1 /*-------------------------------------------------------------------------
4 * routines to convert a string (legal ascii representation of node) back
7 * Portions Copyright (c) 1996-2012, PostgreSQL Global Development Group
8 * Portions Copyright (c) 1994, Regents of the University of California
12 * src/backend/nodes/read.c
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 (void) val; /* avoid compiler warning if unused */
228 if (endptr != token + length || errno == ERANGE
229 #ifdef HAVE_LONG_INT_64
230 /* if long > 32 bits, check for overflow of int4 */
231 || val != (long) ((int32) val)
239 * these three cases do not need length checks, since pg_strtok() will
240 * always treat them as single-byte tokens
242 else if (*token == '(')
244 else if (*token == ')')
245 retval = RIGHT_PAREN;
246 else if (*token == '{')
248 else if (*token == '\"' && length > 1 && token[length - 1] == '\"')
250 else if (*token == 'b')
251 retval = T_BitString;
253 retval = OTHER_TOKEN;
259 * Slightly higher-level reader.
261 * This routine applies some semantic knowledge on top of the purely
262 * lexical tokenizer pg_strtok(). It can read
263 * * Value token nodes (integers, floats, or strings);
264 * * General nodes (via parseNodeString() from readfuncs.c);
265 * * Lists of the above;
266 * * Lists of integers or OIDs.
267 * The return value is declared void *, not Node *, to avoid having to
268 * cast it explicitly in callers that assign to fields of different types.
270 * External callers should always pass NULL/0 for the arguments. Internally
271 * a non-NULL token may be passed when the upper recursion level has already
272 * scanned the first token of a node's representation.
274 * We assume pg_strtok is already initialized with a string to read (hence
275 * this should only be invoked from within a stringToNode operation).
278 nodeRead(char *token, int tok_len)
283 if (token == NULL) /* need to read a token? */
285 token = pg_strtok(&tok_len);
287 if (token == NULL) /* end of input */
291 type = nodeTokenType(token, tok_len);
296 result = parseNodeString();
297 token = pg_strtok(&tok_len);
298 if (token == NULL || token[0] != '}')
299 elog(ERROR, "did not find '}' at end of input node");
306 * Could be an integer list: (i int int ...)
307 * or an OID list: (o int int ...)
308 * or a list of nodes/values: (node node ...)
311 token = pg_strtok(&tok_len);
313 elog(ERROR, "unterminated List structure");
314 if (tok_len == 1 && token[0] == 'i')
316 /* List of integers */
322 token = pg_strtok(&tok_len);
324 elog(ERROR, "unterminated List structure");
327 val = (int) strtol(token, &endptr, 10);
328 if (endptr != token + tok_len)
329 elog(ERROR, "unrecognized integer: \"%.*s\"",
331 l = lappend_int(l, val);
334 else if (tok_len == 1 && token[0] == 'o')
342 token = pg_strtok(&tok_len);
344 elog(ERROR, "unterminated List structure");
347 val = (Oid) strtoul(token, &endptr, 10);
348 if (endptr != token + tok_len)
349 elog(ERROR, "unrecognized OID: \"%.*s\"",
351 l = lappend_oid(l, val);
356 /* List of other node types */
359 /* We have already scanned next token... */
362 l = lappend(l, nodeRead(token, tok_len));
363 token = pg_strtok(&tok_len);
365 elog(ERROR, "unterminated List structure");
372 elog(ERROR, "unexpected right parenthesis");
373 result = NULL; /* keep compiler happy */
378 /* must be "<>" --- represents a null pointer */
383 elog(ERROR, "unrecognized token: \"%.*s\"", tok_len, token);
384 result = NULL; /* keep compiler happy */
390 * we know that the token terminates on a char atol will stop at
392 result = (Node *) makeInteger(atol(token));
396 char *fval = (char *) palloc(tok_len + 1);
398 memcpy(fval, token, tok_len);
399 fval[tok_len] = '\0';
400 result = (Node *) makeFloat(fval);
404 /* need to remove leading and trailing quotes, and backslashes */
405 result = (Node *) makeString(debackslash(token + 1, tok_len - 2));
409 char *val = palloc(tok_len);
411 /* skip leading 'b' */
412 memcpy(val, token + 1, tok_len - 1);
413 val[tok_len - 1] = '\0';
414 result = (Node *) makeBitString(val);
418 elog(ERROR, "unrecognized node type: %d", (int) type);
419 result = NULL; /* keep compiler happy */
423 return (void *) result;