case. If
3091 you do this, you'll lose the ability to match hi-bit non-UTF-8
3092 characters; for instance, if your UTF-8 string contains "v196.172", and
3093 you skip that character, you can never match a chr(200) in a non-UTF-8
3094 string. So don't do that!
3095 (Note that we don't have to test for invariant characters in the
3097 examples above. The functions work on any well-formed UTF-8 input.
3098 It's just that its faster to avoid the function overhead when it's not
3099 needed.)
3100 How does Perl store UTF-8 strings?
3102 Currently, Perl deals with UTF-8 strings and non-UTF-8 strings slightly
3103 differently. A flag in the SV, "SVf_UTF8", indicates that the string
3104 is internally encoded as UTF-8. Without it, the byte value is the
3105 codepoint number and vice versa. This flag is only meaningful if the
3106 SV is "SvPOK" or immediately after stringification via "SvPV" or a
3107 similar macro. You can check and manipulate this flag with the
3108 following macros:
3109 SvUTF8(sv)
3111 SvUTF8_on(sv)
3112 SvUTF8_off(sv)
3113 This flag has an important effect on Perl's treatment of the string: if
3115 UTF-8 data is not properly distinguished, regular expressions,
3116 "length", "substr" and other string handling operations will have
3117 undesirable (wrong) results.
3118 The problem comes when you have, for instance, a string that isn't
3120 flagged as UTF-8, and contains a byte sequence that could be UTF-8 --
3121 especially when combining non-UTF-8 and UTF-8 strings.
3122 Never forget that the "SVf_UTF8" flag is separate from the PV value;
3124 you need to be sure you don't accidentally knock it off while you're
3125 manipulating SVs. More specifically, you cannot expect to do this:
3126 SV *sv;
3128 SV *nsv;
3129 STRLEN len;
3130 char *p;
3131 p = SvPV(sv, len);
3133 frobnicate(p);
3134 nsv = newSVpvn(p, len);
3135 The "char*" string does not tell you the whole story, and you can't
3137 copy or reconstruct an SV just by copying the string value. Check if
3138 the old SV has the UTF8 flag set (after the "SvPV" call), and act
3139 accordingly:
3140 p = SvPV(sv, len);
3142 is_utf8 = SvUTF8(sv);
3143 frobnicate(p, is_utf8);
3144 nsv = newSVpvn(p, len);
3145 if (is_utf8)
3146 SvUTF8_on(nsv);
3147 In the above, your "frobnicate" function has been changed to be made
3149 aware of whether or not it's dealing with UTF-8 data, so that it can
3150 handle the string appropriately.
3151 Since just passing an SV to an XS function and copying the data of the
3153 SV is not enough to copy the UTF8 flags, even less right is just
3154 passing a "char *" to an XS function.
3155 For full generality, use the "DO_UTF8" macro to see if the string in an
3157 SV is to be treated as UTF-8. This takes into account if the call to
3158 the XS function is being made from within the scope of "use bytes". If
3159 so, the underlying bytes that comprise the UTF-8 string are to be
3160 exposed, rather than the character they represent. But this pragma
3161 should only really be used for debugging and perhaps low-level testing
3162 at the byte level. Hence most XS code need not concern itself with
3163 this, but various areas of the perl core do need to support it.
3164 And this isn't the whole story. Starting in Perl v5.12, strings that
3166 aren't encoded in UTF-8 may also be treated as Unicode under various
3167 conditions (see "ASCII Rules versus Unicode Rules" in perlunicode).
3168 This is only really a problem for characters whose ordinals are between
3169 128 and 255, and their behavior varies under ASCII versus Unicode rules
3170 in ways that your code cares about (see "The "Unicode Bug"" in
3171 perlunicode). There is no published API for dealing with this, as it
3172 is subject to change, but you can look at the code for "pp_lc" in pp.c
3173 for an example as to how it's currently done.
3174 How do I pass a Perl string to a C library?
3176 A Perl string, conceptually, is an opaque sequence of code points.
3177 Many C libraries expect their inputs to be "classical" C strings, which
3178 are arrays of octets 1-255, terminated with a NUL byte. Your job when
3179 writing an interface between Perl and a C library is to define the
3180 mapping between Perl and that library.
3181 Generally speaking, "SvPVbyte" and related macros suit this task well.
3183 These assume that your Perl string is a "byte string", i.e., is either
3184 raw, undecoded input into Perl or is pre-encoded to, e.g., UTF-8.
3185 Alternatively, if your C library expects UTF-8 text, you can use
3187 "SvPVutf8" and related macros. This has the same effect as encoding to
3188 UTF-8 then calling the corresponding "SvPVbyte"-related macro.
3189 Some C libraries may expect other encodings (e.g., UTF-16LE). To give
3191 Perl strings to such libraries you must either do that encoding in Perl
3192 then use "SvPVbyte", or use an intermediary C library to convert from
3193 however Perl stores the string to the desired encoding.
3194 Take care also that NULs in your Perl string don't confuse the C
3196 library. If possible, give the string's length to the C library; if
3197 that's not possible, consider rejecting strings that contain NUL bytes.
3198 What about "SvPV", "SvPV_nolen", etc.?
3200 Consider a 3-character Perl string "$foo = "\x64\x78\x8c"". Perl can
3202 store these 3 characters either of two ways:
3203 • bytes: 0x64 0x78 0x8c
3205 • UTF-8: 0x64 0x78 0xc2 0x8c
3207 Now let's say you convert $foo to a C string thus:
3209 STRLEN strlen;
3211 char *str = SvPV(foo_sv, strlen);
3212 At this point "str" could point to a 3-byte C string or a 4-byte one.
3214 Generally speaking, we want "str" to be the same regardless of how Perl
3216 stores $foo, so the ambiguity here is undesirable. "SvPVbyte" and
3217 "SvPVutf8" solve that by giving predictable output: use "SvPVbyte" if
3218 your C library expects byte strings, or "SvPVutf8" if it expects UTF-8.
3219 If your C library happens to support both encodings, then
3221 "SvPV"--always in tandem with lookups to "SvUTF8"!--may be safe and
3222 (slightly) more efficient.
3223 TESTING TIP: Use utf8's "upgrade" and "downgrade" functions in your
3225 tests to ensure consistent handling regardless of Perl's internal
3226 encoding.
3227 How do I convert a string to UTF-8?
3229 If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to
3230 upgrade the non-UTF-8 strings to UTF-8. If you've got an SV, the
3231 easiest way to do this is:
3232 sv_utf8_upgrade(sv);
3234 However, you must not do this, for example:
3236 if (!SvUTF8(left))
3238 sv_utf8_upgrade(left);
3239 If you do this in a binary operator, you will actually change one of
3241 the strings that came into the operator, and, while it shouldn't be
3242 noticeable by the end user, it can cause problems in deficient code.
3243 Instead, "bytes_to_utf8" will give you a UTF-8-encoded copy of its
3245 string argument. This is useful for having the data available for
3246 comparisons and so on, without harming the original SV. There's also
3247 "utf8_to_bytes" to go the other way, but naturally, this will fail if
3248 the string contains any characters above 255 that can't be represented
3249 in a single byte.
3250 How do I compare strings?
3252 "sv_cmp" in perlapi and "sv_cmp_flags" in perlapi do a lexigraphic
3253 comparison of two SV's, and handle UTF-8ness properly. Note, however,
3254 that Unicode specifies a much fancier mechanism for collation,
3255 available via the Unicode::Collate module.
3256 To just compare two strings for equality/non-equality, you can just use
3258 memEQ() and memNE() as usual, except the strings must be both UTF-8 or
3259 not UTF-8 encoded.
3260 To compare two strings case-insensitively, use foldEQ_utf8() (the
3262 strings don't have to have the same UTF-8ness).
3263 Is there anything else I need to know?
3265 Not really. Just remember these things:
3266 • There's no way to tell if a "char *" or "U8 *" string is UTF-8 or
3268 not. But you can tell if an SV is to be treated as UTF-8 by calling
3269 "DO_UTF8" on it, after stringifying it with "SvPV" or a similar
3270 macro. And, you can tell if SV is actually UTF-8 (even if it is not
3271 to be treated as such) by looking at its "SvUTF8" flag (again after
3272 stringifying it). Don't forget to set the flag if something should
3273 be UTF-8. Treat the flag as part of the PV, even though it's not --
3274 if you pass on the PV to somewhere, pass on the flag too.
3275 • If a string is UTF-8, always use "utf8_to_uvchr_buf" to get at the
3277 value, unless UTF8_IS_INVARIANT(*s) in which case you can use *s.
3278 • When writing a character UV to a UTF-8 string, always use
3280 "uvchr_to_utf8", unless "UVCHR_IS_INVARIANT(uv))" in which case you
3281 can use "*s = uv".
3282 • Mixing UTF-8 and non-UTF-8 strings is tricky. Use "bytes_to_utf8"
3284 to get a new string which is UTF-8 encoded, and then combine them.
3285
3287 Custom operator support is an experimental feature that allows you to
3288 define your own ops. This is primarily to allow the building of
3289 interpreters for other languages in the Perl core, but it also allows
3290 optimizations through the creation of "macro-ops" (ops which perform
3291 the functions of multiple ops which are usually executed together, such
3292 as "gvsv, gvsv, add".)
3293 This feature is implemented as a new op type, "OP_CUSTOM". The Perl
3295 core does not "know" anything special about this op type, and so it
3296 will not be involved in any optimizations. This also means that you
3297 can define your custom ops to be any op structure -- unary, binary,
3298 list and so on -- you like.
3299 It's important to know what custom operators won't do for you. They
3301 won't let you add new syntax to Perl, directly. They won't even let
3302 you add new keywords, directly. In fact, they won't change the way
3303 Perl compiles a program at all. You have to do those changes yourself,
3304 after Perl has compiled the program. You do this either by
3305 manipulating the op tree using a "CHECK" block and the "B::Generate"
3306 module, or by adding a custom peephole optimizer with the "optimize"
3307 module.
3308 When you do this, you replace ordinary Perl ops with custom ops by
3310 creating ops with the type "OP_CUSTOM" and the "op_ppaddr" of your own
3311 PP function. This should be defined in XS code, and should look like
3312 the PP ops in "pp_*.c". You are responsible for ensuring that your op
3313 takes the appropriate number of values from the stack, and you are
3314 responsible for adding stack marks if necessary.
3315 You should also "register" your op with the Perl interpreter so that it
3317 can produce sensible error and warning messages. Since it is possible
3318 to have multiple custom ops within the one "logical" op type
3319 "OP_CUSTOM", Perl uses the value of "o->op_ppaddr" to determine which
3320 custom op it is dealing with. You should create an "XOP" structure for
3321 each ppaddr you use, set the properties of the custom op with
3322 "XopENTRY_set", and register the structure against the ppaddr using
3323 "Perl_custom_op_register". A trivial example might look like:
3324 static XOP my_xop;
3326 static OP *my_pp(pTHX);
3327 BOOT:
3329 XopENTRY_set(&my_xop, xop_name, "myxop");
3330 XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
3331 Perl_custom_op_register(aTHX_ my_pp, &my_xop);
3332 The available fields in the structure are:
3334 xop_name
3336 A short name for your op. This will be included in some error
3337 messages, and will also be returned as "$op->name" by the B module,
3338 so it will appear in the output of module like B::Concise.
3339 xop_desc
3341 A short description of the function of the op.
3342 xop_class
3344 Which of the various *OP structures this op uses. This should be
3345 one of the "OA_*" constants from op.h, namely
3346 OA_BASEOP
3348 OA_UNOP
3349 OA_BINOP
3350 OA_LOGOP
3351 OA_LISTOP
3352 OA_PMOP
3353 OA_SVOP
3354 OA_PADOP
3355 OA_PVOP_OR_SVOP
3356 This should be interpreted as '"PVOP"' only. The "_OR_SVOP" is
3357 because the only core "PVOP", "OP_TRANS", can sometimes be a
3358 "SVOP" instead.
3359 OA_LOOP
3361 OA_COP
3362 The other "OA_*" constants should not be used.
3364 xop_peep
3366 This member is of type "Perl_cpeep_t", which expands to "void
3367 (*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)". If it is set, this
3368 function will be called from "Perl_rpeep" when ops of this type are
3369 encountered by the peephole optimizer. o is the OP that needs
3370 optimizing; oldop is the previous OP optimized, whose "op_next"
3371 points to o.
3372 "B::Generate" directly supports the creation of custom ops by name.
3374
3376 Descriptions above occasionally refer to "the stack", but there are in
3377 fact many stack-like data structures within the perl interpreter. When
3378 otherwise unqualified, "the stack" usually refers to the value stack.
3379 The various stacks have different purposes, and operate in slightly
3381 different ways. Their differences are noted below.
3382 Value Stack
3384 This stack stores the values that regular perl code is operating on,
3385 usually intermediate values of expressions within a statement. The
3386 stack itself is formed of an array of SV pointers.
3387 The base of this stack is pointed to by the interpreter variable
3389 "PL_stack_base", of type "SV **".
3390 The head of the stack is "PL_stack_sp", and points to the most
3392 recently-pushed item.
3393 Items are pushed to the stack by using the PUSHs() macro or its
3395 variants described above; XPUSHs(), mPUSHs(), mXPUSHs() and the typed
3396 versions. Note carefully that the non-"X" versions of these macros do
3397 not check the size of the stack and assume it to be big enough. These
3398 must be paired with a suitable check of the stack's size, such as the
3399 "EXTEND" macro to ensure it is large enough. For example
3400 EXTEND(SP, 4);
3402 mPUSHi(10);
3403 mPUSHi(20);
3404 mPUSHi(30);
3405 mPUSHi(40);
3406 This is slightly more performant than making four separate checks in
3408 four separate mXPUSHi() calls.
3409 As a further performance optimisation, the various "PUSH" macros all
3411 operate using a local variable "SP", rather than the interpreter-global
3412 variable "PL_stack_sp". This variable is declared by the "dSP" macro -
3413 though it is normally implied by XSUBs and similar so it is rare you
3414 have to consider it directly. Once declared, the "PUSH" macros will
3415 operate only on this local variable, so before invoking any other perl
3416 core functions you must use the "PUTBACK" macro to return the value
3417 from the local "SP" variable back to the interpreter variable.
3418 Similarly, after calling a perl core function which may have had reason
3419 to move the stack or push/pop values to it, you must use the "SPAGAIN"
3420 macro which refreshes the local "SP" value back from the interpreter
3421 one.
3422 Items are popped from the stack by using the "POPs" macro or its typed
3424 versions, There is also a macro "TOPs" that inspects the topmost item
3425 without removing it.
3426 Note specifically that SV pointers on the value stack do not contribute
3428 to the overall reference count of the xVs being referred to. If newly-
3429 created xVs are being pushed to the stack you must arrange for them to
3430 be destroyed at a suitable time; usually by using one of the "mPUSH*"
3431 macros or sv_2mortal() to mortalise the xV.
3432 Mark Stack
3434 The value stack stores individual perl scalar values as temporaries
3435 between expressions. Some perl expressions operate on entire lists; for
3436 that purpose we need to know where on the stack each list begins. This
3437 is the purpose of the mark stack.
3438 The mark stack stores integers as I32 values, which are the height of
3440 the value stack at the time before the list began; thus the mark itself
3441 actually points to the value stack entry one before the list. The list
3442 itself starts at "mark + 1".
3443 The base of this stack is pointed to by the interpreter variable
3445 "PL_markstack", of type "I32 *".
3446 The head of the stack is "PL_markstack_ptr", and points to the most
3448 recently-pushed item.
3449 Items are pushed to the stack by using the PUSHMARK() macro. Even
3451 though the stack itself stores (value) stack indices as integers, the
3452 "PUSHMARK" macro should be given a stack pointer directly; it will
3453 calculate the index offset by comparing to the "PL_stack_sp" variable.
3454 Thus almost always the code to perform this is
3455 PUSHMARK(SP);
3457 Items are popped from the stack by the "POPMARK" macro. There is also a
3459 macro "TOPMARK" that inspects the topmost item without removing it.
3460 These macros return I32 index values directly. There is also the
3461 "dMARK" macro which declares a new SV double-pointer variable, called
3462 "mark", which points at the marked stack slot; this is the usual macro
3463 that C code will use when operating on lists given on the stack.
3464 As noted above, the "mark" variable itself will point at the most
3466 recently pushed value on the value stack before the list begins, and so
3467 the list itself starts at "mark + 1". The values of the list may be
3468 iterated by code such as
3469 for(SV **svp = mark + 1; svp <= PL_stack_sp; svp++) {
3471 SV *item = *svp;
3472 ...
3473 }
3474 Note specifically in the case that the list is already empty, "mark"
3476 will equal "PL_stack_sp".
3477 Because the "mark" variable is converted to a pointer on the value
3479 stack, extra care must be taken if "EXTEND" or any of the "XPUSH"
3480 macros are invoked within the function, because the stack may need to
3481 be moved to extend it and so the existing pointer will now be invalid.
3482 If this may be a problem, a possible solution is to track the mark
3483 offset as an integer and track the mark itself later on after the stack
3484 had been moved.
3485 I32 markoff = POPMARK;
3487 ...
3489 SP **mark = PL_stack_base + markoff;
3491 Temporaries Stack
3493 As noted above, xV references on the main value stack do not contribute
3494 to the reference count of an xV, and so another mechanism is used to
3495 track when temporary values which live on the stack must be released.
3496 This is the job of the temporaries stack.
3497 The temporaries stack stores pointers to xVs whose reference counts
3499 will be decremented soon.
3500 The base of this stack is pointed to by the interpreter variable
3502 "PL_tmps_stack", of type "SV **".
3503 The head of the stack is indexed by "PL_tmps_ix", an integer which
3505 stores the index in the array of the most recently-pushed item.
3506 There is no public API to directly push items to the temporaries stack.
3508 Instead, the API function sv_2mortal() is used to mortalize an xV,
3509 adding its address to the temporaries stack.
3510 Likewise, there is no public API to read values from the temporaries
3512 stack. Instead, the macros "SAVETMPS" and "FREETMPS" are used. The
3513 "SAVETMPS" macro establishes the base levels of the temporaries stack,
3514 by capturing the current value of "PL_tmps_ix" into "PL_tmps_floor" and
3515 saving the previous value to the save stack. Thereafter, whenever
3516 "FREETMPS" is invoked all of the temporaries that have been pushed
3517 since that level are reclaimed.
3518 While it is common to see these two macros in pairs within an "ENTER"/
3520 "LEAVE" pair, it is not necessary to match them. It is permitted to
3521 invoke "FREETMPS" multiple times since the most recent "SAVETMPS"; for
3522 example in a loop iterating over elements of a list. While you can
3523 invoke "SAVETMPS" multiple times within a scope pair, it is unlikely to
3524 be useful. Subsequent invocations will move the temporaries floor
3525 further up, thus effectively trapping the existing temporaries to only
3526 be released at the end of the scope.
3527 Save Stack
3529 The save stack is used by perl to implement the "local" keyword and
3530 other similar behaviours; any cleanup operations that need to be
3531 performed when leaving the current scope. Items pushed to this stack
3532 generally capture the current value of some internal variable or state,
3533 which will be restored when the scope is unwound due to leaving,
3534 "return", "die", "goto" or other reasons.
3535 Whereas other perl internal stacks store individual items all of the
3537 same type (usually SV pointers or integers), the items pushed to the
3538 save stack are formed of many different types, having multiple fields
3539 to them. For example, the "SAVEt_INT" type needs to store both the
3540 address of the "int" variable to restore, and the value to restore it
3541 to. This information could have been stored using fields of a "struct",
3542 but would have to be large enough to store three pointers in the
3543 largest case, which would waste a lot of space in most of the smaller
3544 cases.
3545 Instead, the stack stores information in a variable-length encoding of
3547 "ANY" structures. The final value pushed is stored in the "UV" field
3548 which encodes the kind of item held by the preceding items; the count
3549 and types of which will depend on what kind of item is being stored.
3550 The kind field is pushed last because that will be the first field to
3551 be popped when unwinding items from the stack.
3552 The base of this stack is pointed to by the interpreter variable
3554 "PL_savestack", of type "ANY *".
3555 The head of the stack is indexed by "PL_savestack_ix", an integer which
3557 stores the index in the array at which the next item should be pushed.
3558 (Note that this is different to most other stacks, which reference the
3559 most recently-pushed item).
3560 Items are pushed to the save stack by using the various "SAVE...()"
3562 macros. Many of these macros take a variable and store both its
3563 address and current value on the save stack, ensuring that value gets
3564 restored on scope exit.
3565 SAVEI8(i8)
3567 SAVEI16(i16)
3568 SAVEI32(i32)
3569 SAVEINT(i)
3570 ...
3571 There are also a variety of other special-purpose macros which save
3573 particular types or values of interest. "SAVETMPS" has already been
3574 mentioned above. Others include "SAVEFREEPV" which arranges for a PV
3575 (i.e. a string buffer) to be freed, or "SAVEDESTRUCTOR" which arranges
3576 for a given function pointer to be invoked on scope exit. A full list
3577 of such macros can be found in scope.h.
3578 There is no public API for popping individual values or items from the
3580 save stack. Instead, via the scope stack, the "ENTER" and "LEAVE" pair
3581 form a way to start and stop nested scopes. Leaving a nested scope via
3582 "LEAVE" will restore all of the saved values that had been pushed since
3583 the most recent "ENTER".
3584 Scope Stack
3586 As with the mark stack to the value stack, the scope stack forms a pair
3587 with the save stack. The scope stack stores the height of the save
3588 stack at which nested scopes begin, and allows the save stack to be
3589 unwound back to that point when the scope is left.
3590 When perl is built with debugging enabled, there is a second part to
3592 this stack storing human-readable string names describing the type of
3593 stack context. Each push operation saves the name as well as the height
3594 of the save stack, and each pop operation checks the topmost name with
3595 what is expected, causing an assertion failure if the name does not
3596 match.
3597 The base of this stack is pointed to by the interpreter variable
3599 "PL_scopestack", of type "I32 *". If enabled, the scope stack names are
3600 stored in a separate array pointed to by "PL_scopestack_name", of type
3601 "const char **".
3602 The head of the stack is indexed by "PL_scopestack_ix", an integer
3604 which stores the index of the array or arrays at which the next item
3605 should be pushed. (Note that this is different to most other stacks,
3606 which reference the most recently-pushed item).
3607 Values are pushed to the scope stack using the "ENTER" macro, which
3609 begins a new nested scope. Any items pushed to the save stack are then
3610 restored at the next nested invocation of the "LEAVE" macro.
3611
3613 Note: this section describes a non-public internal API that is subject
3614 to change without notice.
3615 Introduction to the context stack
3617 In Perl, dynamic scoping refers to the runtime nesting of things like
3618 subroutine calls, evals etc, as well as the entering and exiting of
3619 block scopes. For example, the restoring of a "local"ised variable is
3620 determined by the dynamic scope.
3621 Perl tracks the dynamic scope by a data structure called the context
3623 stack, which is an array of "PERL_CONTEXT" structures, and which is
3624 itself a big union for all the types of context. Whenever a new scope
3625 is entered (such as a block, a "for" loop, or a subroutine call), a new
3626 context entry is pushed onto the stack. Similarly when leaving a block
3627 or returning from a subroutine call etc. a context is popped. Since the
3628 context stack represents the current dynamic scope, it can be searched.
3629 For example, "next LABEL" searches back through the stack looking for a
3630 loop context that matches the label; "return" pops contexts until it
3631 finds a sub or eval context or similar; "caller" examines sub contexts
3632 on the stack.
3633 Each context entry is labelled with a context type, "cx_type". Typical
3635 context types are "CXt_SUB", "CXt_EVAL" etc., as well as "CXt_BLOCK"
3636 and "CXt_NULL" which represent a basic scope (as pushed by "pp_enter")
3637 and a sort block. The type determines which part of the context union
3638 are valid.
3639 The main division in the context struct is between a substitution scope
3641 ("CXt_SUBST") and block scopes, which are everything else. The former
3642 is just used while executing "s///e", and won't be discussed further
3643 here.
3644 All the block scope types share a common base, which corresponds to
3646 "CXt_BLOCK". This stores the old values of various scope-related
3647 variables like "PL_curpm", as well as information about the current
3648 scope, such as "gimme". On scope exit, the old variables are restored.
3649 Particular block scope types store extra per-type information. For
3651 example, "CXt_SUB" stores the currently executing CV, while the various
3652 for loop types might hold the original loop variable SV. On scope exit,
3653 the per-type data is processed; for example the CV has its reference
3654 count decremented, and the original loop variable is restored.
3655 The macro "cxstack" returns the base of the current context stack,
3657 while "cxstack_ix" is the index of the current frame within that stack.
3658 In fact, the context stack is actually part of a stack-of-stacks
3660 system; whenever something unusual is done such as calling a "DESTROY"
3661 or tie handler, a new stack is pushed, then popped at the end.
3662 Note that the API described here changed considerably in perl 5.24;
3664 prior to that, big macros like "PUSHBLOCK" and "POPSUB" were used; in
3665 5.24 they were replaced by the inline static functions described below.
3666 In addition, the ordering and detail of how these macros/function work
3667 changed in many ways, often subtly. In particular they didn't handle
3668 saving the savestack and temps stack positions, and required additional
3669 "ENTER", "SAVETMPS" and "LEAVE" compared to the new functions. The old-
3670 style macros will not be described further.
3671 Pushing contexts
3673 For pushing a new context, the two basic functions are "cx =
3674 cx_pushblock()", which pushes a new basic context block and returns its
3675 address, and a family of similar functions with names like
3676 cx_pushsub(cx) which populate the additional type-dependent fields in
3677 the "cx" struct. Note that "CXt_NULL" and "CXt_BLOCK" don't have their
3678 own push functions, as they don't store any data beyond that pushed by
3679 "cx_pushblock".
3680 The fields of the context struct and the arguments to the "cx_*"
3682 functions are subject to change between perl releases, representing
3683 whatever is convenient or efficient for that release.
3684 A typical context stack pushing can be found in "pp_entersub"; the
3686 following shows a simplified and stripped-down example of a non-XS
3687 call, along with comments showing roughly what each function does.
3688 dMARK;
3690 U8 gimme = GIMME_V;
3691 bool hasargs = cBOOL(PL_op->op_flags & OPf_STACKED);
3692 OP *retop = PL_op->op_next;
3693 I32 old_ss_ix = PL_savestack_ix;
3694 CV *cv = ....;
3695 /* ... make mortal copies of stack args which are PADTMPs here ... */
3697 /* ... do any additional savestack pushes here ... */
3699 /* Now push a new context entry of type 'CXt_SUB'; initially just
3701 * doing the actions common to all block types: */
3702 cx = cx_pushblock(CXt_SUB, gimme, MARK, old_ss_ix);
3704 /* this does (approximately):
3706 CXINC; /* cxstack_ix++ (grow if necessary) */
3707 cx = CX_CUR(); /* and get the address of new frame */
3708 cx->cx_type = CXt_SUB;
3709 cx->blk_gimme = gimme;
3710 cx->blk_oldsp = MARK - PL_stack_base;
3711 cx->blk_oldsaveix = old_ss_ix;
3712 cx->blk_oldcop = PL_curcop;
3713 cx->blk_oldmarksp = PL_markstack_ptr - PL_markstack;
3714 cx->blk_oldscopesp = PL_scopestack_ix;
3715 cx->blk_oldpm = PL_curpm;
3716 cx->blk_old_tmpsfloor = PL_tmps_floor;
3717 PL_tmps_floor = PL_tmps_ix;
3719 */
3720 /* then update the new context frame with subroutine-specific info,
3723 * such as the CV about to be executed: */
3724 cx_pushsub(cx, cv, retop, hasargs);
3726 /* this does (approximately):
3728 cx->blk_sub.cv = cv;
3729 cx->blk_sub.olddepth = CvDEPTH(cv);
3730 cx->blk_sub.prevcomppad = PL_comppad;
3731 cx->cx_type |= (hasargs) ? CXp_HASARGS : 0;
3732 cx->blk_sub.retop = retop;
3733 SvREFCNT_inc_simple_void_NN(cv);
3734 */
3735 Note that cx_pushblock() sets two new floors: for the args stack (to
3737 "MARK") and the temps stack (to "PL_tmps_ix"). While executing at this
3738 scope level, every "nextstate" (amongst others) will reset the args and
3739 tmps stack levels to these floors. Note that since "cx_pushblock" uses
3740 the current value of "PL_tmps_ix" rather than it being passed as an
3741 arg, this dictates at what point "cx_pushblock" should be called. In
3742 particular, any new mortals which should be freed only on scope exit
3743 (rather than at the next "nextstate") should be created first.
3744 Most callers of "cx_pushblock" simply set the new args stack floor to
3746 the top of the previous stack frame, but for "CXt_LOOP_LIST" it stores
3747 the items being iterated over on the stack, and so sets "blk_oldsp" to
3748 the top of these items instead. Note that, contrary to its name,
3749 "blk_oldsp" doesn't always represent the value to restore "PL_stack_sp"
3750 to on scope exit.
3751 Note the early capture of "PL_savestack_ix" to "old_ss_ix", which is
3753 later passed as an arg to "cx_pushblock". In the case of "pp_entersub",
3754 this is because, although most values needing saving are stored in
3755 fields of the context struct, an extra value needs saving only when the
3756 debugger is running, and it doesn't make sense to bloat the struct for
3757 this rare case. So instead it is saved on the savestack. Since this
3758 value gets calculated and saved before the context is pushed, it is
3759 necessary to pass the old value of "PL_savestack_ix" to "cx_pushblock",
3760 to ensure that the saved value gets freed during scope exit. For most
3761 users of "cx_pushblock", where nothing needs pushing on the save stack,
3762 "PL_savestack_ix" is just passed directly as an arg to "cx_pushblock".
3763 Note that where possible, values should be saved in the context struct
3765 rather than on the save stack; it's much faster that way.
3766 Normally "cx_pushblock" should be immediately followed by the
3768 appropriate "cx_pushfoo", with nothing between them; this is because if
3769 code in-between could die (e.g. a warning upgraded to fatal), then the
3770 context stack unwinding code in "dounwind" would see (in the example
3771 above) a "CXt_SUB" context frame, but without all the subroutine-
3772 specific fields set, and crashes would soon ensue.
3773 Where the two must be separate, initially set the type to "CXt_NULL" or
3775 "CXt_BLOCK", and later change it to "CXt_foo" when doing the
3776 "cx_pushfoo". This is exactly what "pp_enteriter" does, once it's
3777 determined which type of loop it's pushing.
3778 Popping contexts
3780 Contexts are popped using cx_popsub() etc. and cx_popblock(). Note
3781 however, that unlike "cx_pushblock", neither of these functions
3782 actually decrement the current context stack index; this is done
3783 separately using CX_POP().
3784 There are two main ways that contexts are popped. During normal
3786 execution as scopes are exited, functions like "pp_leave",
3787 "pp_leaveloop" and "pp_leavesub" process and pop just one context using
3788 "cx_popfoo" and "cx_popblock". On the other hand, things like
3789 "pp_return" and "next" may have to pop back several scopes until a sub
3790 or loop context is found, and exceptions (such as "die") need to pop
3791 back contexts until an eval context is found. Both of these are
3792 accomplished by dounwind(), which is capable of processing and popping
3793 all contexts above the target one.
3794 Here is a typical example of context popping, as found in "pp_leavesub"
3796 (simplified slightly):
3797 U8 gimme;
3799 PERL_CONTEXT *cx;
3800 SV **oldsp;
3801 OP *retop;
3802 cx = CX_CUR();
3804 gimme = cx->blk_gimme;
3806 oldsp = PL_stack_base + cx->blk_oldsp; /* last arg of previous frame */
3807 if (gimme == G_VOID)
3809 PL_stack_sp = oldsp;
3810 else
3811 leave_adjust_stacks(oldsp, oldsp, gimme, 0);
3812 CX_LEAVE_SCOPE(cx);
3814 cx_popsub(cx);
3815 cx_popblock(cx);
3816 retop = cx->blk_sub.retop;
3817 CX_POP(cx);
3818 return retop;
3820 The steps above are in a very specific order, designed to be the
3822 reverse order of when the context was pushed. The first thing to do is
3823 to copy and/or protect any return arguments and free any temps in the
3824 current scope. Scope exits like an rvalue sub normally return a mortal
3825 copy of their return args (as opposed to lvalue subs). It is important
3826 to make this copy before the save stack is popped or variables are
3827 restored, or bad things like the following can happen:
3828 sub f { my $x =...; $x } # $x freed before we get to copy it
3830 sub f { /(...)/; $1 } # PL_curpm restored before $1 copied
3831 Although we wish to free any temps at the same time, we have to be
3833 careful not to free any temps which are keeping return args alive; nor
3834 to free the temps we have just created while mortal copying return
3835 args. Fortunately, leave_adjust_stacks() is capable of making mortal
3836 copies of return args, shifting args down the stack, and only
3837 processing those entries on the temps stack that are safe to do so.
3838 In void context no args are returned, so it's more efficient to skip
3840 calling leave_adjust_stacks(). Also in void context, a "nextstate" op
3841 is likely to be imminently called which will do a "FREETMPS", so
3842 there's no need to do that either.
3843 The next step is to pop savestack entries: CX_LEAVE_SCOPE(cx) is just
3845 defined as LEAVE_SCOPE(cx->blk_oldsaveix). Note that during the
3846 popping, it's possible for perl to call destructors, call "STORE" to
3847 undo localisations of tied vars, and so on. Any of these can die or
3848 call exit(). In this case, dounwind() will be called, and the current
3849 context stack frame will be re-processed. Thus it is vital that all
3850 steps in popping a context are done in such a way to support
3851 reentrancy. The other alternative, of decrementing "cxstack_ix" before
3852 processing the frame, would lead to leaks and the like if something
3853 died halfway through, or overwriting of the current frame.
3854 "CX_LEAVE_SCOPE" itself is safely re-entrant: if only half the
3856 savestack items have been popped before dying and getting trapped by
3857 eval, then the "CX_LEAVE_SCOPE"s in "dounwind" or "pp_leaveeval" will
3858 continue where the first one left off.
3859 The next step is the type-specific context processing; in this case
3861 "cx_popsub". In part, this looks like:
3862 cv = cx->blk_sub.cv;
3864 CvDEPTH(cv) = cx->blk_sub.olddepth;
3865 cx->blk_sub.cv = NULL;
3866 SvREFCNT_dec(cv);
3867 where its processing the just-executed CV. Note that before it
3869 decrements the CV's reference count, it nulls the "blk_sub.cv". This
3870 means that if it re-enters, the CV won't be freed twice. It also means
3871 that you can't rely on such type-specific fields having useful values
3872 after the return from "cx_popfoo".
3873 Next, "cx_popblock" restores all the various interpreter vars to their
3875 previous values or previous high water marks; it expands to:
3876 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
3878 PL_scopestack_ix = cx->blk_oldscopesp;
3879 PL_curpm = cx->blk_oldpm;
3880 PL_curcop = cx->blk_oldcop;
3881 PL_tmps_floor = cx->blk_old_tmpsfloor;
3882 Note that it doesn't restore "PL_stack_sp"; as mentioned earlier, which
3884 value to restore it to depends on the context type (specifically "for
3885 (list) {}"), and what args (if any) it returns; and that will already
3886 have been sorted out earlier by leave_adjust_stacks().
3887 Finally, the context stack pointer is actually decremented by
3889 CX_POP(cx). After this point, it's possible that that the current
3890 context frame could be overwritten by other contexts being pushed.
3891 Although things like ties and "DESTROY" are supposed to work within a
3892 new context stack, it's best not to assume this. Indeed on debugging
3893 builds, CX_POP(cx) deliberately sets "cx" to null to detect code that
3894 is still relying on the field values in that context frame. Note in the
3895 pp_leavesub() example above, we grab "blk_sub.retop" before calling
3896 "CX_POP".
3897 Redoing contexts
3899 Finally, there is cx_topblock(cx), which acts like a super-"nextstate"
3900 as regards to resetting various vars to their base values. It is used
3901 in places like "pp_next", "pp_redo" and "pp_goto" where rather than
3902 exiting a scope, we want to re-initialise the scope. As well as
3903 resetting "PL_stack_sp" like "nextstate", it also resets
3904 "PL_markstack_ptr", "PL_scopestack_ix" and "PL_curpm". Note that it
3905 doesn't do a "FREETMPS".
3906
3908 Note: this section describes a non-public internal API that is subject
3909 to change without notice.
3910 Perl's internal error-handling mechanisms implement "die" (and its
3912 internal equivalents) using longjmp. If this occurs during lexing,
3913 parsing or compilation, we must ensure that any ops allocated as part
3914 of the compilation process are freed. (Older Perl versions did not
3915 adequately handle this situation: when failing a parse, they would leak
3916 ops that were stored in C "auto" variables and not linked anywhere
3917 else.)
3918 To handle this situation, Perl uses op slabs that are attached to the
3920 currently-compiling CV. A slab is a chunk of allocated memory. New ops
3921 are allocated as regions of the slab. If the slab fills up, a new one
3922 is created (and linked from the previous one). When an error occurs and
3923 the CV is freed, any ops remaining are freed.
3924 Each op is preceded by two pointers: one points to the next op in the
3926 slab, and the other points to the slab that owns it. The next-op
3927 pointer is needed so that Perl can iterate over a slab and free all its
3928 ops. (Op structures are of different sizes, so the slab's ops can't
3929 merely be treated as a dense array.) The slab pointer is needed for
3930 accessing a reference count on the slab: when the last op on a slab is
3931 freed, the slab itself is freed.
3932 The slab allocator puts the ops at the end of the slab first. This will
3934 tend to allocate the leaves of the op tree first, and the layout will
3935 therefore hopefully be cache-friendly. In addition, this means that
3936 there's no need to store the size of the slab (see below on why slabs
3937 vary in size), because Perl can follow pointers to find the last op.
3938 It might seem possible to eliminate slab reference counts altogether,
3940 by having all ops implicitly attached to "PL_compcv" when allocated and
3941 freed when the CV is freed. That would also allow "op_free" to skip
3942 "FreeOp" altogether, and thus free ops faster. But that doesn't work in
3943 those cases where ops need to survive beyond their CVs, such as re-
3944 evals.
3945 The CV also has to have a reference count on the slab. Sometimes the
3947 first op created is immediately freed. If the reference count of the
3948 slab reaches 0, then it will be freed with the CV still pointing to it.
3949 CVs use the "CVf_SLABBED" flag to indicate that the CV has a reference
3951 count on the slab. When this flag is set, the slab is accessible via
3952 "CvSTART" when "CvROOT" is not set, or by subtracting two pointers
3953 "(2*sizeof(I32 *))" from "CvROOT" when it is set. The alternative to
3954 this approach of sneaking the slab into "CvSTART" during compilation
3955 would be to enlarge the "xpvcv" struct by another pointer. But that
3956 would make all CVs larger, even though slab-based op freeing is
3957 typically of benefit only for programs that make significant use of
3958 string eval.
3959 When the "CVf_SLABBED" flag is set, the CV takes responsibility for
3961 freeing the slab. If "CvROOT" is not set when the CV is freed or
3962 undeffed, it is assumed that a compilation error has occurred, so the
3963 op slab is traversed and all the ops are freed.
3964 Under normal circumstances, the CV forgets about its slab (decrementing
3966 the reference count) when the root is attached. So the slab reference
3967 counting that happens when ops are freed takes care of freeing the
3968 slab. In some cases, the CV is told to forget about the slab
3969 ("cv_forget_slab") precisely so that the ops can survive after the CV
3970 is done away with.
3971 Forgetting the slab when the root is attached is not strictly
3973 necessary, but avoids potential problems with "CvROOT" being written
3974 over. There is code all over the place, both in core and on CPAN, that
3975 does things with "CvROOT", so forgetting the slab makes things more
3976 robust and avoids potential problems.
3977 Since the CV takes ownership of its slab when flagged, that flag is
3979 never copied when a CV is cloned, as one CV could free a slab that
3980 another CV still points to, since forced freeing of ops ignores the
3981 reference count (but asserts that it looks right).
3982 To avoid slab fragmentation, freed ops are marked as freed and attached
3984 to the slab's freed chain (an idea stolen from DBM::Deep). Those freed
3985 ops are reused when possible. Not reusing freed ops would be simpler,
3986 but it would result in significantly higher memory usage for programs
3987 with large "if (DEBUG) {...}" blocks.
3988 "SAVEFREEOP" is slightly problematic under this scheme. Sometimes it
3990 can cause an op to be freed after its CV. If the CV has forcibly freed
3991 the ops on its slab and the slab itself, then we will be fiddling with
3992 a freed slab. Making "SAVEFREEOP" a no-op doesn't help, as sometimes an
3993 op can be savefreed when there is no compilation error, so the op would
3994 never be freed. It holds a reference count on the slab, so the whole
3995 slab would leak. So "SAVEFREEOP" now sets a special flag on the op
3996 ("->op_savefree"). The forced freeing of ops after a compilation error
3997 won't free any ops thus marked.
3998 Since many pieces of code create tiny subroutines consisting of only a
4000 few ops, and since a huge slab would be quite a bit of baggage for
4001 those to carry around, the first slab is always very small. To avoid
4002 allocating too many slabs for a single CV, each subsequent slab is
4003 twice the size of the previous.
4004 Smartmatch expects to be able to allocate an op at run time, run it,
4006 and then throw it away. For that to work the op is simply malloced when
4007 "PL_compcv" hasn't been set up. So all slab-allocated ops are marked as
4008 such ("->op_slabbed"), to distinguish them from malloced ops.
4009
4011 Until May 1997, this document was maintained by Jeff Okamoto
4012 <okamoto@corp.hp.com>. It is now maintained as part of Perl itself by
4013 the Perl 5 Porters <perl5-porters@perl.org>.
4014 With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
4016 Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
4017 Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
4018 Stephen McCamant, and Gurusamy Sarathy.
4019
4021 perlapi, perlintern, perlxs, perlembed
4022perl v5.38.2 2023-11-30 PERLGUTS(1)