1
/*-------------------------------------------------------------------------
4
* routines to convert a string (legal ascii representation of node) back
7
* Portions Copyright (c) 1996-2009, PostgreSQL Global Development Group
8
* Portions Copyright (c) 1994, Regents of the University of California
15
* AUTHOR DATE MAJOR EVENT
16
* Andrew Yu Nov 2, 1994 file creation
18
*-------------------------------------------------------------------------
24
#include "nodes/pg_list.h"
25
#include "nodes/readfuncs.h"
26
#include "nodes/value.h"
29
/* Static state for pg_strtok */
30
static char *pg_strtok_ptr = NULL;
35
* returns a Node with a given legal ASCII representation
38
stringToNode(char *str)
44
* We save and restore the pre-existing state of pg_strtok. This makes the
45
* world safe for re-entrant invocation of stringToNode, without incurring
46
* a lot of notational overhead by having to pass the next-character
47
* pointer around through all the readfuncs.c code.
49
save_strtok = pg_strtok_ptr;
51
pg_strtok_ptr = str; /* point pg_strtok at the string to read */
53
retval = nodeRead(NULL, 0); /* do the reading */
55
pg_strtok_ptr = save_strtok;
60
/*****************************************************************************
62
* the lisp token parser
64
*****************************************************************************/
67
* pg_strtok --- retrieve next "token" from a string.
69
* Works kinda like strtok, except it never modifies the source string.
70
* (Instead of storing nulls into the string, the length of the token
71
* is returned to the caller.)
72
* Also, the rules about what is a token are hard-wired rather than being
73
* configured by passing a set of terminating characters.
75
* The string is assumed to have been initialized already by stringToNode.
77
* The rules for tokens are:
78
* * Whitespace (space, tab, newline) always separates tokens.
79
* * The characters '(', ')', '{', '}' form individual tokens even
80
* without any whitespace around them.
81
* * Otherwise, a token is all the characters up to the next whitespace
82
* or occurrence of one of the four special characters.
83
* * A backslash '\' can be used to quote whitespace or one of the four
84
* special characters, so that it is treated as a plain token character.
85
* Backslashes themselves must also be backslashed for consistency.
86
* Any other character can be, but need not be, backslashed as well.
87
* * If the resulting token is '<>' (with no backslash), it is returned
88
* as a non-NULL pointer to the token but with length == 0. Note that
89
* there is no other way to get a zero-length token.
91
* Returns a pointer to the start of the next token, and the length of the
92
* token (including any embedded backslashes!) in *length. If there are
93
* no more tokens, NULL and 0 are returned.
95
* NOTE: this routine doesn't remove backslashes; the caller must do so
96
* if necessary (see "debackslash").
98
* NOTE: prior to release 7.0, this routine also had a special case to treat
99
* a token starting with '"' as extending to the next '"'. This code was
100
* broken, however, since it would fail to cope with a string containing an
101
* embedded '"'. I have therefore removed this special case, and instead
102
* introduced rules for using backslashes to quote characters. Higher-level
103
* code should add backslashes to a string constant to ensure it is treated
107
pg_strtok(int *length)
109
char *local_str; /* working pointer to string */
110
char *ret_str; /* start of token to return */
112
local_str = pg_strtok_ptr;
114
while (*local_str == ' ' || *local_str == '\n' || *local_str == '\t')
117
if (*local_str == '\0')
120
pg_strtok_ptr = local_str;
121
return NULL; /* no more tokens */
125
* Now pointing at start of next token.
129
if (*local_str == '(' || *local_str == ')' ||
130
*local_str == '{' || *local_str == '}')
132
/* special 1-character token */
137
/* Normal token, possibly containing backslashes */
138
while (*local_str != '\0' &&
139
*local_str != ' ' && *local_str != '\n' &&
140
*local_str != '\t' &&
141
*local_str != '(' && *local_str != ')' &&
142
*local_str != '{' && *local_str != '}')
144
if (*local_str == '\\' && local_str[1] != '\0')
151
*length = local_str - ret_str;
153
/* Recognize special case for "empty" token */
154
if (*length == 2 && ret_str[0] == '<' && ret_str[1] == '>')
157
pg_strtok_ptr = local_str;
164
* create a palloc'd string holding the given token.
165
* any protective backslashes in the token are removed.
168
debackslash(char *token, int length)
170
char *result = palloc(length + 1);
175
if (*token == '\\' && length > 1)
184
#define RIGHT_PAREN (1000000 + 1)
185
#define LEFT_PAREN (1000000 + 2)
186
#define LEFT_BRACE (1000000 + 3)
187
#define OTHER_TOKEN (1000000 + 4)
191
* returns the type of the node token contained in token.
192
* It returns one of the following valid NodeTags:
193
* T_Integer, T_Float, T_String, T_BitString
194
* and some of its own:
195
* RIGHT_PAREN, LEFT_PAREN, LEFT_BRACE, OTHER_TOKEN
197
* Assumption: the ascii representation is legal
200
nodeTokenType(char *token, int length)
207
* Check if the token is a number
211
if (*numptr == '+' || *numptr == '-')
213
if ((numlen > 0 && isdigit((unsigned char) *numptr)) ||
214
(numlen > 1 && *numptr == '.' && isdigit((unsigned char) numptr[1])))
217
* Yes. Figure out whether it is integral or float; this requires
218
* both a syntax check and a range check. strtol() can do both for us.
219
* We know the token will end at a character that strtol will stop at,
220
* so we do not need to modify the string.
226
val = strtol(token, &endptr, 10);
227
if (endptr != token + length || errno == ERANGE
228
#ifdef HAVE_LONG_INT_64
229
/* if long > 32 bits, check for overflow of int4 */
230
|| val != (long) ((int32) val)
238
* these three cases do not need length checks, since pg_strtok() will
239
* always treat them as single-byte tokens
241
else if (*token == '(')
243
else if (*token == ')')
244
retval = RIGHT_PAREN;
245
else if (*token == '{')
247
else if (*token == '\"' && length > 1 && token[length - 1] == '\"')
249
else if (*token == 'b')
250
retval = T_BitString;
252
retval = OTHER_TOKEN;
258
* Slightly higher-level reader.
260
* This routine applies some semantic knowledge on top of the purely
261
* lexical tokenizer pg_strtok(). It can read
262
* * Value token nodes (integers, floats, or strings);
263
* * General nodes (via parseNodeString() from readfuncs.c);
264
* * Lists of the above;
265
* * Lists of integers or OIDs.
266
* The return value is declared void *, not Node *, to avoid having to
267
* cast it explicitly in callers that assign to fields of different types.
269
* External callers should always pass NULL/0 for the arguments. Internally
270
* a non-NULL token may be passed when the upper recursion level has already
271
* scanned the first token of a node's representation.
273
* We assume pg_strtok is already initialized with a string to read (hence
274
* this should only be invoked from within a stringToNode operation).
277
nodeRead(char *token, int tok_len)
282
if (token == NULL) /* need to read a token? */
284
token = pg_strtok(&tok_len);
286
if (token == NULL) /* end of input */
290
type = nodeTokenType(token, tok_len);
295
result = parseNodeString();
296
token = pg_strtok(&tok_len);
297
if (token == NULL || token[0] != '}')
298
elog(ERROR, "did not find '}' at end of input node");
305
* Could be an integer list: (i int int ...)
306
* or an OID list: (o int int ...)
307
* or a list of nodes/values: (node node ...)
310
token = pg_strtok(&tok_len);
312
elog(ERROR, "unterminated List structure");
313
if (tok_len == 1 && token[0] == 'i')
315
/* List of integers */
321
token = pg_strtok(&tok_len);
323
elog(ERROR, "unterminated List structure");
326
val = (int) strtol(token, &endptr, 10);
327
if (endptr != token + tok_len)
328
elog(ERROR, "unrecognized integer: \"%.*s\"",
330
l = lappend_int(l, val);
333
else if (tok_len == 1 && token[0] == 'o')
341
token = pg_strtok(&tok_len);
343
elog(ERROR, "unterminated List structure");
346
val = (Oid) strtoul(token, &endptr, 10);
347
if (endptr != token + tok_len)
348
elog(ERROR, "unrecognized OID: \"%.*s\"",
350
l = lappend_oid(l, val);
355
/* List of other node types */
358
/* We have already scanned next token... */
361
l = lappend(l, nodeRead(token, tok_len));
362
token = pg_strtok(&tok_len);
364
elog(ERROR, "unterminated List structure");
371
elog(ERROR, "unexpected right parenthesis");
372
result = NULL; /* keep compiler happy */
377
/* must be "<>" --- represents a null pointer */
382
elog(ERROR, "unrecognized token: \"%.*s\"", tok_len, token);
383
result = NULL; /* keep compiler happy */
389
* we know that the token terminates on a char atol will stop at
391
result = (Node *) makeInteger(atol(token));
395
char *fval = (char *) palloc(tok_len + 1);
397
memcpy(fval, token, tok_len);
398
fval[tok_len] = '\0';
399
result = (Node *) makeFloat(fval);
403
/* need to remove leading and trailing quotes, and backslashes */
404
result = (Node *) makeString(debackslash(token + 1, tok_len - 2));
408
char *val = palloc(tok_len);
410
/* skip leading 'b' */
411
memcpy(val, token + 1, tok_len - 1);
412
val[tok_len - 1] = '\0';
413
result = (Node *) makeBitString(val);
417
elog(ERROR, "unrecognized node type: %d", (int) type);
418
result = NULL; /* keep compiler happy */
422
return (void *) result;