1Bytes(3) OCaml library Bytes(3)
2
6 Bytes - Byte sequence operations.
7
9 Module Bytes
10
12 Module Bytes
13 : sig end
14 Byte sequence operations.
17 A byte sequence is a mutable data structure that contains a
19 fixed-length sequence of bytes. Each byte can be indexed in constant
20 time for reading or writing.
21 Given a byte sequence s of length l , we can access each of the l bytes
23 of s via its index in the sequence. Indexes start at 0 , and we will
24 call an index valid in s if it falls within the range [0...l-1] (inclu‐
25 sive). A position is the point between two bytes or at the beginning or
26 end of the sequence. We call a position valid in s if it falls within
27 the range [0...l] (inclusive). Note that the byte at index n is between
28 positions n and n+1 .
29 Two parameters start and len are said to designate a valid range of s
31 if len >= 0 and start and start+len are valid positions in s .
32 Byte sequences can be modified in place, for instance via the set and
34 blit functions described below. See also strings (module String ),
35 which are almost the same data structure, but cannot be modified in
36 place.
37 Bytes are represented by the OCaml type char .
39 The labeled version of this module can be used as described in the Std‐
41 Labels module.
42 Since 4.02.0
45 val length : bytes -> int
52 Return the length (number of bytes) of the argument.
54 val get : bytes -> int -> char
58 get s n returns the byte at index n in argument s .
61 Raises Invalid_argument if n is not a valid index in s .
64 val set : bytes -> int -> char -> unit
68 set s n c modifies s in place, replacing the byte at index n with c .
71 Raises Invalid_argument if n is not a valid index in s .
74 val create : int -> bytes
78 create n returns a new byte sequence of length n . The sequence is
81 uninitialized and contains arbitrary bytes.
82 Raises Invalid_argument if n < 0 or n > Sys.max_string_length .
85 val make : int -> char -> bytes
89 make n c returns a new byte sequence of length n , filled with the byte
92 c .
93 Raises Invalid_argument if n < 0 or n > Sys.max_string_length .
96 val init : int -> (int -> char) -> bytes
100 init n f returns a fresh byte sequence of length n , with character i
103 initialized to the result of f i (in increasing index order).
104 Raises Invalid_argument if n < 0 or n > Sys.max_string_length .
107 val empty : bytes
111 A byte sequence of size 0.
113 val copy : bytes -> bytes
117 Return a new byte sequence that contains the same bytes as the argu‐
119 ment.
120 val of_string : string -> bytes
124 Return a new byte sequence that contains the same bytes as the given
126 string.
127 val to_string : bytes -> string
131 Return a new string that contains the same bytes as the given byte se‐
133 quence.
134 val sub : bytes -> int -> int -> bytes
138 sub s pos len returns a new byte sequence of length len , containing
141 the subsequence of s that starts at position pos and has length len .
142 Raises Invalid_argument if pos and len do not designate a valid range
145 of s .
146 val sub_string : bytes -> int -> int -> string
150 Same as Bytes.sub but return a string instead of a byte sequence.
152 val extend : bytes -> int -> int -> bytes
156 extend s left right returns a new byte sequence that contains the bytes
159 of s , with left uninitialized bytes prepended and right uninitialized
160 bytes appended to it. If left or right is negative, then bytes are re‐
161 moved (instead of appended) from the corresponding side of s .
162 Since 4.05.0 in BytesLabels
165 Raises Invalid_argument if the result length is negative or longer than
168 Sys.max_string_length bytes.
169 val fill : bytes -> int -> int -> char -> unit
173 fill s pos len c modifies s in place, replacing len characters with c ,
176 starting at pos .
177 Raises Invalid_argument if pos and len do not designate a valid range
180 of s .
181 val blit : bytes -> int -> bytes -> int -> int -> unit
185 blit src src_pos dst dst_pos len copies len bytes from sequence src ,
188 starting at index src_pos , to sequence dst , starting at index dst_pos
189 . It works correctly even if src and dst are the same byte sequence,
190 and the source and destination intervals overlap.
191 Raises Invalid_argument if src_pos and len do not designate a valid
194 range of src , or if dst_pos and len do not designate a valid range of
195 dst .
196 val blit_string : string -> int -> bytes -> int -> int -> unit
200 blit src src_pos dst dst_pos len copies len bytes from string src ,
203 starting at index src_pos , to byte sequence dst , starting at index
204 dst_pos .
205 Since 4.05.0 in BytesLabels
208 Raises Invalid_argument if src_pos and len do not designate a valid
211 range of src , or if dst_pos and len do not designate a valid range of
212 dst .
213 val concat : bytes -> bytes list -> bytes
217 concat sep sl concatenates the list of byte sequences sl , inserting
220 the separator byte sequence sep between each, and returns the result as
221 a new byte sequence.
222 Raises Invalid_argument if the result is longer than
225 Sys.max_string_length bytes.
226 val cat : bytes -> bytes -> bytes
230 cat s1 s2 concatenates s1 and s2 and returns the result as a new byte
233 sequence.
234 Since 4.05.0 in BytesLabels
237 Raises Invalid_argument if the result is longer than
240 Sys.max_string_length bytes.
241 val iter : (char -> unit) -> bytes -> unit
245 iter f s applies function f in turn to all the bytes of s . It is
248 equivalent to f (get s 0); f (get s 1); ...; f (get s
249 (length s - 1)); () .
250 val iteri : (int -> char -> unit) -> bytes -> unit
254 Same as Bytes.iter , but the function is applied to the index of the
256 byte as first argument and the byte itself as second argument.
257 val map : (char -> char) -> bytes -> bytes
261 map f s applies function f in turn to all the bytes of s (in increasing
264 index order) and stores the resulting bytes in a new sequence that is
265 returned as the result.
266 val mapi : (int -> char -> char) -> bytes -> bytes
270 mapi f s calls f with each character of s and its index (in increasing
273 index order) and stores the resulting bytes in a new sequence that is
274 returned as the result.
275 val fold_left : ('a -> char -> 'a) -> 'a -> bytes -> 'a
279 fold_left f x s computes f (... (f (f x (get s 0)) (get s 1)) ...) (get
282 s (n-1)) , where n is the length of s .
283 Since 4.13.0
286 val fold_right : (char -> 'a -> 'a) -> bytes -> 'a -> 'a
290 fold_right f s x computes f (get s 0) (f (get s 1) ( ... (f (get s
293 (n-1)) x) ...)) , where n is the length of s .
294 Since 4.13.0
297 val for_all : (char -> bool) -> bytes -> bool
301 for_all p s checks if all characters in s satisfy the predicate p .
304 Since 4.13.0
307 val exists : (char -> bool) -> bytes -> bool
311 exists p s checks if at least one character of s satisfies the predi‐
314 cate p .
315 Since 4.13.0
318 val trim : bytes -> bytes
322 Return a copy of the argument, without leading and trailing whitespace.
324 The bytes regarded as whitespace are the ASCII characters ' ' , '\012'
325 , '\n' , '\r' , and '\t' .
326 val escaped : bytes -> bytes
330 Return a copy of the argument, with special characters represented by
332 escape sequences, following the lexical conventions of OCaml. All
333 characters outside the ASCII printable range (32..126) are escaped, as
334 well as backslash and double-quote.
335 Raises Invalid_argument if the result is longer than
338 Sys.max_string_length bytes.
339 val index : bytes -> char -> int
343 index s c returns the index of the first occurrence of byte c in s .
346 Raises Not_found if c does not occur in s .
349 val index_opt : bytes -> char -> int option
353 index_opt s c returns the index of the first occurrence of byte c in s
356 or None if c does not occur in s .
357 Since 4.05
360 val rindex : bytes -> char -> int
364 rindex s c returns the index of the last occurrence of byte c in s .
367 Raises Not_found if c does not occur in s .
370 val rindex_opt : bytes -> char -> int option
374 rindex_opt s c returns the index of the last occurrence of byte c in s
377 or None if c does not occur in s .
378 Since 4.05
381 val index_from : bytes -> int -> char -> int
385 index_from s i c returns the index of the first occurrence of byte c in
388 s after position i . index s c is equivalent to index_from s 0 c .
389 Raises Invalid_argument if i is not a valid position in s .
392 Raises Not_found if c does not occur in s after position i .
395 val index_from_opt : bytes -> int -> char -> int option
399 index_from_opt s i c returns the index of the first occurrence of byte
402 c in s after position i or None if c does not occur in s after position
403 i . index_opt s c is equivalent to index_from_opt s 0 c .
404 Since 4.05
407 Raises Invalid_argument if i is not a valid position in s .
410 val rindex_from : bytes -> int -> char -> int
414 rindex_from s i c returns the index of the last occurrence of byte c in
417 s before position i+1 . rindex s c is equivalent to rindex_from s
418 (length s - 1) c .
419 Raises Invalid_argument if i+1 is not a valid position in s .
422 Raises Not_found if c does not occur in s before position i+1 .
425 val rindex_from_opt : bytes -> int -> char -> int option
429 rindex_from_opt s i c returns the index of the last occurrence of byte
432 c in s before position i+1 or None if c does not occur in s before po‐
433 sition i+1 . rindex_opt s c is equivalent to rindex_from s (length s -
434 1) c .
435 Since 4.05
438 Raises Invalid_argument if i+1 is not a valid position in s .
441 val contains : bytes -> char -> bool
445 contains s c tests if byte c appears in s .
448 val contains_from : bytes -> int -> char -> bool
452 contains_from s start c tests if byte c appears in s after position
455 start . contains s c is equivalent to contains_from
456 s 0 c .
457 Raises Invalid_argument if start is not a valid position in s .
460 val rcontains_from : bytes -> int -> char -> bool
464 rcontains_from s stop c tests if byte c appears in s before position
467 stop+1 .
468 Raises Invalid_argument if stop < 0 or stop+1 is not a valid position
471 in s .
472 val uppercase_ascii : bytes -> bytes
476 Return a copy of the argument, with all lowercase letters translated to
478 uppercase, using the US-ASCII character set.
479 Since 4.03.0 (4.05.0 in BytesLabels)
482 val lowercase_ascii : bytes -> bytes
486 Return a copy of the argument, with all uppercase letters translated to
488 lowercase, using the US-ASCII character set.
489 Since 4.03.0 (4.05.0 in BytesLabels)
492 val capitalize_ascii : bytes -> bytes
496 Return a copy of the argument, with the first character set to upper‐
498 case, using the US-ASCII character set.
499 Since 4.03.0 (4.05.0 in BytesLabels)
502 val uncapitalize_ascii : bytes -> bytes
506 Return a copy of the argument, with the first character set to lower‐
508 case, using the US-ASCII character set.
509 Since 4.03.0 (4.05.0 in BytesLabels)
512 type t = bytes
515 An alias for the type of byte sequences.
518 val compare : t -> t -> int
522 The comparison function for byte sequences, with the same specification
524 as compare . Along with the type t , this function compare allows the
525 module Bytes to be passed as argument to the functors Set.Make and
526 Map.Make .
527 val equal : t -> t -> bool
531 The equality function for byte sequences.
533 Since 4.03.0 (4.05.0 in BytesLabels)
536 val starts_with : prefix:bytes -> bytes -> bool
540 starts_with ~prefix s is true if and only if s starts with prefix .
543 Since 4.13.0
546 val ends_with : suffix:bytes -> bytes -> bool
550 ends_with ~suffix s is true if and only if s ends with suffix .
553 Since 4.13.0
556 Unsafe conversions (for advanced users)
561 This section describes unsafe, low-level conversion functions between
562 bytes and string . They do not copy the internal data; used improperly,
563 they can break the immutability invariant on strings provided by the
564 -safe-string option. They are available for expert library authors, but
565 for most purposes you should use the always-correct Bytes.to_string and
566 Bytes.of_string instead.
567 val unsafe_to_string : bytes -> string
569 Unsafely convert a byte sequence into a string.
571 To reason about the use of unsafe_to_string , it is convenient to con‐
573 sider an "ownership" discipline. A piece of code that manipulates some
574 data "owns" it; there are several disjoint ownership modes, including:
575 -Unique ownership: the data may be accessed and mutated
577 -Shared ownership: the data has several owners, that may only access
579 it, not mutate it.
580 Unique ownership is linear: passing the data to another piece of code
582 means giving up ownership (we cannot write the data again). A unique
583 owner may decide to make the data shared (giving up mutation rights on
584 it), but shared data may not become uniquely-owned again.
585 unsafe_to_string s can only be used when the caller owns the byte se‐
588 quence s -- either uniquely or as shared immutable data. The caller
589 gives up ownership of s , and gains ownership of the returned string.
590 There are two valid use-cases that respect this ownership discipline:
592 1. Creating a string by initializing and mutating a byte sequence that
594 is never changed after initialization is performed.
595 let string_init len f : string =
598 let s = Bytes.create len in
599 for i = 0 to len - 1 do Bytes.set s i (f i) done;
600 Bytes.unsafe_to_string s
601 This function is safe because the byte sequence s will never be ac‐
604 cessed or mutated after unsafe_to_string is called. The string_init
605 code gives up ownership of s , and returns the ownership of the result‐
606 ing string to its caller.
607 Note that it would be unsafe if s was passed as an additional parameter
609 to the function f as it could escape this way and be mutated in the fu‐
610 ture -- string_init would give up ownership of s to pass it to f , and
611 could not call unsafe_to_string safely.
612 We have provided the String.init , String.map and String.mapi functions
614 to cover most cases of building new strings. You should prefer those
615 over to_string or unsafe_to_string whenever applicable.
616 2. Temporarily giving ownership of a byte sequence to a function that
618 expects a uniquely owned string and returns ownership back, so that we
619 can mutate the sequence again after the call ended.
620 let bytes_length (s : bytes) =
623 String.length (Bytes.unsafe_to_string s)
624 In this use-case, we do not promise that s will never be mutated after
627 the call to bytes_length s . The String.length function temporarily
628 borrows unique ownership of the byte sequence (and sees it as a string
629 ), but returns this ownership back to the caller, which may assume that
630 s is still a valid byte sequence after the call. Note that this is only
631 correct because we know that String.length does not capture its argu‐
632 ment -- it could escape by a side-channel such as a memoization combi‐
633 nator.
634 The caller may not mutate s while the string is borrowed (it has tempo‐
636 rarily given up ownership). This affects concurrent programs, but also
637 higher-order functions: if String.length returned a closure to be
638 called later, s should not be mutated until this closure is fully ap‐
639 plied and returns ownership.
640 val unsafe_of_string : string -> bytes
644 Unsafely convert a shared string to a byte sequence that should not be
646 mutated.
647 The same ownership discipline that makes unsafe_to_string correct ap‐
649 plies to unsafe_of_string : you may use it if you were the owner of the
650 string value, and you will own the return bytes in the same mode.
651 In practice, unique ownership of string values is extremely difficult
653 to reason about correctly. You should always assume strings are shared,
654 never uniquely owned.
655 For example, string literals are implicitly shared by the compiler, so
657 you never uniquely own them.
658 let incorrect = Bytes.unsafe_of_string "hello"
661 let s = Bytes.of_string "hello"
662 The first declaration is incorrect, because the string literal "hello"
665 could be shared by the compiler with other parts of the program, and
666 mutating incorrect is a bug. You must always use the second version,
667 which performs a copy and is thus correct.
668 Assuming unique ownership of strings that are not string literals, but
670 are (partly) built from string literals, is also incorrect. For exam‐
671 ple, mutating unsafe_of_string ("foo" ^ s) could mutate the shared
672 string "foo" -- assuming a rope-like representation of strings. More
673 generally, functions operating on strings will assume shared ownership,
674 they do not preserve unique ownership. It is thus incorrect to assume
675 unique ownership of the result of unsafe_of_string .
676 The only case we have reasonable confidence is safe is if the produced
678 bytes is shared -- used as an immutable byte sequence. This is possibly
679 useful for incremental migration of low-level programs that manipulate
680 immutable sequences of bytes (for example Marshal.from_bytes ) and pre‐
681 viously used the string type for this purpose.
682 val split_on_char : char -> bytes -> bytes list
686 split_on_char sep s returns the list of all (possibly empty) subse‐
689 quences of s that are delimited by the sep character.
690 The function's output is specified by the following invariants:
692 -The list is not empty.
695 -Concatenating its elements using sep as a separator returns a byte se‐
697 quence equal to the input ( Bytes.concat (Bytes.make 1 sep)
698 (Bytes.split_on_char sep s) = s ).
699 -No byte sequence in the result contains the sep character.
701 Since 4.13.0
705 Iterators
710 val to_seq : t -> char Seq.t
711 Iterate on the string, in increasing index order. Modifications of the
713 string during iteration will be reflected in the sequence.
714 Since 4.07
717 val to_seqi : t -> (int * char) Seq.t
721 Iterate on the string, in increasing order, yielding indices along
723 chars
724 Since 4.07
727 val of_seq : char Seq.t -> t
731 Create a string from the generator
733 Since 4.07
736 UTF codecs and validations
741 UTF-8
742 val get_utf_8_uchar : t -> int -> Uchar.utf_decode
743 get_utf_8_uchar b i decodes an UTF-8 character at index i in b .
746 val set_utf_8_uchar : t -> int -> Uchar.t -> int
750 set_utf_8_uchar b i u UTF-8 encodes u at index i in b and returns the
753 number of bytes n that were written starting at i . If n is 0 there was
754 not enough space to encode u at i and b was left untouched. Otherwise a
755 new character can be encoded at i + n .
756 val is_valid_utf_8 : t -> bool
760 is_valid_utf_8 b is true if and only if b contains valid UTF-8 data.
763 UTF-16BE
768 val get_utf_16be_uchar : t -> int -> Uchar.utf_decode
769 get_utf_16be_uchar b i decodes an UTF-16BE character at index i in b .
772 val set_utf_16be_uchar : t -> int -> Uchar.t -> int
776 set_utf_16be_uchar b i u UTF-16BE encodes u at index i in b and returns
779 the number of bytes n that were written starting at i . If n is 0 there
780 was not enough space to encode u at i and b was left untouched. Other‐
781 wise a new character can be encoded at i + n .
782 val is_valid_utf_16be : t -> bool
786 is_valid_utf_16be b is true if and only if b contains valid UTF-16BE
789 data.
790 UTF-16LE
795 val get_utf_16le_uchar : t -> int -> Uchar.utf_decode
796 get_utf_16le_uchar b i decodes an UTF-16LE character at index i in b .
799 val set_utf_16le_uchar : t -> int -> Uchar.t -> int
803 set_utf_16le_uchar b i u UTF-16LE encodes u at index i in b and returns
806 the number of bytes n that were written starting at i . If n is 0 there
807 was not enough space to encode u at i and b was left untouched. Other‐
808 wise a new character can be encoded at i + n .
809 val is_valid_utf_16le : t -> bool
813 is_valid_utf_16le b is true if and only if b contains valid UTF-16LE
816 data.
817 Binary encoding/decoding of integers
822 The functions in this section binary encode and decode integers to and
823 from byte sequences.
824 All following functions raise Invalid_argument if the space needed at
826 index i to decode or encode the integer is not available.
827 Little-endian (resp. big-endian) encoding means that least (resp. most)
829 significant bytes are stored first. Big-endian is also known as net‐
830 work byte order. Native-endian encoding is either little-endian or
831 big-endian depending on Sys.big_endian .
832 32-bit and 64-bit integers are represented by the int32 and int64
834 types, which can be interpreted either as signed or unsigned numbers.
835 8-bit and 16-bit integers are represented by the int type, which has
837 more bits than the binary encoding. These extra bits are handled as
838 follows:
839 -Functions that decode signed (resp. unsigned) 8-bit or 16-bit integers
841 represented by int values sign-extend (resp. zero-extend) their result.
842 -Functions that encode 8-bit or 16-bit integers represented by int val‐
844 ues truncate their input to their least significant bytes.
845 val get_uint8 : bytes -> int -> int
848 get_uint8 b i is b 's unsigned 8-bit integer starting at byte index i .
851 Since 4.08
854 val get_int8 : bytes -> int -> int
858 get_int8 b i is b 's signed 8-bit integer starting at byte index i .
861 Since 4.08
864 val get_uint16_ne : bytes -> int -> int
868 get_uint16_ne b i is b 's native-endian unsigned 16-bit integer start‐
871 ing at byte index i .
872 Since 4.08
875 val get_uint16_be : bytes -> int -> int
879 get_uint16_be b i is b 's big-endian unsigned 16-bit integer starting
882 at byte index i .
883 Since 4.08
886 val get_uint16_le : bytes -> int -> int
890 get_uint16_le b i is b 's little-endian unsigned 16-bit integer start‐
893 ing at byte index i .
894 Since 4.08
897 val get_int16_ne : bytes -> int -> int
901 get_int16_ne b i is b 's native-endian signed 16-bit integer starting
904 at byte index i .
905 Since 4.08
908 val get_int16_be : bytes -> int -> int
912 get_int16_be b i is b 's big-endian signed 16-bit integer starting at
915 byte index i .
916 Since 4.08
919 val get_int16_le : bytes -> int -> int
923 get_int16_le b i is b 's little-endian signed 16-bit integer starting
926 at byte index i .
927 Since 4.08
930 val get_int32_ne : bytes -> int -> int32
934 get_int32_ne b i is b 's native-endian 32-bit integer starting at byte
937 index i .
938 Since 4.08
941 val get_int32_be : bytes -> int -> int32
945 get_int32_be b i is b 's big-endian 32-bit integer starting at byte in‐
948 dex i .
949 Since 4.08
952 val get_int32_le : bytes -> int -> int32
956 get_int32_le b i is b 's little-endian 32-bit integer starting at byte
959 index i .
960 Since 4.08
963 val get_int64_ne : bytes -> int -> int64
967 get_int64_ne b i is b 's native-endian 64-bit integer starting at byte
970 index i .
971 Since 4.08
974 val get_int64_be : bytes -> int -> int64
978 get_int64_be b i is b 's big-endian 64-bit integer starting at byte in‐
981 dex i .
982 Since 4.08
985 val get_int64_le : bytes -> int -> int64
989 get_int64_le b i is b 's little-endian 64-bit integer starting at byte
992 index i .
993 Since 4.08
996 val set_uint8 : bytes -> int -> int -> unit
1000 set_uint8 b i v sets b 's unsigned 8-bit integer starting at byte index
1003 i to v .
1004 Since 4.08
1007 val set_int8 : bytes -> int -> int -> unit
1011 set_int8 b i v sets b 's signed 8-bit integer starting at byte index i
1014 to v .
1015 Since 4.08
1018 val set_uint16_ne : bytes -> int -> int -> unit
1022 set_uint16_ne b i v sets b 's native-endian unsigned 16-bit integer
1025 starting at byte index i to v .
1026 Since 4.08
1029 val set_uint16_be : bytes -> int -> int -> unit
1033 set_uint16_be b i v sets b 's big-endian unsigned 16-bit integer start‐
1036 ing at byte index i to v .
1037 Since 4.08
1040 val set_uint16_le : bytes -> int -> int -> unit
1044 set_uint16_le b i v sets b 's little-endian unsigned 16-bit integer
1047 starting at byte index i to v .
1048 Since 4.08
1051 val set_int16_ne : bytes -> int -> int -> unit
1055 set_int16_ne b i v sets b 's native-endian signed 16-bit integer start‐
1058 ing at byte index i to v .
1059 Since 4.08
1062 val set_int16_be : bytes -> int -> int -> unit
1066 set_int16_be b i v sets b 's big-endian signed 16-bit integer starting
1069 at byte index i to v .
1070 Since 4.08
1073 val set_int16_le : bytes -> int -> int -> unit
1077 set_int16_le b i v sets b 's little-endian signed 16-bit integer start‐
1080 ing at byte index i to v .
1081 Since 4.08
1084 val set_int32_ne : bytes -> int -> int32 -> unit
1088 set_int32_ne b i v sets b 's native-endian 32-bit integer starting at
1091 byte index i to v .
1092 Since 4.08
1095 val set_int32_be : bytes -> int -> int32 -> unit
1099 set_int32_be b i v sets b 's big-endian 32-bit integer starting at byte
1102 index i to v .
1103 Since 4.08
1106 val set_int32_le : bytes -> int -> int32 -> unit
1110 set_int32_le b i v sets b 's little-endian 32-bit integer starting at
1113 byte index i to v .
1114 Since 4.08
1117 val set_int64_ne : bytes -> int -> int64 -> unit
1121 set_int64_ne b i v sets b 's native-endian 64-bit integer starting at
1124 byte index i to v .
1125 Since 4.08
1128 val set_int64_be : bytes -> int -> int64 -> unit
1132 set_int64_be b i v sets b 's big-endian 64-bit integer starting at byte
1135 index i to v .
1136 Since 4.08
1139 val set_int64_le : bytes -> int -> int64 -> unit
1143 set_int64_le b i v sets b 's little-endian 64-bit integer starting at
1146 byte index i to v .
1147 Since 4.08
1150 Byte sequences and concurrency safety
1155 Care must be taken when concurrently accessing byte sequences from mul‐
1156 tiple domains: accessing a byte sequence will never crash a program,
1157 but unsynchronized accesses might yield surprising (non-sequen‐
1158 tially-consistent) results.
1159 Atomicity
1162 Every byte sequence operation that accesses more than one byte is not
1163 atomic. This includes iteration and scanning.
1164 For example, consider the following program:
1166 let size = 100_000_000
1167 let b = Bytes.make size ' '
1168 let update b f () =
1169 Bytes.iteri (fun i x -> Bytes.set b i (Char.chr (f (Char.code x)))) b
1170 let d1 = Domain.spawn (update b (fun x -> x + 1))
1171 let d2 = Domain.spawn (update b (fun x -> 2 * x + 1))
1172 let () = Domain.join d1; Domain.join d2
1173 the bytes sequence b may contain a non-deterministic mixture of '!' ,
1175 'A' , 'B' , and 'C' values.
1176 After executing this code, each byte of the sequence b is either '!' ,
1178 'A' , 'B' , or 'C' . If atomicity is required, then the user must im‐
1179 plement their own synchronization (for example, using Mutex.t ).
1180 Data races
1183 If two domains only access disjoint parts of a byte sequence, then the
1184 observed behaviour is the equivalent to some sequential interleaving of
1185 the operations from the two domains.
1186 A data race is said to occur when two domains access the same byte
1188 without synchronization and at least one of the accesses is a write.
1189 In the absence of data races, the observed behaviour is equivalent to
1190 some sequential interleaving of the operations from different domains.
1191 Whenever possible, data races should be avoided by using synchroniza‐
1193 tion to mediate the accesses to the elements of the sequence.
1194 Indeed, in the presence of data races, programs will not crash but the
1196 observed behaviour may not be equivalent to any sequential interleaving
1197 of operations from different domains. Nevertheless, even in the pres‐
1198 ence of data races, a read operation will return the value of some
1199 prior write to that location.
1200 Mixed-size accesses
1203 Another subtle point is that if a data race involves mixed-size writes
1204 and reads to the same location, the order in which those writes and
1205 reads are observed by domains is not specified. For instance, the fol‐
1206 lowing code write sequentially a 32-bit integer and a char to the same
1207 index
1208 let b = Bytes.make 10 '\000'
1209 let d1 = Domain.spawn (fun () -> Bytes.set_int32_ne b 0 100; b.[0] <- 'd' )
1210 In this situation, a domain that observes the write of 'd' to b. 0 is
1213 not guaranteed to also observe the write to indices 1 , 2 , or 3 .
1214OCamldoc 2023-07-20 Bytes(3)