1SQL::Interp(3) User Contributed Perl Documentation SQL::Interp(3)
2
3
4
6 SQL::Interp - Interpolate Perl variables into SQL statements
7
9 use SQL::Interp ':all';
10
11 my ($sql, @bind) = sql_interp 'INSERT INTO table', \%item;
12 my ($sql, @bind) = sql_interp 'UPDATE table SET', \%item, 'WHERE y <> ', \2;
13 my ($sql, @bind) = sql_interp 'DELETE FROM table WHERE y = ', \2;
14
15 # These two select syntax produce the same result
16 my ($sql, @bind) = sql_interp 'SELECT * FROM table WHERE x = ', \$s, 'AND y IN', \@v;
17 my ($sql, @bind) = sql_interp 'SELECT * FROM table WHERE', {x => $s, y => \@v};
18
20 SQL::Interp converts a list of intermixed SQL fragments and variable
21 references into a conventional SQL string and list of bind values
22 suitable for passing onto DBI. This simple technique creates database
23 calls that are simpler to create and easier to read, while still giving
24 you full access to custom SQL.
25
26 SQL::Interp properly binds or escapes variables. This recommended
27 practice safeguards against "SQL injection" attacks. The DBI
28 documentation has several links on the topic.
29
30 Besides the simple techniques shown above, The SQL::Interp integrates
31 directly with DBIx::Simple for an excellent alternative to raw DBI
32 access:
33
34 use DBIx::Simple;
35
36 ...
37
38 my $rows = $db->iquery("
39 SELECT title
40 FROM threads
41 WHERE date > ",\$x," AND subject IN ",\@subjects
42 )->arrays;
43
44 Since DBIx::Simple still allows you complete access to the DBI API,
45 using it as wrapper is recommended for most applications.
46
48 "sql_interp"
49 ($sql, @bind) = sql_interp @params;
50
51 "sql_interp()" is the one central function you need to know.
52 "sql_interp()" strings together the given list of elements and returns
53 both an SQL string ($sql) with placeholders ("?") and a corresponding
54 list of bind values (@bind) suitable for passing to DBI.
55
56 The interpolation list can contain elements of these types:
57
58 * SQL - string containing a raw SQL fragment such as 'SELECT * FROM
59 mytable WHERE'.
60
61 * variable reference - scalarref, arrayref, hashref, or A sql_type()
62 object referring to data to interpolate between the SQL.
63
64 * other interpolation list - an interpolation list can be nested inside
65 another interpolation list. This is possible with the sql() function.
66
67 Interpolation Examples
68
69 The following variable names will be used in the below examples:
70
71 $sref = \3; # scalarref
72 $aref = [1, 2]; # arrayref
73 $href = {m => 1, n => undef}; # hashref
74 $hv = {v => $v, s => $$s}; # hashref containing arrayref
75 $vv = [$v, $v]; # arrayref of arrayref
76 $vh = [$h, $h]; # arrayref of hashref
77
78 Let $x stand for any of these.
79
80 Default scalarref behavior
81
82 A scalarref becomes a single bind value.
83
84 IN: 'foo', $sref, 'bar'
85 OUT: 'foo ? bar', $$sref
86
87 Default hashref behavior
88
89 A hashref becomes a logical AND
90
91 IN: 'WHERE', $href
92 OUT: 'WHERE (m=? AND n IS NULL)', $h->{m},
93
94 IN: 'WHERE', $hv
95 OUT: 'WHERE (v IN (?, ?) AND s = ?)', @$v, $$s
96
97 Default arrayref of (hashref or arrayref) behavior
98
99 This is not commonly used.
100
101 IN: $vv
102 OUT: '(SELECT ?, ? UNION ALL SELECT ?, ?)',
103 map {@$_} @$v
104
105 IN: $vh
106 OUT: '(SELECT ? as m, ? as n UNION ALL
107 SELECT ?, ?)',
108 $vh->[0]->{m}, $vh->[0]->{n},
109 $vh->[1]->{m}, $vh->[1]->{n}
110
111 # Typical usage:
112 IN: $x
113 IN: $x, 'UNION [ALL|DISTINCT]', $x
114 IN: 'INSERT INTO mytable', $x
115 IN: 'SELECT * FROM mytable WHERE x IN', $x
116
117 Context ('IN', $x)
118
119 A scalarref or arrayref can used to form an "IN" clause. As a
120 convenience, a reference to an arrayref is also accepted. This way,
121 you can simply provide a reference to a value which may be a single-
122 valued scalar or a multi-valued arrayref.
123
124 IN: 'WHERE x IN', $aref
125 OUT: 'WHERE x IN (?, ?)', @$aref
126
127 IN: 'WHERE x IN', $sref
128 OUT: 'WHERE x IN (?)', $$sref
129
130 IN: 'WHERE x IN', []
131 OUT: 'WHERE 1=0'
132
133 IN: 'WHERE x NOT IN', []
134 OUT: 'WHERE 1=1'
135
136 Context ('INSERT INTO tablename', $x)
137
138 IN: 'INSERT INTO mytable', $href
139 OUT: 'INSERT INTO mytable (m, n) VALUES(?, ?)', $href->{m}, $href->{n}
140
141 IN: 'INSERT INTO mytable', $aref
142 OUT: 'INSERT INTO mytable VALUES(?, ?)', @$aref;
143
144 IN: 'INSERT INTO mytable', $sref
145 OUT: 'INSERT INTO mytable VALUES(?)', $$sref;
146
147 MySQL's "REPLACE INTO" is supported the same way.
148
149 Context ('SET', $x)
150
151 IN: 'UPDATE mytable SET', $href
152 OUT: 'UPDATE mytable SET m = ?, n = ?', $href->{m}, $href->{n}
153
154 MySQL's "ON DUPLICATE KEY UPDATE" is supported the same way.
155
156 Context ('FROM | JOIN', $x)
157
158 This is not commonly used.
159
160 IN: 'SELECT * FROM', $vv
161 OUT: 'SELECT * FROM
162 (SELECT ?, ? UNION ALL SELECT ?, ?) as t001',
163 map {@$_} @$v
164
165 IN: 'SELECT * FROM', $vh
166 OUT: 'SELECT * FROM
167 (SELECT ? as m, ? as n UNION ALL SELECT ?, ?) as temp001',
168 $vh->[0]->{m}, $vh->[0]->{n},
169 $vh->[1]->{m}, $vh->[1]->{n}
170
171 IN: 'SELECT * FROM', $vv, 'AS t'
172 OUT: 'SELECT * FROM
173 (SELECT ?, ? UNION ALL SELECT ?, ?) AS t',
174 map {@$_} @$v
175
176 # Example usage (where $x and $y are table references):
177 'SELECT * FROM', $x, 'JOIN', $y
178
179 Other Rules
180
181 Whitespace is automatically added between parameters:
182
183 IN: 'UPDATE', 'mytable SET', {x => 2}, 'WHERE y IN', \@colors;
184 OUT: 'UPDATE mytable SET x = ? WHERE y in (?, ?)', 2, @colors
185
186 Variables must be passed as references; otherwise, they will processed
187 as SQL fragments and interpolated verbatim into the result SQL string,
188 negating the security and performance benefits of binding values.
189
190 In contrast, any scalar values inside an arrayref or hashref are by
191 default treated as binding variables, not SQL. The contained elements
192 may be also be sql_type() or sql().
193
195 The "sql_interp" function has a security weakness. Consider these two
196 statements, one easily a typo of the other:
197
198 sql_interp("SELECT * FROM foo WHERE a = ",\$b)
199 sql_interp("SELECT * FROM foo WHERE a = ",$b)
200
201 Both would produce valid SQL, but the first would be secure due to use
202 of bind variables, while the second is potentially insecure, because $b
203 is added directly to the SQL statement. If $b contains a malicious
204 value, it could be used for a SQL injection attack.
205
206 To prevent this accident, we also supply "sql_interp_strict()", which
207 works exactly the same as sql_interp(), but with an additional check
208 that two non-references never appear in a row . If they do, an
209 exception will be thrown.
210
211 This does mean some previously safe-but-valid SQL be need to be
212 rewritten, such as when you are building a complex query from pieces.
213 Here's a contrived example:
214
215 sql_interp("SELECT * FROM ","foo","WHERE a = ",\$b);
216
217 To work under strict mode, you need to concatenate the strings instead:
218
219 sql_interp("SELECT * FROM "."foo"."WHERE a = ",\$b);
220
222 "sql()"
223 sql_interp 'INSERT INTO mytable',
224 {x => $x, y => sql('CURRENT_TIMESTAMP')};
225 # OUT: 'INSERT INTO mytable (x, y) VALUES(?, CURRENT_TIMESTAMP)', $x
226
227 sql() is useful if you want insert raw SQL as a value in an arrayref or
228 hashref.
229
230 "sql_type()"
231 my $sqlvar = sql_type($value_ref, type => $sql_type, %params);
232
233 "sql_type()" provides a general way to represent a binding variable
234 along with metadata. It is necessary in rare applications which you
235 need to explicity give the bind type of a SQL variable.
236
237 $value_ref - variable reference contained
238
239 $sql_type - any DBI SQL_DATA_TYPE (e.g. SQL_INTEGER). Optional.
240 Default is undef.
241
242 Any other named parameters (%params) passed in will be saved into the
243 object as attributes.
244
245 sql_type objects are useful only in special cases where additional
246 information should be tagged onto the variable. For example, DBI
247 allows bind variables to be given an explicit type:
248
249 my ($sql, @bind) = sql_interp 'SELECT * FROM mytable WHERE',
250 'x=', \$x, 'AND y=', sql_type(\$y, SQL_VARCHAR), 'AND z IN',
251 sql_type([1, 2], SQL_INTEGER);
252 # RESULT: @bind =
253 # ([$x, sql_type(\$x)], [$y, sql_type(\$y, type => SQL_VARCHAR)],
254 # [1, sql_type([1, 2], type => SQL_INTEGER)],
255 # [2, sql_type([1, 2], type => SQL_INTEGER)]);
256
257 my $idx = 1;
258 for my $var (@bind) {
259 $sth->bind_param($idx++, $var->[0], $var->[1]->{type});
260 }
261 $sth->execute();
262 my $ret = $sth->selectall_arrayref();
263
264 If the interpolation list contains at least one sql_type object, then
265 all the variable references are transparently converted into sql_type
266 objects, and the elements of @bind take a special form: an arrayref
267 consisting of the bind value and the sql_type object that generated the
268 bind value. Note that a single sql_type holding an aggregate (arrayref
269 or hashref) may generate multiple bind values.
270
272 To have the generated SQL and bind variables sent to STDOUT, you can
273 set the environment variable "TRACE_SQL" to "1"
274
275 TRACE_SQL=1 perl my_script.pl
276
277 Here's some example output:
278
279 DEBUG:interp[sql=INSERT INTO mytable VALUES(?),bind=5]
280
282 The query language is SQL. There are other modules, such as
283 SQL::Abstract, that hide SQL behind method calls and/or Perl data
284 structures (hashes and arrays). The former may be undesirable in some
285 cases since it replaces one language with another and hides the full
286 capabilities and expressiveness of your database's native SQL language.
287 The latter may load too much meaning into the syntax of "{, "[" and "\"
288 thereby rendering the meaning less clear:
289
290 SQL::Abstract example:
291 %where = (lname => {like => '%son%'},
292 age => [-and => {'>=', 10}, {'<=', 20}])
293 Plain SQL:
294 "lname LIKE '%son' AND (age >= 10 AND age <= 20)"
295
296 In contrast, SQL::Interp does not abstract away your SQL but rather
297 makes it easier to interpolate Perl variables into your SQL. Now,
298 SQL::Interp does load some meaning into "{, "[" and "\", but we try to
299 limit its use to obvious cases. Since your raw SQL is exposed, you can
300 use your particular dialect of SQL.
301
303 Some types of interpolation are context-sensitive and involve
304 examination of your SQL fragments. The examination could fail on
305 obscure syntax, but it is generally robust. Look at the examples to
306 see the types of interpolation that are accepted, and if doubt, examine
307 the SQL output yourself with the TRACE_SQL environment variable set.
308 If needed, you can disable context sensitivity by inserting a null-
309 string before a variable.
310
311 "SET", "", \$x
312
313 A few things are just not possible with the ('WHERE', \%hashref)
314 syntax, so in such case, use a more direct syntax:
315
316 # ok--direct syntax
317 sql_interp '...WHERE', {x => $x, y => $y}, 'AND y = z';
318 # bad--trying to impose a hashref but keys must be scalars and be unique
319 sql_interp '...WHERE',
320 {sql_type(\$x) => sql('x'), y => $y, y => sql('z')};
321
322 In the cases where this module parses or generates SQL fragments, this
323 module should work for many databases, but its been tested mostly on
324 MySQL and PostgreSQL. Please inform the author of any
325 incompatibilities.
326
328 David Manura (<http://math2.org/david/contact>) (author). Mark
329 Stosberg (<http://mark.stosberg.com/>) created and maintains the
330 SQL::Interp fork. Also thanks to: Mark Tiefenbruck (syntax), Wojciech
331 Pietron (Oracle compat), Jim Chromie (DBIx::Interp idea), Juerd
332 Waalboer, Terrence Brannon (early feedback), and others.
333
334 If you like SQL::Interp, please consider supporting the project by
335 adding support for the 'quote_char' and 'name_sep' options.
336 SQL::Abstract has code that can be borrowed for this. See this bug
337 report for details: http://rt.cpan.org/Public/Bug/Display.html?id=31488
338
339 If you use SQL::Interp with PostgreSQL and are interested in a further
340 performance improvement, considering working on this optimization:
341 "RT#39778: wish: optimize IN() to be ANY() for compatible PostgreSQL
342 versions": https://rt.cpan.org/Ticket/Display.html?id=39778
343
344 SQL::Interp now has a code repository hosted on Github:
345
346 L<https://github.com/markstos/SQL-Interp>
347
349 Use rt.cpan.org for bug reports.
350
352 Copyright (c) 2003-2005, David Manura. This module is free software.
353 It may be used, redistributed and/or modified under the same terms as
354 Perl itself. See <http://www.perl.com/perl/misc/Artistic.html>.
355
357 Fork
358 This module was forked from SQL::Interpolate, around version 0.40. The
359 core functionality remains unchanged, but the following incompatible
360 changes have been made:
361
362 · The optional source filtering feature was removed.
363
364 · The optional "macro" feature was removed.
365
366 · A legacy, deprecated function "sql_literal" was removed.
367
368 · The docs were overhauled to be simpler and clearer.
369
370 So if you want those removed features, you should use SQL::Interpolate.
371 I used it for years without those optional features and never missed
372 them.
373
374 Other modules in this distribution
375 DBIx::Interp allows DBI methods to accept an "sql_interp()"-like
376 interpolation list rather than the traditional ($statement, \%attr,
377 @bind_values) parameter list. However, consider using DBIx::Simple
378 instead-- it even more user friendly.
379
380 Related modules
381 SQL::Abstract
382
383 SQL::Abstract shares with "SQL::Interp" the purpose of making SQL
384 generation easier. SQL::Abstract differs in that it expresses queries
385 in terms of OO method calls. It's syntax may impair readability because
386 it uses the subtle difference between a brace and bracket to denote the
387 difference between AND and OR in a query (the user can change whether a
388 bracket implies "AND" or "OR"). Some complex where clauses are
389 difficult or impossible with SQL::Abstract. SQL::Interp gives the
390 author more direct access to the underlying SQL. This permits using
391 the full expressivity of the database query language.
392
393 DBIx::Simple
394
395 DBIx::Simple strives to simplify SQL generation as well as the data
396 structures returned from DBI. "DBIx::Simple" currently can use
397 SQL::Interp to help generate SQL.
398
399 Class::DBI
400
401 Class::DBI is a popular "complete" solution for abstract database
402 access through an OO interface. It currently has a plugin called
403 Class::DBI::AbstractSearch that allows it to use "SQL::Abstract" to
404 generate SQL. It's possible that "SQL::Interp" could be integrated with
405 it as well.
406
407 SQL::Preproc
408
409 SQL::Preproc provides an "embedded SQL" approach where the Perl and SQL
410 languages are extended (via source filtering) to support interwoven
411 Perl and SQL. The syntax supports interpolating Perl variables into
412 SQL and passing query results back into Perl variables. In contrast,
413 SQL::Interp extends neither SQL nor Perl, and it deals only with
414 interpolating Perl variables into queries, whereas returning variables
415 from queries is the job of something like DBI, DBIx::Interp, or
416 DBIx::Simple.
417
418 SQL::String
419
420 SQL::String shares a number of similiarities to SQL::Interp but it is
421 more rudimentary. Both let you combine "chunks" of SQL that have their
422 parameters attached to them and then transform it into an SQL string
423 and list of bind parameters suitable for passing to DBI.
424
425 SQL::KeywordSearch
426
427 SQL::KeywordSearch helps generate SQL for a keyword search. It can
428 return the result in a format compatible with SQL::Interp to become
429 part of a larger query.
430
431
432
433perl v5.28.1 2019-02-02 SQL::Interp(3)