Line data Source code
1 : %top{
2 : /*-------------------------------------------------------------------------
3 : *
4 : * psqlscan.l
5 : * lexical scanner for SQL commands
6 : *
7 : * This lexer used to be part of psql, and that heritage is reflected in
8 : * the file name as well as function and typedef names, though it can now
9 : * be used by other frontend programs as well. It's also possible to extend
10 : * this lexer with a compatible add-on lexer to handle program-specific
11 : * backslash commands.
12 : *
13 : * This code is mainly concerned with determining where the end of a SQL
14 : * statement is: we are looking for semicolons that are not within quotes,
15 : * comments, or parentheses. The most reliable way to handle this is to
16 : * borrow the backend's flex lexer rules, lock, stock, and barrel. The rules
17 : * below are (except for a few) the same as the backend's, but their actions
18 : * are just ECHO whereas the backend's actions generally do other things.
19 : *
20 : * XXX The rules in this file must be kept in sync with the backend lexer!!!
21 : *
22 : * XXX Avoid creating backtracking cases --- see the backend lexer for info.
23 : *
24 : * See psqlscan_int.h for additional commentary.
25 : *
26 : *
27 : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
28 : * Portions Copyright (c) 1994, Regents of the University of California
29 : *
30 : * IDENTIFICATION
31 : * src/fe_utils/psqlscan.l
32 : *
33 : *-------------------------------------------------------------------------
34 : */
35 : #include "postgres_fe.h"
36 :
37 : #include "common/logging.h"
38 : #include "fe_utils/psqlscan.h"
39 :
40 : #include "libpq-fe.h"
41 : }
42 :
43 : %{
44 :
45 : /* LCOV_EXCL_START */
46 :
47 : #include "fe_utils/psqlscan_int.h"
48 :
49 : /*
50 : * We must have a typedef YYSTYPE for yylex's first argument, but this lexer
51 : * doesn't presently make use of that argument, so just declare it as int.
52 : */
53 : typedef int YYSTYPE;
54 :
55 :
56 : /* Return values from yylex() */
57 : #define LEXRES_EOL 0 /* end of input */
58 : #define LEXRES_SEMI 1 /* command-terminating semicolon found */
59 : #define LEXRES_BACKSLASH 2 /* backslash command start */
60 :
61 :
62 : #define ECHO psqlscan_emit(cur_state, yytext, yyleng)
63 :
64 : %}
65 :
66 : %option reentrant
67 : %option bison-bridge
68 : %option 8bit
69 : %option never-interactive
70 : %option nodefault
71 : %option noinput
72 : %option nounput
73 : %option noyywrap
74 : %option warn
75 : %option prefix="psql_yy"
76 :
77 : /*
78 : * Set the type of yyextra; we use it as a pointer back to the containing
79 : * PsqlScanState.
80 : */
81 : %option extra-type="PsqlScanState"
82 :
83 : /*
84 : * All of the following definitions and rules should exactly match
85 : * src/backend/parser/scan.l so far as the flex patterns are concerned.
86 : * The rule bodies are just ECHO as opposed to what the backend does,
87 : * however. (But be sure to duplicate code that affects the lexing process,
88 : * such as BEGIN() and yyless().) Also, psqlscan uses a single <<EOF>> rule
89 : * whereas scan.l has a separate one for each exclusive state.
90 : */
91 :
92 : /*
93 : * OK, here is a short description of lex/flex rules behavior.
94 : * The longest pattern which matches an input string is always chosen.
95 : * For equal-length patterns, the first occurring in the rules list is chosen.
96 : * INITIAL is the starting state, to which all non-conditional rules apply.
97 : * Exclusive states change parsing rules while the state is active. When in
98 : * an exclusive state, only those rules defined for that state apply.
99 : *
100 : * We use exclusive states for quoted strings, extended comments,
101 : * and to eliminate parsing troubles for numeric strings.
102 : * Exclusive states:
103 : * <xb> bit string literal
104 : * <xc> extended C-style comments
105 : * <xd> delimited identifiers (double-quoted identifiers)
106 : * <xh> hexadecimal byte string
107 : * <xq> standard quoted strings
108 : * <xqs> quote stop (detect continued strings)
109 : * <xe> extended quoted strings (support backslash escape sequences)
110 : * <xdolq> $foo$ quoted strings
111 : * <xui> quoted identifier with Unicode escapes
112 : * <xus> quoted string with Unicode escapes
113 : *
114 : * Note: we intentionally don't mimic the backend's <xeu> state; we have
115 : * no need to distinguish it from <xe> state, and no good way to get out
116 : * of it in error cases. The backend just throws yyerror() in those
117 : * cases, but that's not an option here.
118 : */
119 :
120 : %x xb
121 : %x xc
122 : %x xd
123 : %x xh
124 : %x xq
125 : %x xqs
126 : %x xe
127 : %x xdolq
128 : %x xui
129 : %x xus
130 :
131 : /*
132 : * In order to make the world safe for Windows and Mac clients as well as
133 : * Unix ones, we accept either \n or \r as a newline. A DOS-style \r\n
134 : * sequence will be seen as two successive newlines, but that doesn't cause
135 : * any problems. Comments that start with -- and extend to the next
136 : * newline are treated as equivalent to a single whitespace character.
137 : *
138 : * NOTE a fine point: if there is no newline following --, we will absorb
139 : * everything to the end of the input as a comment. This is correct. Older
140 : * versions of Postgres failed to recognize -- as a comment if the input
141 : * did not end with a newline.
142 : *
143 : * non_newline_space tracks all space characters except newlines.
144 : *
145 : * XXX if you change the set of whitespace characters, fix scanner_isspace()
146 : * to agree.
147 : */
148 :
149 : space [ \t\n\r\f\v]
150 : non_newline_space [ \t\f\v]
151 : newline [\n\r]
152 : non_newline [^\n\r]
153 :
154 : comment ("--"{non_newline}*)
155 :
156 : whitespace ({space}+|{comment})
157 :
158 : /*
159 : * SQL requires at least one newline in the whitespace separating
160 : * string literals that are to be concatenated. Silly, but who are we
161 : * to argue? Note that {whitespace_with_newline} should not have * after
162 : * it, whereas {whitespace} should generally have a * after it...
163 : */
164 :
165 : special_whitespace ({space}+|{comment}{newline})
166 : non_newline_whitespace ({non_newline_space}|{comment})
167 : whitespace_with_newline ({non_newline_whitespace}*{newline}{special_whitespace}*)
168 :
169 : quote '
170 : /* If we see {quote} then {quotecontinue}, the quoted string continues */
171 : quotecontinue {whitespace_with_newline}{quote}
172 :
173 : /*
174 : * {quotecontinuefail} is needed to avoid lexer backup when we fail to match
175 : * {quotecontinue}. It might seem that this could just be {whitespace}*,
176 : * but if there's a dash after {whitespace_with_newline}, it must be consumed
177 : * to see if there's another dash --- which would start a {comment} and thus
178 : * allow continuation of the {quotecontinue} token.
179 : */
180 : quotecontinuefail {whitespace}*"-"?
181 :
182 : /* Bit string
183 : * It is tempting to scan the string for only those characters
184 : * which are allowed. However, this leads to silently swallowed
185 : * characters if illegal characters are included in the string.
186 : * For example, if xbinside is [01] then B'ABCD' is interpreted
187 : * as a zero-length string, and the ABCD' is lost!
188 : * Better to pass the string forward and let the input routines
189 : * validate the contents.
190 : */
191 : xbstart [bB]{quote}
192 : xbinside [^']*
193 :
194 : /* Hexadecimal byte string */
195 : xhstart [xX]{quote}
196 : xhinside [^']*
197 :
198 : /* National character */
199 : xnstart [nN]{quote}
200 :
201 : /* Quoted string that allows backslash escapes */
202 : xestart [eE]{quote}
203 : xeinside [^\\']+
204 : xeescape [\\][^0-7]
205 : xeoctesc [\\][0-7]{1,3}
206 : xehexesc [\\]x[0-9A-Fa-f]{1,2}
207 : xeunicode [\\](u[0-9A-Fa-f]{4}|U[0-9A-Fa-f]{8})
208 : xeunicodefail [\\](u[0-9A-Fa-f]{0,3}|U[0-9A-Fa-f]{0,7})
209 :
210 : /* Extended quote
211 : * xqdouble implements embedded quote, ''''
212 : */
213 : xqstart {quote}
214 : xqdouble {quote}{quote}
215 : xqinside [^']+
216 :
217 : /* $foo$ style quotes ("dollar quoting")
218 : * The quoted string starts with $foo$ where "foo" is an optional string
219 : * in the form of an identifier, except that it may not contain "$",
220 : * and extends to the first occurrence of an identical string.
221 : * There is *no* processing of the quoted text.
222 : *
223 : * {dolqfailed} is an error rule to avoid scanner backup when {dolqdelim}
224 : * fails to match its trailing "$".
225 : */
226 : dolq_start [A-Za-z\200-\377_]
227 : dolq_cont [A-Za-z\200-\377_0-9]
228 : dolqdelim \$({dolq_start}{dolq_cont}*)?\$
229 : dolqfailed \${dolq_start}{dolq_cont}*
230 : dolqinside [^$]+
231 :
232 : /* Double quote
233 : * Allows embedded spaces and other special characters into identifiers.
234 : */
235 : dquote \"
236 : xdstart {dquote}
237 : xdstop {dquote}
238 : xddouble {dquote}{dquote}
239 : xdinside [^"]+
240 :
241 : /* Quoted identifier with Unicode escapes */
242 : xuistart [uU]&{dquote}
243 :
244 : /* Quoted string with Unicode escapes */
245 : xusstart [uU]&{quote}
246 :
247 : /* error rule to avoid backup */
248 : xufailed [uU]&
249 :
250 :
251 : /* C-style comments
252 : *
253 : * The "extended comment" syntax closely resembles allowable operator syntax.
254 : * The tricky part here is to get lex to recognize a string starting with
255 : * slash-star as a comment, when interpreting it as an operator would produce
256 : * a longer match --- remember lex will prefer a longer match! Also, if we
257 : * have something like plus-slash-star, lex will think this is a 3-character
258 : * operator whereas we want to see it as a + operator and a comment start.
259 : * The solution is two-fold:
260 : * 1. append {op_chars}* to xcstart so that it matches as much text as
261 : * {operator} would. Then the tie-breaker (first matching rule of same
262 : * length) ensures xcstart wins. We put back the extra stuff with yyless()
263 : * in case it contains a star-slash that should terminate the comment.
264 : * 2. In the operator rule, check for slash-star within the operator, and
265 : * if found throw it back with yyless(). This handles the plus-slash-star
266 : * problem.
267 : * Dash-dash comments have similar interactions with the operator rule.
268 : */
269 : xcstart \/\*{op_chars}*
270 : xcstop \*+\/
271 : xcinside [^*/]+
272 :
273 : ident_start [A-Za-z\200-\377_]
274 : ident_cont [A-Za-z\200-\377_0-9\$]
275 :
276 : identifier {ident_start}{ident_cont}*
277 :
278 : /* Assorted special-case operators and operator-like tokens */
279 : typecast "::"
280 : dot_dot \.\.
281 : colon_equals ":="
282 :
283 : /*
284 : * These operator-like tokens (unlike the above ones) also match the {operator}
285 : * rule, which means that they might be overridden by a longer match if they
286 : * are followed by a comment start or a + or - character. Accordingly, if you
287 : * add to this list, you must also add corresponding code to the {operator}
288 : * block to return the correct token in such cases. (This is not needed in
289 : * psqlscan.l since the token value is ignored there.)
290 : */
291 : equals_greater "=>"
292 : less_equals "<="
293 : greater_equals ">="
294 : less_greater "<>"
295 : not_equals "!="
296 : /* Note there is no need for left_arrow, since "<-" is not a single operator. */
297 : right_arrow "->"
298 :
299 : /*
300 : * "self" is the set of chars that should be returned as single-character
301 : * tokens. "op_chars" is the set of chars that can make up "Op" tokens,
302 : * which can be one or more characters long (but if a single-char token
303 : * appears in the "self" set, it is not to be returned as an Op). Note
304 : * that the sets overlap, but each has some chars that are not in the other.
305 : *
306 : * If you change either set, adjust the character lists appearing in the
307 : * rule for "operator"!
308 : */
309 : self [,()\[\].;\:\|\+\-\*\/\%\^\<\>\=]
310 : op_chars [\~\!\@\#\^\&\|\`\?\+\-\*\/\%\<\>\=]
311 : operator {op_chars}+
312 :
313 : /*
314 : * Numbers
315 : *
316 : * Unary minus is not part of a number here. Instead we pass it separately to
317 : * the parser, and there it gets coerced via doNegate().
318 : *
319 : * {numericfail} is used because we would like "1..10" to lex as 1, dot_dot, 10.
320 : *
321 : * {realfail} is added to prevent the need for scanner
322 : * backup when the {real} rule fails to match completely.
323 : */
324 : decdigit [0-9]
325 : hexdigit [0-9A-Fa-f]
326 : octdigit [0-7]
327 : bindigit [0-1]
328 :
329 : decinteger {decdigit}(_?{decdigit})*
330 : hexinteger 0[xX](_?{hexdigit})+
331 : octinteger 0[oO](_?{octdigit})+
332 : bininteger 0[bB](_?{bindigit})+
333 :
334 : hexfail 0[xX]_?
335 : octfail 0[oO]_?
336 : binfail 0[bB]_?
337 :
338 : numeric (({decinteger}\.{decinteger}?)|(\.{decinteger}))
339 : numericfail {decinteger}\.\.
340 :
341 : real ({decinteger}|{numeric})[Ee][-+]?{decinteger}
342 : realfail ({decinteger}|{numeric})[Ee][-+]
343 :
344 : /* Positional parameters don't accept underscores. */
345 : param \${decdigit}+
346 :
347 : /*
348 : * An identifier immediately following an integer literal is disallowed because
349 : * in some cases it's ambiguous what is meant: for example, 0x1234 could be
350 : * either a hexinteger or a decinteger "0" and an identifier "x1234". We can
351 : * detect such problems by seeing if integer_junk matches a longer substring
352 : * than any of the XXXinteger patterns (decinteger, hexinteger, octinteger,
353 : * bininteger). One "junk" pattern is sufficient because
354 : * {decinteger}{identifier} will match all the same strings we'd match with
355 : * {hexinteger}{identifier} etc.
356 : *
357 : * Note that the rule for integer_junk must appear after the ones for
358 : * XXXinteger to make this work correctly: 0x1234 will match both hexinteger
359 : * and integer_junk, and we need hexinteger to be chosen in that case.
360 : *
361 : * Also disallow strings matched by numeric_junk, real_junk and param_junk
362 : * for consistency.
363 : */
364 : integer_junk {decinteger}{identifier}
365 : numeric_junk {numeric}{identifier}
366 : real_junk {real}{identifier}
367 : param_junk \${decdigit}+{identifier}
368 :
369 : /* psql-specific: characters allowed in variable names */
370 : variable_char [A-Za-z\200-\377_0-9]
371 :
372 : other .
373 :
374 : /*
375 : * Dollar quoted strings are totally opaque, and no escaping is done on them.
376 : * Other quoted strings must allow some special characters such as single-quote
377 : * and newline.
378 : * Embedded single-quotes are implemented both in the SQL standard
379 : * style of two adjacent single quotes "''" and in the Postgres/Java style
380 : * of escaped-quote "\'".
381 : * Other embedded escaped characters are matched explicitly and the leading
382 : * backslash is dropped from the string.
383 : * Note that xcstart must appear before operator, as explained above!
384 : * Also whitespace (comment) must appear before operator.
385 : */
386 :
387 : %%
388 :
389 : %{
390 : /* Declare some local variables inside yylex(), for convenience */
391 : PsqlScanState cur_state = yyextra;
392 790313 : PQExpBuffer output_buf = cur_state->output_buf;
393 790313 :
394 : /*
395 : * Force flex into the state indicated by start_state. This has a
396 : * couple of purposes: it lets some of the functions below set a new
397 : * starting state without ugly direct access to flex variables, and it
398 : * allows us to transition from one flex lexer to another so that we
399 : * can lex different parts of the source string using separate lexers.
400 : */
401 : BEGIN(cur_state->start_state);
402 790313 : %}
403 :
404 : {whitespace} {
405 : /*
406 : * Note that the whitespace rule includes both true
407 : * whitespace and single-line ("--" style) comments.
408 : * We suppress whitespace until we have collected some
409 : * non-whitespace data. (This interacts with some
410 : * decisions in MainLoop(); see there for details.)
411 : */
412 : if (output_buf->len > 0)
413 1874509 : ECHO;
414 1762957 : }
415 :
416 1874509 : {xcstart} {
417 448 : cur_state->xcdepth = 0;
418 448 : BEGIN(xc);
419 448 : /* Put back any characters past slash-star; see above */
420 : yyless(2);
421 448 : ECHO;
422 448 : }
423 :
424 448 : <xc>{
425 : {xcstart} {
426 12 : cur_state->xcdepth++;
427 12 : /* Put back any characters past slash-star; see above */
428 : yyless(2);
429 12 : ECHO;
430 12 : }
431 :
432 12 : {xcstop} {
433 460 : if (cur_state->xcdepth <= 0)
434 460 : BEGIN(INITIAL);
435 448 : else
436 : cur_state->xcdepth--;
437 12 : ECHO;
438 460 : }
439 :
440 460 : {xcinside} {
441 1029 : ECHO;
442 1029 : }
443 :
444 1029 : {op_chars} {
445 277 : ECHO;
446 277 : }
447 :
448 277 : \*+ {
449 0 : ECHO;
450 0 : }
451 : } /* <xc> */
452 0 :
453 : {xbstart} {
454 500 : BEGIN(xb);
455 500 : ECHO;
456 500 : }
457 : <xh>{xhinside} |
458 500 : <xb>{xbinside} {
459 2695 : ECHO;
460 2695 : }
461 :
462 2695 : {xhstart} {
463 2215 : /* Hexadecimal bit type.
464 : * At some point we should simply pass the string
465 : * forward to the parser and label it there.
466 : * In the meantime, place a leading "x" on the string
467 : * to mark it for the input routine as a hex string.
468 : */
469 : BEGIN(xh);
470 2215 : ECHO;
471 2215 : }
472 :
473 2215 : {xnstart} {
474 0 : yyless(1); /* eat only 'n' this time */
475 0 : ECHO;
476 0 : }
477 :
478 0 : {xqstart} {
479 157645 : if (cur_state->std_strings)
480 157645 : BEGIN(xq);
481 157645 : else
482 : BEGIN(xe);
483 0 : ECHO;
484 157645 : }
485 : {xestart} {
486 157645 : BEGIN(xe);
487 868 : ECHO;
488 868 : }
489 : {xusstart} {
490 868 : BEGIN(xus);
491 368 : ECHO;
492 368 : }
493 :
494 368 : <xb,xh,xq,xe,xus>{quote} {
495 161596 : /*
496 : * When we are scanning a quoted string and see an end
497 : * quote, we must look ahead for a possible continuation.
498 : * If we don't see one, we know the end quote was in fact
499 : * the end of the string. To reduce the lexer table size,
500 : * we use a single "xqs" state to do the lookahead for all
501 : * types of strings.
502 : */
503 : cur_state->state_before_str_stop = YYSTATE;
504 161596 : BEGIN(xqs);
505 161596 : ECHO;
506 161596 : }
507 : <xqs>{quotecontinue} {
508 161596 : /*
509 0 : * Found a quote continuation, so return to the in-quote
510 : * state and continue scanning the literal. Nothing is
511 : * added to the literal's contents.
512 : */
513 : BEGIN(cur_state->state_before_str_stop);
514 0 : ECHO;
515 0 : }
516 : <xqs>{quotecontinuefail} |
517 0 : <xqs>{other} {
518 160781 : /*
519 : * Failed to see a quote continuation. Throw back
520 : * everything after the end quote, and handle the string
521 : * according to the state we were in previously.
522 : */
523 : yyless(0);
524 160781 : BEGIN(INITIAL);
525 160781 : /* There's nothing to echo ... */
526 : }
527 :
528 160781 : <xq,xe,xus>{xqdouble} {
529 4130 : ECHO;
530 4130 : }
531 : <xq,xus>{xqinside} {
532 4130 : ECHO;
533 165297 : }
534 : <xe>{xeinside} {
535 165297 : ECHO;
536 1659 : }
537 : <xe>{xeunicode} {
538 1659 : ECHO;
539 124 : }
540 : <xe>{xeunicodefail} {
541 124 : ECHO;
542 8 : }
543 : <xe>{xeescape} {
544 8 : ECHO;
545 959 : }
546 : <xe>{xeoctesc} {
547 959 : ECHO;
548 14 : }
549 : <xe>{xehexesc} {
550 14 : ECHO;
551 6 : }
552 : <xe>. {
553 6 : /* This is only needed for \ just before EOF */
554 0 : ECHO;
555 0 : }
556 :
557 0 : {dolqdelim} {
558 4547 : cur_state->dolqstart = pg_strdup(yytext);
559 4547 : BEGIN(xdolq);
560 4547 : ECHO;
561 4547 : }
562 : {dolqfailed} {
563 4547 : /* throw back all but the initial "$" */
564 0 : yyless(1);
565 0 : ECHO;
566 0 : }
567 : <xdolq>{dolqdelim} {
568 0 : if (strcmp(yytext, cur_state->dolqstart) == 0)
569 4763 : {
570 : free(cur_state->dolqstart);
571 4547 : cur_state->dolqstart = NULL;
572 4547 : BEGIN(INITIAL);
573 4547 : }
574 : else
575 : {
576 : /*
577 : * When we fail to match $...$ to dolqstart, transfer
578 : * the $... part to the output, but put back the final
579 : * $ for rescanning. Consider $delim$...$junk$delim$
580 : */
581 : yyless(yyleng - 1);
582 216 : }
583 : ECHO;
584 4763 : }
585 : <xdolq>{dolqinside} {
586 4763 : ECHO;
587 24031 : }
588 : <xdolq>{dolqfailed} {
589 24031 : ECHO;
590 574 : }
591 : <xdolq>. {
592 574 : /* This is only needed for $ inside the quoted text */
593 1588 : ECHO;
594 1588 : }
595 :
596 1588 : {xdstart} {
597 6573 : BEGIN(xd);
598 6573 : ECHO;
599 6573 : }
600 : {xuistart} {
601 6573 : BEGIN(xui);
602 16 : ECHO;
603 16 : }
604 : <xd>{xdstop} {
605 16 : BEGIN(INITIAL);
606 6573 : ECHO;
607 6573 : }
608 : <xui>{dquote} {
609 6573 : BEGIN(INITIAL);
610 16 : ECHO;
611 16 : }
612 : <xd,xui>{xddouble} {
613 16 : ECHO;
614 67 : }
615 : <xd,xui>{xdinside} {
616 67 : ECHO;
617 6650 : }
618 :
619 6650 : {xufailed} {
620 0 : /* throw back all but the initial u/U */
621 : yyless(1);
622 0 : ECHO;
623 0 : }
624 :
625 0 : {typecast} {
626 36055 : ECHO;
627 36055 : }
628 :
629 36055 : {dot_dot} {
630 0 : ECHO;
631 0 : }
632 :
633 0 : {colon_equals} {
634 1673 : ECHO;
635 1673 : }
636 :
637 1673 : {equals_greater} {
638 1361 : ECHO;
639 1361 : }
640 :
641 1361 : {less_equals} {
642 1412 : ECHO;
643 1412 : }
644 :
645 1412 : {greater_equals} {
646 4190 : ECHO;
647 4190 : }
648 :
649 4190 : {less_greater} {
650 895 : ECHO;
651 895 : }
652 :
653 895 : {not_equals} {
654 1524 : ECHO;
655 1524 : }
656 :
657 1524 : {right_arrow} {
658 773 : ECHO;
659 773 : }
660 :
661 773 : /*
662 : * These rules are specific to psql --- they implement parenthesis
663 : * counting and detection of command-ending semicolon. These must
664 : * appear before the {self} rule so that they take precedence over it.
665 : */
666 :
667 253931 : "(" {
668 : cur_state->paren_depth++;
669 253931 : ECHO;
670 253931 : }
671 :
672 253931 : ")" {
673 253922 : if (cur_state->paren_depth > 0)
674 253922 : cur_state->paren_depth--;
675 253922 : ECHO;
676 253922 : }
677 :
678 253922 : ";" {
679 242975 : ECHO;
680 242975 : if (cur_state->paren_depth == 0 && cur_state->begin_depth == 0)
681 242975 : {
682 : /* Terminate lexing temporarily */
683 : cur_state->start_state = YY_START;
684 242812 : cur_state->identifier_count = 0;
685 242812 : return LEXRES_SEMI;
686 242812 : }
687 : }
688 :
689 163 : /*
690 : * psql-specific rules to handle backslash commands and variable
691 : * substitution. We want these before {self}, also.
692 : */
693 :
694 512 : "\\"[;:] {
695 : /* Force a semi-colon or colon into the query buffer */
696 : psqlscan_emit(cur_state, yytext + 1, 1);
697 512 : if (yytext[1] == ';')
698 512 : cur_state->identifier_count = 0;
699 512 : }
700 :
701 512 : "\\" {
702 32404 : /* Terminate lexing temporarily */
703 : cur_state->start_state = YY_START;
704 32404 : return LEXRES_BACKSLASH;
705 32404 : }
706 :
707 : :{variable_char}+ {
708 1782 : /* Possible psql variable substitution */
709 : char *varname;
710 : char *value;
711 :
712 : varname = psqlscan_extract_substring(cur_state,
713 1782 : yytext + 1,
714 1782 : yyleng - 1);
715 1782 : if (cur_state->callbacks->get_variable)
716 1782 : value = cur_state->callbacks->get_variable(varname,
717 1186 : PQUOTE_PLAIN,
718 : cur_state->cb_passthrough);
719 : else
720 : value = NULL;
721 596 :
722 : if (value)
723 1782 : {
724 : /* It is a variable, check for recursion */
725 : if (psqlscan_var_is_current_source(cur_state, varname))
726 878 : {
727 : /* Recursive expansion --- don't go there */
728 : pg_log_warning("skipping recursive expansion of variable \"%s\"",
729 0 : varname);
730 : /* Instead copy the string as is */
731 : ECHO;
732 0 : }
733 : else
734 : {
735 : /* OK, perform substitution */
736 : psqlscan_push_new_buffer(cur_state, value, varname);
737 878 : /* yy_scan_string already made buffer active */
738 : }
739 : free(value);
740 878 : }
741 : else
742 : {
743 : /*
744 : * if the variable doesn't exist we'll copy the string
745 : * as is
746 : */
747 : ECHO;
748 904 : }
749 :
750 : free(varname);
751 1782 : }
752 :
753 1782 : :'{variable_char}+' {
754 616 : psqlscan_escape_variable(cur_state, yytext, yyleng,
755 616 : PQUOTE_SQL_LITERAL);
756 : }
757 :
758 616 : :\"{variable_char}+\" {
759 21 : psqlscan_escape_variable(cur_state, yytext, yyleng,
760 21 : PQUOTE_SQL_IDENT);
761 : }
762 :
763 21 : :\{\?{variable_char}+\} {
764 8 : psqlscan_test_variable(cur_state, yytext, yyleng);
765 8 : }
766 :
767 8 : /*
768 : * These rules just avoid the need for scanner backup if one of the
769 : * three rules above fails to match completely.
770 : */
771 :
772 0 : :'{variable_char}* {
773 : /* Throw back everything but the colon */
774 : yyless(1);
775 0 : ECHO;
776 0 : }
777 :
778 0 : :\"{variable_char}* {
779 0 : /* Throw back everything but the colon */
780 : yyless(1);
781 0 : ECHO;
782 0 : }
783 :
784 0 : :\{\?{variable_char}* {
785 0 : /* Throw back everything but the colon */
786 : yyless(1);
787 0 : ECHO;
788 0 : }
789 : :\{ {
790 0 : /* Throw back everything but the colon */
791 0 : yyless(1);
792 0 : ECHO;
793 0 : }
794 :
795 0 : /*
796 : * Back to backend-compatible rules.
797 : */
798 :
799 442067 : {self} {
800 : ECHO;
801 442067 : }
802 :
803 442067 : {operator} {
804 12738 : /*
805 : * Check for embedded slash-star or dash-dash; those
806 : * are comment starts, so operator must stop there.
807 : * Note that slash-star or dash-dash at the first
808 : * character will match a prior rule, not this one.
809 : */
810 : int nchars = yyleng;
811 12738 : char *slashstar = strstr(yytext, "/*");
812 12738 : char *dashdash = strstr(yytext, "--");
813 12738 :
814 : if (slashstar && dashdash)
815 12738 : {
816 : /* if both appear, take the first one */
817 : if (slashstar > dashdash)
818 0 : slashstar = dashdash;
819 0 : }
820 : else if (!slashstar)
821 12738 : slashstar = dashdash;
822 12698 : if (slashstar)
823 12738 : nchars = slashstar - yytext;
824 48 :
825 : /*
826 : * For SQL compatibility, '+' and '-' cannot be the
827 : * last char of a multi-char operator unless the operator
828 : * contains chars that are not in SQL operators.
829 : * The idea is to lex '=-' as two operators, but not
830 : * to forbid operator names like '?-' that could not be
831 : * sequences of SQL operators.
832 : */
833 : if (nchars > 1 &&
834 12738 : (yytext[nchars - 1] == '+' ||
835 11708 : yytext[nchars - 1] == '-'))
836 11704 : {
837 : int ic;
838 :
839 : for (ic = nchars - 2; ic >= 0; ic--)
840 385 : {
841 : char c = yytext[ic];
842 326 : if (c == '~' || c == '!' || c == '@' ||
843 326 : c == '#' || c == '^' || c == '&' ||
844 270 : c == '|' || c == '`' || c == '?' ||
845 106 : c == '%')
846 : break;
847 : }
848 : if (ic < 0)
849 291 : {
850 : /*
851 : * didn't find a qualifying character, so remove
852 : * all trailing [+-]
853 : */
854 : do {
855 : nchars--;
856 59 : } while (nchars > 1 &&
857 59 : (yytext[nchars - 1] == '+' ||
858 23 : yytext[nchars - 1] == '-'));
859 23 : }
860 : }
861 :
862 : if (nchars < yyleng)
863 12738 : {
864 : /* Strip the unwanted chars from the token */
865 : yyless(nchars);
866 107 : }
867 : ECHO;
868 12738 : }
869 :
870 12738 : {param} {
871 789 : ECHO;
872 789 : }
873 : {param_junk} {
874 789 : ECHO;
875 8 : }
876 :
877 8 : {decinteger} {
878 139826 : ECHO;
879 139826 : }
880 : {hexinteger} {
881 139826 : ECHO;
882 83 : }
883 : {octinteger} {
884 83 : ECHO;
885 40 : }
886 : {bininteger} {
887 40 : ECHO;
888 40 : }
889 : {hexfail} {
890 40 : ECHO;
891 4 : }
892 : {octfail} {
893 4 : ECHO;
894 4 : }
895 : {binfail} {
896 4 : ECHO;
897 4 : }
898 : {numeric} {
899 4 : ECHO;
900 5286 : }
901 : {numericfail} {
902 5286 : /* throw back the .., and treat as integer */
903 0 : yyless(yyleng - 2);
904 0 : ECHO;
905 0 : }
906 : {real} {
907 0 : ECHO;
908 418 : }
909 : {realfail} {
910 418 : ECHO;
911 4 : }
912 : {integer_junk} {
913 4 : ECHO;
914 44 : }
915 : {numeric_junk} {
916 44 : ECHO;
917 32 : }
918 : {real_junk} {
919 32 : ECHO;
920 0 : }
921 :
922 0 :
923 1822555 : {identifier} {
924 : /*
925 : * We need to track if we are inside a BEGIN .. END block
926 : * in a function definition, so that semicolons contained
927 : * therein don't terminate the whole statement. Short of
928 : * writing a full parser here, the following heuristic
929 : * should work. First, we track whether the beginning of
930 : * the statement matches CREATE [OR REPLACE]
931 : * {FUNCTION|PROCEDURE|SCHEMA}. (Allowing this in
932 : * CREATE SCHEMA, without tracking whether we're within a
933 : * CREATE FUNCTION/PROCEDURE subcommand, is a bit shaky
934 : * but should be okay with the present set of valid
935 : * subcommands.)
936 : */
937 :
938 : if (cur_state->identifier_count == 0)
939 1822555 : memset(cur_state->identifiers, 0, sizeof(cur_state->identifiers));
940 249526 :
941 : if (cur_state->identifier_count < sizeof(cur_state->identifiers))
942 1822555 : {
943 : if (pg_strcasecmp(yytext, "create") == 0 ||
944 1666983 : pg_strcasecmp(yytext, "function") == 0 ||
945 1617915 : pg_strcasecmp(yytext, "procedure") == 0 ||
946 1610902 : pg_strcasecmp(yytext, "or") == 0 ||
947 1609034 : pg_strcasecmp(yytext, "replace") == 0 ||
948 1606133 : pg_strcasecmp(yytext, "schema") == 0)
949 802335 : cur_state->identifiers[cur_state->identifier_count] = pg_tolower((unsigned char) yytext[0]);
950 53734 : }
951 :
952 : cur_state->identifier_count++;
953 1822555 :
954 : if (cur_state->identifiers[0] == 'c' &&
955 1822555 : (cur_state->identifiers[1] == 'f' || cur_state->identifiers[1] == 'p' ||
956 436694 : (cur_state->identifiers[1] == 'o' && cur_state->identifiers[2] == 'r' &&
957 398694 : (cur_state->identifiers[3] == 'f' || cur_state->identifiers[3] == 'p')) ||
958 14937 : cur_state->identifiers[1] == 's') &&
959 389347 : cur_state->paren_depth == 0)
960 53932 : {
961 : if (pg_strcasecmp(yytext, "begin") == 0)
962 42491 : cur_state->begin_depth++;
963 115 : else if (pg_strcasecmp(yytext, "case") == 0)
964 42376 : {
965 : /*
966 : * CASE also ends with END. We only need to track
967 : * this if we are already inside a BEGIN.
968 : */
969 : if (cur_state->begin_depth >= 1)
970 4 : cur_state->begin_depth++;
971 4 : }
972 : else if (pg_strcasecmp(yytext, "end") == 0)
973 42372 : {
974 : if (cur_state->begin_depth > 0)
975 119 : cur_state->begin_depth--;
976 119 : }
977 : }
978 :
979 : ECHO;
980 1822555 : }
981 :
982 1822555 : {other} {
983 8 : ECHO;
984 8 : }
985 :
986 8 : <<EOF>> {
987 515975 : if (cur_state->buffer_stack == NULL)
988 515975 : {
989 : cur_state->start_state = YY_START;
990 515097 : return LEXRES_EOL; /* end of input reached */
991 515097 : }
992 :
993 : /*
994 : * We were expanding a variable, so pop the inclusion
995 : * stack and keep lexing
996 : */
997 : psqlscan_pop_buffer_stack(cur_state);
998 878 : psqlscan_select_top_buffer(cur_state);
999 878 : }
1000 :
1001 878 : %%
1002 0 :
1003 : /* LCOV_EXCL_STOP */
1004 :
1005 : /*
1006 : * Create a lexer working state struct.
1007 : *
1008 : * callbacks is a struct of function pointers that encapsulate some
1009 : * behavior we need from the surrounding program. This struct must
1010 : * remain valid for the lifespan of the PsqlScanState.
1011 : */
1012 : PsqlScanState
1013 : psql_scan_create(const PsqlScanCallbacks *callbacks)
1014 10431 : {
1015 : PsqlScanState state;
1016 :
1017 : state = pg_malloc0_object(PsqlScanStateData);
1018 10431 :
1019 : state->callbacks = callbacks;
1020 10431 :
1021 : yylex_init(&state->scanner);
1022 10431 :
1023 : yyset_extra(state, state->scanner);
1024 10431 :
1025 : psql_scan_reset(state);
1026 10431 :
1027 : return state;
1028 10431 : }
1029 :
1030 : /*
1031 : * Destroy a lexer working state struct, releasing all resources.
1032 : */
1033 : void
1034 : psql_scan_destroy(PsqlScanState state)
1035 10375 : {
1036 : psql_scan_finish(state);
1037 10375 :
1038 : psql_scan_reset(state);
1039 10375 :
1040 : yylex_destroy(state->scanner);
1041 10375 :
1042 : free(state);
1043 10375 : }
1044 10375 :
1045 : /*
1046 : * Set the callback passthrough pointer for the lexer.
1047 : *
1048 : * This could have been integrated into psql_scan_create, but keeping it
1049 : * separate allows the application to change the pointer later, which might
1050 : * be useful.
1051 : */
1052 : void
1053 : psql_scan_set_passthrough(PsqlScanState state, void *passthrough)
1054 9828 : {
1055 : state->cb_passthrough = passthrough;
1056 9828 : }
1057 9828 :
1058 : /*
1059 : * Set up to perform lexing of the given input line.
1060 : *
1061 : * The text at *line, extending for line_len bytes, will be scanned by
1062 : * subsequent calls to the psql_scan routines. psql_scan_finish should
1063 : * be called when scanning is complete. Note that the lexer retains
1064 : * a pointer to the storage at *line --- this string must not be altered
1065 : * or freed until after psql_scan_finish is called.
1066 : *
1067 : * encoding is the libpq identifier for the character encoding in use,
1068 : * and std_strings says whether standard_conforming_strings is on.
1069 : */
1070 : void
1071 : psql_scan_setup(PsqlScanState state,
1072 515422 : const char *line, int line_len,
1073 : int encoding, bool std_strings)
1074 : {
1075 : /* Mustn't be scanning already */
1076 : Assert(state->scanbufhandle == NULL);
1077 : Assert(state->buffer_stack == NULL);
1078 :
1079 : /* Do we need to hack the character set encoding? */
1080 : state->encoding = encoding;
1081 515422 : state->safe_encoding = pg_valid_server_encoding_id(encoding);
1082 515422 :
1083 : /* Save standard-strings flag as well */
1084 : state->std_strings = std_strings;
1085 515422 :
1086 : /* Set up flex input buffer with appropriate translation and padding */
1087 : state->scanbufhandle = psqlscan_prepare_buffer(state, line, line_len,
1088 515422 : &state->scanbuf);
1089 : state->scanline = line;
1090 515422 :
1091 : /* Set lookaside data in case we have to map unsafe encoding */
1092 : state->curline = state->scanbuf;
1093 515422 : state->refline = state->scanline;
1094 515422 :
1095 : /* Initialize state for psql_scan_get_location() */
1096 : state->cur_line_no = 0; /* yylex not called yet */
1097 515422 : state->cur_line_ptr = state->scanbuf;
1098 515422 : }
1099 515422 :
1100 : /*
1101 : * Do lexical analysis of SQL command text.
1102 : *
1103 : * The text previously passed to psql_scan_setup is scanned, and appended
1104 : * (possibly with transformation) to query_buf.
1105 : *
1106 : * The return value indicates the condition that stopped scanning:
1107 : *
1108 : * PSCAN_SEMICOLON: found a command-ending semicolon. (The semicolon is
1109 : * transferred to query_buf.) The command accumulated in query_buf should
1110 : * be executed, then clear query_buf and call again to scan the remainder
1111 : * of the line.
1112 : *
1113 : * PSCAN_BACKSLASH: found a backslash that starts a special command.
1114 : * Any previous data on the line has been transferred to query_buf.
1115 : * The caller will typically next apply a separate flex lexer to scan
1116 : * the special command.
1117 : *
1118 : * PSCAN_INCOMPLETE: the end of the line was reached, but we have an
1119 : * incomplete SQL command. *prompt is set to the appropriate prompt type.
1120 : *
1121 : * PSCAN_EOL: the end of the line was reached, and there is no lexical
1122 : * reason to consider the command incomplete. The caller may or may not
1123 : * choose to send it. *prompt is set to the appropriate prompt type if
1124 : * the caller chooses to collect more input.
1125 : *
1126 : * In the PSCAN_INCOMPLETE and PSCAN_EOL cases, psql_scan_finish() should
1127 : * be called next, then the cycle may be repeated with a fresh input line.
1128 : *
1129 : * In all cases, *prompt is set to an appropriate prompt type code for the
1130 : * next line-input operation.
1131 : */
1132 : PsqlScanResult
1133 : psql_scan(PsqlScanState state,
1134 790313 : PQExpBuffer query_buf,
1135 : promptStatus_t *prompt)
1136 : {
1137 : PsqlScanResult result;
1138 : int lexresult;
1139 :
1140 : /* Must be scanning already */
1141 : Assert(state->scanbufhandle != NULL);
1142 :
1143 : /* Set current output target */
1144 : state->output_buf = query_buf;
1145 790313 :
1146 : /* Set input source */
1147 : if (state->buffer_stack != NULL)
1148 790313 : yy_switch_to_buffer(state->buffer_stack->buf, state->scanner);
1149 60 : else
1150 : yy_switch_to_buffer(state->scanbufhandle, state->scanner);
1151 790253 :
1152 : /* And lex. */
1153 : lexresult = yylex(NULL, state->scanner);
1154 790313 :
1155 : /* Notify psql_scan_get_location() that a yylex call has been made. */
1156 : if (state->cur_line_no == 0)
1157 790313 : state->cur_line_no = 1;
1158 515420 :
1159 : /*
1160 : * Check termination state and return appropriate result info.
1161 : */
1162 : switch (lexresult)
1163 790313 : {
1164 : case LEXRES_EOL: /* end of input */
1165 515097 : switch (state->start_state)
1166 515097 : {
1167 : case INITIAL:
1168 483588 : case xqs: /* we treat this like INITIAL */
1169 : if (state->paren_depth > 0)
1170 483588 : {
1171 : result = PSCAN_INCOMPLETE;
1172 42388 : *prompt = PROMPT_PAREN;
1173 42388 : }
1174 : else if (state->begin_depth > 0)
1175 441200 : {
1176 : result = PSCAN_INCOMPLETE;
1177 665 : *prompt = PROMPT_CONTINUE;
1178 665 : }
1179 : else if (query_buf->len > 0)
1180 440535 : {
1181 : result = PSCAN_EOL;
1182 93066 : *prompt = PROMPT_CONTINUE;
1183 93066 : }
1184 : else
1185 : {
1186 : /* never bother to send an empty buffer */
1187 : result = PSCAN_INCOMPLETE;
1188 347469 : *prompt = PROMPT_READY;
1189 347469 : }
1190 : break;
1191 483588 : case xb:
1192 0 : result = PSCAN_INCOMPLETE;
1193 0 : *prompt = PROMPT_SINGLEQUOTE;
1194 0 : break;
1195 0 : case xc:
1196 513 : result = PSCAN_INCOMPLETE;
1197 513 : *prompt = PROMPT_COMMENT;
1198 513 : break;
1199 513 : case xd:
1200 23 : result = PSCAN_INCOMPLETE;
1201 23 : *prompt = PROMPT_DOUBLEQUOTE;
1202 23 : break;
1203 23 : case xh:
1204 0 : result = PSCAN_INCOMPLETE;
1205 0 : *prompt = PROMPT_SINGLEQUOTE;
1206 0 : break;
1207 0 : case xe:
1208 301 : result = PSCAN_INCOMPLETE;
1209 301 : *prompt = PROMPT_SINGLEQUOTE;
1210 301 : break;
1211 301 : case xq:
1212 7022 : result = PSCAN_INCOMPLETE;
1213 7022 : *prompt = PROMPT_SINGLEQUOTE;
1214 7022 : break;
1215 7022 : case xdolq:
1216 23650 : result = PSCAN_INCOMPLETE;
1217 23650 : *prompt = PROMPT_DOLLARQUOTE;
1218 23650 : break;
1219 23650 : case xui:
1220 0 : result = PSCAN_INCOMPLETE;
1221 0 : *prompt = PROMPT_DOUBLEQUOTE;
1222 0 : break;
1223 0 : case xus:
1224 0 : result = PSCAN_INCOMPLETE;
1225 0 : *prompt = PROMPT_SINGLEQUOTE;
1226 0 : break;
1227 0 : default:
1228 0 : /* can't get here */
1229 : fprintf(stderr, "invalid YY_START\n");
1230 0 : exit(1);
1231 0 : }
1232 : break;
1233 515097 : case LEXRES_SEMI: /* semicolon */
1234 242812 : result = PSCAN_SEMICOLON;
1235 242812 : *prompt = PROMPT_READY;
1236 242812 : break;
1237 242812 : case LEXRES_BACKSLASH: /* backslash */
1238 32404 : result = PSCAN_BACKSLASH;
1239 32404 : *prompt = PROMPT_READY;
1240 32404 : break;
1241 32404 : default:
1242 0 : /* can't get here */
1243 : fprintf(stderr, "invalid yylex result\n");
1244 0 : exit(1);
1245 0 : }
1246 :
1247 : return result;
1248 790313 : }
1249 :
1250 : /*
1251 : * Clean up after scanning a string. This flushes any unread input and
1252 : * releases resources (but not the PsqlScanState itself). Note however
1253 : * that this does not reset the lexer scan state; that can be done by
1254 : * psql_scan_reset(), which is an orthogonal operation.
1255 : *
1256 : * It is legal to call this when not scanning anything (makes it easier
1257 : * to deal with error recovery).
1258 : */
1259 : void
1260 : psql_scan_finish(PsqlScanState state)
1261 525422 : {
1262 : /* Drop any incomplete variable expansions. */
1263 : while (state->buffer_stack != NULL)
1264 525422 : psqlscan_pop_buffer_stack(state);
1265 0 :
1266 : /* Done with the outer scan buffer, too */
1267 : if (state->scanbufhandle)
1268 525422 : yy_delete_buffer(state->scanbufhandle, state->scanner);
1269 515367 : state->scanbufhandle = NULL;
1270 525422 : if (state->scanbuf)
1271 525422 : free(state->scanbuf);
1272 515367 : state->scanbuf = NULL;
1273 525422 : }
1274 525422 :
1275 : /*
1276 : * Reset lexer scanning state to start conditions. This is appropriate
1277 : * for executing \r psql commands (or any other time that we discard the
1278 : * prior contents of query_buf). It is not, however, necessary to do this
1279 : * when we execute and clear the buffer after getting a PSCAN_SEMICOLON or
1280 : * PSCAN_EOL scan result, because the scan state must be INITIAL when those
1281 : * conditions are returned.
1282 : *
1283 : * Note that this is unrelated to flushing unread input; that task is
1284 : * done by psql_scan_finish().
1285 : */
1286 : void
1287 : psql_scan_reset(PsqlScanState state)
1288 22911 : {
1289 : state->start_state = INITIAL;
1290 22911 : state->paren_depth = 0;
1291 22911 : state->xcdepth = 0; /* not really necessary */
1292 22911 : if (state->dolqstart)
1293 22911 : free(state->dolqstart);
1294 0 : state->dolqstart = NULL;
1295 22911 : state->identifier_count = 0;
1296 22911 : state->begin_depth = 0;
1297 22911 : }
1298 22911 :
1299 : /*
1300 : * Reselect this lexer (psqlscan.l) after using another one.
1301 : *
1302 : * Currently and for foreseeable uses, it's sufficient to reset to INITIAL
1303 : * state, because we'd never switch to another lexer in a different state.
1304 : * However, we don't want to reset e.g. paren_depth, so this can't be
1305 : * the same as psql_scan_reset().
1306 : *
1307 : * Note: psql setjmp error recovery just calls psql_scan_reset(), so that
1308 : * must be a superset of this.
1309 : *
1310 : * Note: it seems likely that other lexers could just assign INITIAL for
1311 : * themselves, since that probably has the value zero in every flex-generated
1312 : * lexer. But let's not assume that.
1313 : */
1314 : void
1315 : psql_scan_reselect_sql_lexer(PsqlScanState state)
1316 153952 : {
1317 : state->start_state = INITIAL;
1318 153952 : }
1319 153952 :
1320 : /*
1321 : * Return true if lexer is currently in an "inside quotes" state.
1322 : *
1323 : * This is pretty grotty but is needed to preserve the old behavior
1324 : * that mainloop.c drops blank lines not inside quotes without even
1325 : * echoing them.
1326 : */
1327 : bool
1328 : psql_scan_in_quote(PsqlScanState state)
1329 99459 : {
1330 : return state->start_state != INITIAL &&
1331 100073 : state->start_state != xqs;
1332 614 : }
1333 :
1334 : /*
1335 : * Return the current scanning location (end+1 of last scanned token),
1336 : * as a line number counted from 1 and an offset from string start.
1337 : *
1338 : * This considers only the outermost input string, and therefore is of
1339 : * limited use for programs that use psqlscan_push_new_buffer().
1340 : *
1341 : * It would be a bit easier probably to use "%option yylineno" to count
1342 : * lines, but the flex manual says that has a performance cost, and only
1343 : * a minority of programs using psqlscan have need for this functionality.
1344 : * So we implement it ourselves without adding overhead to the lexer itself.
1345 : */
1346 : void
1347 : psql_scan_get_location(PsqlScanState state,
1348 1737 : int *lineno, int *offset)
1349 : {
1350 : const char *line_end;
1351 :
1352 : /*
1353 : * We rely on flex's having stored a NUL after the current token in
1354 : * scanbuf. Therefore we must specially handle the state before yylex()
1355 : * has been called, when obviously that won't have happened yet.
1356 : */
1357 : if (state->cur_line_no == 0)
1358 1737 : {
1359 : *lineno = 1;
1360 0 : *offset = 0;
1361 0 : return;
1362 0 : }
1363 :
1364 : /*
1365 : * Advance cur_line_no/cur_line_ptr past whatever has been lexed so far.
1366 : * Doing this prevents repeated calls from being O(N^2) for long inputs.
1367 : */
1368 : while ((line_end = strchr(state->cur_line_ptr, '\n')) != NULL)
1369 2210 : {
1370 : state->cur_line_no++;
1371 473 : state->cur_line_ptr = line_end + 1;
1372 473 : }
1373 : state->cur_line_ptr += strlen(state->cur_line_ptr);
1374 1737 :
1375 : /* Report current location. */
1376 : *lineno = state->cur_line_no;
1377 1737 : *offset = state->cur_line_ptr - state->scanbuf;
1378 1737 : }
1379 :
1380 : /*
1381 : * Push the given string onto the stack of stuff to scan.
1382 : *
1383 : * NOTE SIDE EFFECT: the new buffer is made the active flex input buffer.
1384 : */
1385 : void
1386 : psqlscan_push_new_buffer(PsqlScanState state, const char *newstr,
1387 878 : const char *varname)
1388 : {
1389 : StackElem *stackelem;
1390 :
1391 : stackelem = pg_malloc_object(StackElem);
1392 878 :
1393 : /*
1394 : * In current usage, the passed varname points at the current flex input
1395 : * buffer; we must copy it before calling psqlscan_prepare_buffer()
1396 : * because that will change the buffer state.
1397 : */
1398 : stackelem->varname = varname ? pg_strdup(varname) : NULL;
1399 878 :
1400 : stackelem->buf = psqlscan_prepare_buffer(state, newstr, strlen(newstr),
1401 878 : &stackelem->bufstring);
1402 : state->curline = stackelem->bufstring;
1403 878 : if (state->safe_encoding)
1404 878 : {
1405 : stackelem->origstring = NULL;
1406 878 : state->refline = stackelem->bufstring;
1407 878 : }
1408 : else
1409 : {
1410 : stackelem->origstring = pg_strdup(newstr);
1411 0 : state->refline = stackelem->origstring;
1412 0 : }
1413 : stackelem->next = state->buffer_stack;
1414 878 : state->buffer_stack = stackelem;
1415 878 : }
1416 878 :
1417 : /*
1418 : * Pop the topmost buffer stack item (there must be one!)
1419 : *
1420 : * NB: after this, the flex input state is unspecified; caller must
1421 : * switch to an appropriate buffer to continue lexing.
1422 : * See psqlscan_select_top_buffer().
1423 : */
1424 : void
1425 : psqlscan_pop_buffer_stack(PsqlScanState state)
1426 878 : {
1427 : StackElem *stackelem = state->buffer_stack;
1428 878 :
1429 : state->buffer_stack = stackelem->next;
1430 878 : yy_delete_buffer(stackelem->buf, state->scanner);
1431 878 : free(stackelem->bufstring);
1432 878 : if (stackelem->origstring)
1433 878 : free(stackelem->origstring);
1434 0 : if (stackelem->varname)
1435 878 : free(stackelem->varname);
1436 878 : free(stackelem);
1437 878 : }
1438 878 :
1439 : /*
1440 : * Select the topmost surviving buffer as the active input.
1441 : */
1442 : void
1443 : psqlscan_select_top_buffer(PsqlScanState state)
1444 878 : {
1445 : StackElem *stackelem = state->buffer_stack;
1446 878 :
1447 : if (stackelem != NULL)
1448 878 : {
1449 : yy_switch_to_buffer(stackelem->buf, state->scanner);
1450 0 : state->curline = stackelem->bufstring;
1451 0 : state->refline = stackelem->origstring ? stackelem->origstring : stackelem->bufstring;
1452 0 : }
1453 : else
1454 : {
1455 : yy_switch_to_buffer(state->scanbufhandle, state->scanner);
1456 878 : state->curline = state->scanbuf;
1457 878 : state->refline = state->scanline;
1458 878 : }
1459 : }
1460 878 :
1461 : /*
1462 : * Check if specified variable name is the source for any string
1463 : * currently being scanned
1464 : */
1465 : bool
1466 : psqlscan_var_is_current_source(PsqlScanState state, const char *varname)
1467 878 : {
1468 : StackElem *stackelem;
1469 :
1470 : for (stackelem = state->buffer_stack;
1471 878 : stackelem != NULL;
1472 878 : stackelem = stackelem->next)
1473 0 : {
1474 : if (stackelem->varname && strcmp(stackelem->varname, varname) == 0)
1475 0 : return true;
1476 0 : }
1477 : return false;
1478 878 : }
1479 :
1480 : /*
1481 : * Set up a flex input buffer to scan the given data. We always make a
1482 : * copy of the data. If working in an unsafe encoding, the copy has
1483 : * multibyte sequences replaced by FFs to avoid fooling the lexer rules.
1484 : *
1485 : * NOTE SIDE EFFECT: the new buffer is made the active flex input buffer.
1486 : */
1487 : YY_BUFFER_STATE
1488 : psqlscan_prepare_buffer(PsqlScanState state, const char *txt, int len,
1489 516300 : char **txtcopy)
1490 : {
1491 : char *newtxt;
1492 :
1493 : /* Flex wants two \0 characters after the actual data */
1494 : newtxt = pg_malloc_array(char, (len + 2));
1495 516300 : *txtcopy = newtxt;
1496 516300 : newtxt[len] = newtxt[len + 1] = YY_END_OF_BUFFER_CHAR;
1497 516300 :
1498 : if (state->safe_encoding)
1499 516300 : memcpy(newtxt, txt, len);
1500 516160 : else
1501 : {
1502 : /* Gotta do it the hard way */
1503 : int i = 0;
1504 140 :
1505 : while (i < len)
1506 808 : {
1507 : int thislen = PQmblen(txt + i, state->encoding);
1508 668 :
1509 : /* first byte should always be okay... */
1510 : newtxt[i] = txt[i];
1511 668 : i++;
1512 668 : while (--thislen > 0 && i < len)
1513 808 : newtxt[i++] = (char) 0xFF;
1514 140 : }
1515 : }
1516 :
1517 : return yy_scan_buffer(newtxt, len + 2, state->scanner);
1518 516300 : }
1519 :
1520 : /*
1521 : * psqlscan_emit() --- body for ECHO macro
1522 : *
1523 : * NB: this must be used for ALL and ONLY the text copied from the flex
1524 : * input data. If you pass it something that is not part of the yytext
1525 : * string, you are making a mistake. Internally generated text can be
1526 : * appended directly to state->output_buf.
1527 : */
1528 : void
1529 : psqlscan_emit(PsqlScanState state, const char *txt, int len)
1530 6523216 : {
1531 : PQExpBuffer output_buf = state->output_buf;
1532 6523216 :
1533 : if (state->safe_encoding)
1534 6523216 : appendBinaryPQExpBuffer(output_buf, txt, len);
1535 6522740 : else
1536 : {
1537 : /* Gotta do it the hard way */
1538 : const char *reference = state->refline;
1539 476 : int i;
1540 :
1541 : reference += (txt - state->curline);
1542 476 :
1543 : for (i = 0; i < len; i++)
1544 1277 : {
1545 : char ch = txt[i];
1546 801 :
1547 : if (ch == (char) 0xFF)
1548 801 : ch = reference[i];
1549 140 : appendPQExpBufferChar(output_buf, ch);
1550 801 : }
1551 : }
1552 : }
1553 6523216 :
1554 : /*
1555 : * psqlscan_extract_substring --- fetch value of (part of) the current token
1556 : *
1557 : * This is like psqlscan_emit(), except that the data is returned as a
1558 : * malloc'd string rather than being pushed directly to state->output_buf.
1559 : */
1560 : char *
1561 : psqlscan_extract_substring(PsqlScanState state, const char *txt, int len)
1562 3324 : {
1563 : char *result = pg_malloc_array(char, (len + 1));
1564 3324 :
1565 : if (state->safe_encoding)
1566 3324 : memcpy(result, txt, len);
1567 3324 : else
1568 : {
1569 : /* Gotta do it the hard way */
1570 : const char *reference = state->refline;
1571 0 : int i;
1572 :
1573 : reference += (txt - state->curline);
1574 0 :
1575 : for (i = 0; i < len; i++)
1576 0 : {
1577 : char ch = txt[i];
1578 0 :
1579 : if (ch == (char) 0xFF)
1580 0 : ch = reference[i];
1581 0 : result[i] = ch;
1582 0 : }
1583 : }
1584 : result[len] = '\0';
1585 3324 : return result;
1586 3324 : }
1587 :
1588 : /*
1589 : * psqlscan_escape_variable --- process :'VARIABLE' or :"VARIABLE"
1590 : *
1591 : * If the variable name is found, escape its value using the appropriate
1592 : * quoting method and emit the value to output_buf. (Since the result is
1593 : * surely quoted, there is never any reason to rescan it.) If we don't
1594 : * find the variable or escaping fails, emit the token as-is.
1595 : */
1596 : void
1597 : psqlscan_escape_variable(PsqlScanState state, const char *txt, int len,
1598 681 : PsqlScanQuoteType quote)
1599 : {
1600 : char *varname;
1601 : char *value;
1602 :
1603 : /* Variable lookup. */
1604 : varname = psqlscan_extract_substring(state, txt + 2, len - 3);
1605 681 : if (state->callbacks->get_variable)
1606 681 : value = state->callbacks->get_variable(varname, quote,
1607 681 : state->cb_passthrough);
1608 : else
1609 : value = NULL;
1610 0 : free(varname);
1611 681 :
1612 : if (value)
1613 681 : {
1614 : /* Emit the suitably-escaped value */
1615 : appendPQExpBufferStr(state->output_buf, value);
1616 644 : free(value);
1617 644 : }
1618 : else
1619 : {
1620 : /* Emit original token as-is */
1621 : psqlscan_emit(state, txt, len);
1622 37 : }
1623 : }
1624 681 :
1625 : void
1626 : psqlscan_test_variable(PsqlScanState state, const char *txt, int len)
1627 21 : {
1628 : char *varname;
1629 : char *value;
1630 :
1631 : varname = psqlscan_extract_substring(state, txt + 3, len - 4);
1632 21 : if (state->callbacks->get_variable)
1633 21 : value = state->callbacks->get_variable(varname, PQUOTE_PLAIN,
1634 21 : state->cb_passthrough);
1635 : else
1636 : value = NULL;
1637 0 : free(varname);
1638 21 :
1639 : if (value != NULL)
1640 21 : {
1641 : appendPQExpBufferStr(state->output_buf, "TRUE");
1642 9 : free(value);
1643 9 : }
1644 : else
1645 : {
1646 : appendPQExpBufferStr(state->output_buf, "FALSE");
1647 12 : }
1648 : }
1649 21 : /* END: function "psqlscan_test_variable" */
|
