Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * read.c
4 : * routines to convert a string (legal ascii representation of node) back
5 : * to nodes
6 : *
7 : * Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
8 : * Portions Copyright (c) 1994, Regents of the University of California
9 : *
10 : *
11 : * IDENTIFICATION
12 : * src/backend/nodes/read.c
13 : *
14 : * HISTORY
15 : * AUTHOR DATE MAJOR EVENT
16 : * Andrew Yu Nov 2, 1994 file creation
17 : *
18 : *-------------------------------------------------------------------------
19 : */
20 : #include "postgres.h"
21 :
22 : #include <ctype.h>
23 :
24 : #include "common/string.h"
25 : #include "nodes/bitmapset.h"
26 : #include "nodes/pg_list.h"
27 : #include "nodes/readfuncs.h"
28 : #include "nodes/value.h"
29 :
30 :
31 : /* Static state for pg_strtok */
32 : static const char *pg_strtok_ptr = NULL;
33 :
34 : /* State flag that determines how readfuncs.c should treat location fields */
35 : #ifdef DEBUG_NODE_TESTS_ENABLED
36 : bool restore_location_fields = false;
37 : #endif
38 :
39 :
40 : /*
41 : * stringToNode -
42 : * builds a Node tree from its string representation (assumed valid)
43 : *
44 : * restore_loc_fields instructs readfuncs.c whether to restore location
45 : * fields rather than set them to -1. This is currently only supported
46 : * in builds with DEBUG_NODE_TESTS_ENABLED defined.
47 : */
48 : static void *
49 2217182 : stringToNodeInternal(const char *str, bool restore_loc_fields)
50 : {
51 : void *retval;
52 : const char *save_strtok;
53 : #ifdef DEBUG_NODE_TESTS_ENABLED
54 : bool save_restore_location_fields;
55 : #endif
56 :
57 : /*
58 : * We save and restore the pre-existing state of pg_strtok. This makes the
59 : * world safe for re-entrant invocation of stringToNode, without incurring
60 : * a lot of notational overhead by having to pass the next-character
61 : * pointer around through all the readfuncs.c code.
62 : */
63 2217182 : save_strtok = pg_strtok_ptr;
64 :
65 2217182 : pg_strtok_ptr = str; /* point pg_strtok at the string to read */
66 :
67 : /*
68 : * If enabled, likewise save/restore the location field handling flag.
69 : */
70 : #ifdef DEBUG_NODE_TESTS_ENABLED
71 2217182 : save_restore_location_fields = restore_location_fields;
72 2217182 : restore_location_fields = restore_loc_fields;
73 : #endif
74 :
75 2217182 : retval = nodeRead(NULL, 0); /* do the reading */
76 :
77 2217178 : pg_strtok_ptr = save_strtok;
78 :
79 : #ifdef DEBUG_NODE_TESTS_ENABLED
80 2217178 : restore_location_fields = save_restore_location_fields;
81 : #endif
82 :
83 2217178 : return retval;
84 : }
85 :
86 : /*
87 : * Externally visible entry points
88 : */
89 : void *
90 323522 : stringToNode(const char *str)
91 : {
92 323522 : return stringToNodeInternal(str, false);
93 : }
94 :
95 : #ifdef DEBUG_NODE_TESTS_ENABLED
96 :
97 : void *
98 1893660 : stringToNodeWithLocations(const char *str)
99 : {
100 1893660 : return stringToNodeInternal(str, true);
101 : }
102 :
103 : #endif
104 :
105 :
106 : /*****************************************************************************
107 : *
108 : * the lisp token parser
109 : *
110 : *****************************************************************************/
111 :
112 : /*
113 : * pg_strtok --- retrieve next "token" from a string.
114 : *
115 : * Works kinda like strtok, except it never modifies the source string.
116 : * (Instead of storing nulls into the string, the length of the token
117 : * is returned to the caller.)
118 : * Also, the rules about what is a token are hard-wired rather than being
119 : * configured by passing a set of terminating characters.
120 : *
121 : * The string is assumed to have been initialized already by stringToNode.
122 : *
123 : * The rules for tokens are:
124 : * * Whitespace (space, tab, newline) always separates tokens.
125 : * * The characters '(', ')', '{', '}' form individual tokens even
126 : * without any whitespace around them.
127 : * * Otherwise, a token is all the characters up to the next whitespace
128 : * or occurrence of one of the four special characters.
129 : * * A backslash '\' can be used to quote whitespace or one of the four
130 : * special characters, so that it is treated as a plain token character.
131 : * Backslashes themselves must also be backslashed for consistency.
132 : * Any other character can be, but need not be, backslashed as well.
133 : * * If the resulting token is '<>' (with no backslash), it is returned
134 : * as a non-NULL pointer to the token but with length == 0. Note that
135 : * there is no other way to get a zero-length token.
136 : *
137 : * Returns a pointer to the start of the next token, and the length of the
138 : * token (including any embedded backslashes!) in *length. If there are
139 : * no more tokens, NULL and 0 are returned.
140 : *
141 : * NOTE: this routine doesn't remove backslashes; the caller must do so
142 : * if necessary (see "debackslash").
143 : *
144 : * NOTE: prior to release 7.0, this routine also had a special case to treat
145 : * a token starting with '"' as extending to the next '"'. This code was
146 : * broken, however, since it would fail to cope with a string containing an
147 : * embedded '"'. I have therefore removed this special case, and instead
148 : * introduced rules for using backslashes to quote characters. Higher-level
149 : * code should add backslashes to a string constant to ensure it is treated
150 : * as a single token.
151 : */
152 : const char *
153 1123444442 : pg_strtok(int *length)
154 : {
155 : const char *local_str; /* working pointer to string */
156 : const char *ret_str; /* start of token to return */
157 :
158 1123444442 : local_str = pg_strtok_ptr;
159 :
160 2081759316 : while (*local_str == ' ' || *local_str == '\n' || *local_str == '\t')
161 958314874 : local_str++;
162 :
163 1123444442 : if (*local_str == '\0')
164 : {
165 0 : *length = 0;
166 0 : pg_strtok_ptr = local_str;
167 0 : return NULL; /* no more tokens */
168 : }
169 :
170 : /*
171 : * Now pointing at start of next token.
172 : */
173 1123444442 : ret_str = local_str;
174 :
175 1123444442 : if (*local_str == '(' || *local_str == ')' ||
176 1052766782 : *local_str == '{' || *local_str == '}')
177 : {
178 : /* special 1-character token */
179 163320108 : local_str++;
180 : }
181 : else
182 : {
183 : /* Normal token, possibly containing backslashes */
184 6698501958 : while (*local_str != '\0' &&
185 6698468052 : *local_str != ' ' && *local_str != '\n' &&
186 5805891174 : *local_str != '\t' &&
187 5805891174 : *local_str != '(' && *local_str != ')' &&
188 5781420936 : *local_str != '{' && *local_str != '}')
189 : {
190 5738377624 : if (*local_str == '\\' && local_str[1] != '\0')
191 1685082 : local_str += 2;
192 : else
193 5736692542 : local_str++;
194 : }
195 : }
196 :
197 1123444442 : *length = local_str - ret_str;
198 :
199 : /* Recognize special case for "empty" token */
200 1123444442 : if (*length == 2 && ret_str[0] == '<' && ret_str[1] == '>')
201 58904152 : *length = 0;
202 :
203 1123444442 : pg_strtok_ptr = local_str;
204 :
205 1123444442 : return ret_str;
206 : }
207 :
208 : /*
209 : * debackslash -
210 : * create a palloc'd string holding the given token.
211 : * any protective backslashes in the token are removed.
212 : */
213 : char *
214 33561578 : debackslash(const char *token, int length)
215 : {
216 33561578 : char *result = palloc(length + 1);
217 33561578 : char *ptr = result;
218 :
219 315050886 : while (length > 0)
220 : {
221 281489308 : if (*token == '\\' && length > 1)
222 1685082 : token++, length--;
223 281489308 : *ptr++ = *token++;
224 281489308 : length--;
225 : }
226 33561578 : *ptr = '\0';
227 33561578 : return result;
228 : }
229 :
230 : #define RIGHT_PAREN (1000000 + 1)
231 : #define LEFT_PAREN (1000000 + 2)
232 : #define LEFT_BRACE (1000000 + 3)
233 : #define OTHER_TOKEN (1000000 + 4)
234 :
235 : /*
236 : * nodeTokenType -
237 : * returns the type of the node token contained in token.
238 : * It returns one of the following valid NodeTags:
239 : * T_Integer, T_Float, T_Boolean, T_String, T_BitString
240 : * and some of its own:
241 : * RIGHT_PAREN, LEFT_PAREN, LEFT_BRACE, OTHER_TOKEN
242 : *
243 : * Assumption: the ascii representation is legal
244 : */
245 : static NodeTag
246 140732922 : nodeTokenType(const char *token, int length)
247 : {
248 : NodeTag retval;
249 : const char *numptr;
250 : int numlen;
251 :
252 : /*
253 : * Check if the token is a number
254 : */
255 140732922 : numptr = token;
256 140732922 : numlen = length;
257 140732922 : if (*numptr == '+' || *numptr == '-')
258 49600 : numptr++, numlen--;
259 140732922 : if ((numlen > 0 && isdigit((unsigned char) *numptr)) ||
260 25401946 : (numlen > 1 && *numptr == '.' && isdigit((unsigned char) numptr[1])))
261 : {
262 : /*
263 : * Yes. Figure out whether it is integral or float; this requires
264 : * both a syntax check and a range check. strtoint() can do both for
265 : * us. We know the token will end at a character that strtoint will
266 : * stop at, so we do not need to modify the string.
267 : */
268 : char *endptr;
269 :
270 535348 : errno = 0;
271 535348 : (void) strtoint(numptr, &endptr, 10);
272 535348 : if (endptr != token + length || errno == ERANGE)
273 12330 : return T_Float;
274 523018 : return T_Integer;
275 : }
276 :
277 : /*
278 : * these three cases do not need length checks, since pg_strtok() will
279 : * always treat them as single-byte tokens
280 : */
281 140197574 : else if (*token == '(')
282 17597144 : retval = LEFT_PAREN;
283 122600430 : else if (*token == ')')
284 0 : retval = RIGHT_PAREN;
285 122600430 : else if (*token == '{')
286 46321232 : retval = LEFT_BRACE;
287 76279198 : else if ((length == 4 && strncmp(token, "true", 4) == 0) ||
288 1660852 : (length == 5 && strncmp(token, "false", 5) == 0))
289 90314 : retval = T_Boolean;
290 76188884 : else if (*token == '"' && length > 1 && token[length - 1] == '"')
291 25307542 : retval = T_String;
292 50881342 : else if (*token == 'b' || *token == 'x')
293 4068 : retval = T_BitString;
294 : else
295 50877274 : retval = OTHER_TOKEN;
296 140197574 : return retval;
297 : }
298 :
299 : /*
300 : * nodeRead -
301 : * Slightly higher-level reader.
302 : *
303 : * This routine applies some semantic knowledge on top of the purely
304 : * lexical tokenizer pg_strtok(). It can read
305 : * * Value token nodes (integers, floats, booleans, or strings);
306 : * * General nodes (via parseNodeString() from readfuncs.c);
307 : * * Lists of the above;
308 : * * Lists of integers, OIDs, or TransactionIds.
309 : * The return value is declared void *, not Node *, to avoid having to
310 : * cast it explicitly in callers that assign to fields of different types.
311 : *
312 : * External callers should always pass NULL/0 for the arguments. Internally
313 : * a non-NULL token may be passed when the upper recursion level has already
314 : * scanned the first token of a node's representation.
315 : *
316 : * We assume pg_strtok is already initialized with a string to read (hence
317 : * this should only be invoked from within a stringToNode operation).
318 : */
319 : void *
320 140732922 : nodeRead(const char *token, int tok_len)
321 : {
322 : Node *result;
323 : NodeTag type;
324 :
325 140732922 : if (token == NULL) /* need to read a token? */
326 : {
327 90205296 : token = pg_strtok(&tok_len);
328 :
329 90205296 : if (token == NULL) /* end of input */
330 0 : return NULL;
331 : }
332 :
333 140732922 : type = nodeTokenType(token, tok_len);
334 :
335 140732922 : switch ((int) type)
336 : {
337 46321232 : case LEFT_BRACE:
338 46321232 : result = parseNodeString();
339 46321216 : token = pg_strtok(&tok_len);
340 46321216 : if (token == NULL || token[0] != '}')
341 0 : elog(ERROR, "did not find '}' at end of input node");
342 46321216 : break;
343 17597144 : case LEFT_PAREN:
344 : {
345 17597144 : List *l = NIL;
346 :
347 : /*----------
348 : * Could be an integer list: (i int int ...)
349 : * or an OID list: (o int int ...)
350 : * or an XID list: (x int int ...)
351 : * or a bitmapset: (b int int ...)
352 : * or a list of nodes/values: (node node ...)
353 : *----------
354 : */
355 17597144 : token = pg_strtok(&tok_len);
356 17597144 : if (token == NULL)
357 0 : elog(ERROR, "unterminated List structure");
358 17597144 : if (tok_len == 1 && token[0] == 'i')
359 : {
360 : /* List of integers */
361 : for (;;)
362 4305588 : {
363 : int val;
364 : char *endptr;
365 :
366 4789794 : token = pg_strtok(&tok_len);
367 4789794 : if (token == NULL)
368 0 : elog(ERROR, "unterminated List structure");
369 4789794 : if (token[0] == ')')
370 484206 : break;
371 4305588 : val = (int) strtol(token, &endptr, 10);
372 4305588 : if (endptr != token + tok_len)
373 0 : elog(ERROR, "unrecognized integer: \"%.*s\"",
374 : tok_len, token);
375 4305588 : l = lappend_int(l, val);
376 : }
377 484206 : result = (Node *) l;
378 : }
379 17112938 : else if (tok_len == 1 && token[0] == 'o')
380 : {
381 : /* List of OIDs */
382 : for (;;)
383 1307516 : {
384 : Oid val;
385 : char *endptr;
386 :
387 1981546 : token = pg_strtok(&tok_len);
388 1981546 : if (token == NULL)
389 0 : elog(ERROR, "unterminated List structure");
390 1981546 : if (token[0] == ')')
391 674030 : break;
392 1307516 : val = (Oid) strtoul(token, &endptr, 10);
393 1307516 : if (endptr != token + tok_len)
394 0 : elog(ERROR, "unrecognized OID: \"%.*s\"",
395 : tok_len, token);
396 1307516 : l = lappend_oid(l, val);
397 : }
398 674030 : result = (Node *) l;
399 : }
400 16438908 : else if (tok_len == 1 && token[0] == 'x')
401 : {
402 : /* List of TransactionIds */
403 : for (;;)
404 0 : {
405 : TransactionId val;
406 : char *endptr;
407 :
408 0 : token = pg_strtok(&tok_len);
409 0 : if (token == NULL)
410 0 : elog(ERROR, "unterminated List structure");
411 0 : if (token[0] == ')')
412 0 : break;
413 0 : val = (TransactionId) strtoul(token, &endptr, 10);
414 0 : if (endptr != token + tok_len)
415 0 : elog(ERROR, "unrecognized Xid: \"%.*s\"",
416 : tok_len, token);
417 0 : l = lappend_xid(l, val);
418 : }
419 0 : result = (Node *) l;
420 : }
421 16438908 : else if (tok_len == 1 && token[0] == 'b')
422 0 : {
423 : /* Bitmapset -- see also _readBitmapset() */
424 0 : Bitmapset *bms = NULL;
425 :
426 : for (;;)
427 0 : {
428 : int val;
429 : char *endptr;
430 :
431 0 : token = pg_strtok(&tok_len);
432 0 : if (token == NULL)
433 0 : elog(ERROR, "unterminated Bitmapset structure");
434 0 : if (tok_len == 1 && token[0] == ')')
435 0 : break;
436 0 : val = (int) strtol(token, &endptr, 10);
437 0 : if (endptr != token + tok_len)
438 0 : elog(ERROR, "unrecognized integer: \"%.*s\"",
439 : tok_len, token);
440 0 : bms = bms_add_member(bms, val);
441 : }
442 0 : result = (Node *) bms;
443 : }
444 : else
445 : {
446 : /* List of other node types */
447 : for (;;)
448 : {
449 : /* We have already scanned next token... */
450 66966526 : if (token[0] == ')')
451 16438900 : break;
452 50527626 : l = lappend(l, nodeRead(token, tok_len));
453 50527618 : token = pg_strtok(&tok_len);
454 50527618 : if (token == NULL)
455 0 : elog(ERROR, "unterminated List structure");
456 : }
457 16438900 : result = (Node *) l;
458 : }
459 17597136 : break;
460 : }
461 0 : case RIGHT_PAREN:
462 0 : elog(ERROR, "unexpected right parenthesis");
463 : result = NULL; /* keep compiler happy */
464 : break;
465 50877274 : case OTHER_TOKEN:
466 50877274 : if (tok_len == 0)
467 : {
468 : /* must be "<>" --- represents a null pointer */
469 50877274 : result = NULL;
470 : }
471 : else
472 : {
473 0 : elog(ERROR, "unrecognized token: \"%.*s\"", tok_len, token);
474 : result = NULL; /* keep compiler happy */
475 : }
476 50877274 : break;
477 523018 : case T_Integer:
478 :
479 : /*
480 : * we know that the token terminates on a char atoi will stop at
481 : */
482 523018 : result = (Node *) makeInteger(atoi(token));
483 523018 : break;
484 12330 : case T_Float:
485 : {
486 12330 : char *fval = (char *) palloc(tok_len + 1);
487 :
488 12330 : memcpy(fval, token, tok_len);
489 12330 : fval[tok_len] = '\0';
490 12330 : result = (Node *) makeFloat(fval);
491 : }
492 12330 : break;
493 90314 : case T_Boolean:
494 90314 : result = (Node *) makeBoolean(token[0] == 't');
495 90314 : break;
496 25307542 : case T_String:
497 : /* need to remove leading and trailing quotes, and backslashes */
498 25307542 : result = (Node *) makeString(debackslash(token + 1, tok_len - 2));
499 25307542 : break;
500 4068 : case T_BitString:
501 : /* need to remove backslashes, but there are no quotes */
502 4068 : result = (Node *) makeBitString(debackslash(token, tok_len));
503 4068 : break;
504 0 : default:
505 0 : elog(ERROR, "unrecognized node type: %d", (int) type);
506 : result = NULL; /* keep compiler happy */
507 : break;
508 : }
509 :
510 140732898 : return (void *) result;
511 : }
|
