/*-------------------------------------------------------------------------
*
- * read.c--
- * routines to convert a string (legal ascii representation of node) back
- * to nodes
+ * read.c
+ * routines to convert a string (legal ascii representation of node) back
+ * to nodes
*
- * Copyright (c) 1994, Regents of the University of California
+ * Portions Copyright (c) 1996-2018, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
- * $Header: /cvsroot/pgsql/src/backend/nodes/read.c,v 1.2 1996/11/08 05:56:48 momjian Exp $
+ * src/backend/nodes/read.c
*
* HISTORY
- * AUTHOR DATE MAJOR EVENT
- * Andrew Yu Nov 2, 1994 file creation
+ * AUTHOR DATE MAJOR EVENT
+ * Andrew Yu Nov 2, 1994 file creation
*
*-------------------------------------------------------------------------
*/
-#include <stdio.h>
-#include <string.h>
-#include <ctype.h>
#include "postgres.h"
+
+#include <ctype.h>
+
#include "nodes/pg_list.h"
#include "nodes/readfuncs.h"
-#include "utils/elog.h"
+#include "nodes/value.h"
+
+
+/* Static state for pg_strtok */
+static char *pg_strtok_ptr = NULL;
+
/*
* stringToNode -
- * returns a Node with a given legal ascii representation
+ * returns a Node with a given legal ASCII representation
*/
void *
stringToNode(char *str)
{
- void *retval;
+ char *save_strtok;
+ void *retval;
+
+ /*
+ * We save and restore the pre-existing state of pg_strtok. This makes the
+ * world safe for re-entrant invocation of stringToNode, without incurring
+ * a lot of notational overhead by having to pass the next-character
+ * pointer around through all the readfuncs.c code.
+ */
+ save_strtok = pg_strtok_ptr;
+
+ pg_strtok_ptr = str; /* point pg_strtok at the string to read */
+
+ retval = nodeRead(NULL, 0); /* do the reading */
- (void) lsptok(str, NULL); /* set the string used in lsptok */
- retval = nodeRead(true); /* start reading */
+ pg_strtok_ptr = save_strtok;
- return retval;
+ return retval;
}
/*****************************************************************************
*
*****************************************************************************/
-#define RIGHT_PAREN (1000000 + 1)
-#define LEFT_PAREN (1000000 + 2)
-#define PLAN_SYM (1000000 + 3)
-#define AT_SYMBOL (1000000 + 4)
-#define ATOM_TOKEN (1000000 + 5)
-
/*
- * nodeTokenType -
- * returns the type of the node token contained in token.
- * It returns one of the following valid NodeTags:
- * T_Integer, T_Float, T_String
- * and some of its own:
- * RIGHT_PAREN, LEFT_PAREN, PLAN_SYM, AT_SYMBOL, ATOM_TOKEN
+ * pg_strtok --- retrieve next "token" from a string.
+ *
+ * Works kinda like strtok, except it never modifies the source string.
+ * (Instead of storing nulls into the string, the length of the token
+ * is returned to the caller.)
+ * Also, the rules about what is a token are hard-wired rather than being
+ * configured by passing a set of terminating characters.
+ *
+ * The string is assumed to have been initialized already by stringToNode.
*
- * Assumption: the ascii representation is legal
+ * The rules for tokens are:
+ * * Whitespace (space, tab, newline) always separates tokens.
+ * * The characters '(', ')', '{', '}' form individual tokens even
+ * without any whitespace around them.
+ * * Otherwise, a token is all the characters up to the next whitespace
+ * or occurrence of one of the four special characters.
+ * * A backslash '\' can be used to quote whitespace or one of the four
+ * special characters, so that it is treated as a plain token character.
+ * Backslashes themselves must also be backslashed for consistency.
+ * Any other character can be, but need not be, backslashed as well.
+ * * If the resulting token is '<>' (with no backslash), it is returned
+ * as a non-NULL pointer to the token but with length == 0. Note that
+ * there is no other way to get a zero-length token.
+ *
+ * Returns a pointer to the start of the next token, and the length of the
+ * token (including any embedded backslashes!) in *length. If there are
+ * no more tokens, NULL and 0 are returned.
+ *
+ * NOTE: this routine doesn't remove backslashes; the caller must do so
+ * if necessary (see "debackslash").
+ *
+ * NOTE: prior to release 7.0, this routine also had a special case to treat
+ * a token starting with '"' as extending to the next '"'. This code was
+ * broken, however, since it would fail to cope with a string containing an
+ * embedded '"'. I have therefore removed this special case, and instead
+ * introduced rules for using backslashes to quote characters. Higher-level
+ * code should add backslashes to a string constant to ensure it is treated
+ * as a single token.
*/
-static NodeTag
-nodeTokenType(char *token, int length)
+char *
+pg_strtok(int *length)
{
- NodeTag retval = 0;
-
- /*
- * Check if the token is a number (decimal or integer,
- * positive or negative
- */
- if (isdigit(*token) ||
- (length>=2 && *token=='-' && isdigit(*(token+1)) ))
+ char *local_str; /* working pointer to string */
+ char *ret_str; /* start of token to return */
+
+ local_str = pg_strtok_ptr;
+
+ while (*local_str == ' ' || *local_str == '\n' || *local_str == '\t')
+ local_str++;
+
+ if (*local_str == '\0')
+ {
+ *length = 0;
+ pg_strtok_ptr = local_str;
+ return NULL; /* no more tokens */
+ }
+
+ /*
+ * Now pointing at start of next token.
+ */
+ ret_str = local_str;
+
+ if (*local_str == '(' || *local_str == ')' ||
+ *local_str == '{' || *local_str == '}')
{
- /*
- * skip the optional '-' (i.e. negative number)
- */
- if (*token == '-') {
- token++;
- }
-
- /*
- * See if there is a decimal point
- */
-
- for (; length && *token != '.'; token++, length--);
-
- /*
- * if there isn't, token's an int, otherwise it's a float.
- */
-
- retval = (*token != '.') ? T_Integer : T_Float;
+ /* special 1-character token */
+ local_str++;
}
- else if (isalpha(*token))
- retval = ATOM_TOKEN;
- else if (*token == '(')
- retval = LEFT_PAREN;
- else if (*token == ')')
- retval = RIGHT_PAREN;
- else if (*token == '@')
- retval = AT_SYMBOL;
- else if (*token == '\"')
- retval = T_String;
- else if (*token == '{')
- retval = PLAN_SYM;
- return(retval);
+ else
+ {
+ /* Normal token, possibly containing backslashes */
+ while (*local_str != '\0' &&
+ *local_str != ' ' && *local_str != '\n' &&
+ *local_str != '\t' &&
+ *local_str != '(' && *local_str != ')' &&
+ *local_str != '{' && *local_str != '}')
+ {
+ if (*local_str == '\\' && local_str[1] != '\0')
+ local_str += 2;
+ else
+ local_str++;
+ }
+ }
+
+ *length = local_str - ret_str;
+
+ /* Recognize special case for "empty" token */
+ if (*length == 2 && ret_str[0] == '<' && ret_str[1] == '>')
+ *length = 0;
+
+ pg_strtok_ptr = local_str;
+
+ return ret_str;
}
/*
- * Works kinda like strtok, except it doesn't put nulls into string.
- *
- * Returns the length in length instead. The string can be set without
- * returning a token by calling lsptok with length == NULL.
- *
+ * debackslash -
+ * create a palloc'd string holding the given token.
+ * any protective backslashes in the token are removed.
*/
char *
-lsptok(char *string, int *length)
+debackslash(char *token, int length)
{
- static char *local_str;
- char *ret_string;
-
- if (string != NULL) {
- local_str = string;
- if (length == NULL) {
- return(NULL);
+ char *result = palloc(length + 1);
+ char *ptr = result;
+
+ while (length > 0)
+ {
+ if (*token == '\\' && length > 1)
+ token++, length--;
+ *ptr++ = *token++;
+ length--;
}
- }
-
- for (; *local_str == ' '
- || *local_str == '\n'
- || *local_str == '\t'; local_str++);
-
- /*
- * Now pointing at next token.
- */
- ret_string = local_str;
- if (*local_str == '\0') return(NULL);
- *length = 1;
-
- if (*local_str == '\"') {
- for (local_str++; *local_str != '\"'; (*length)++, local_str++);
- (*length)++; local_str++;
- }else if (*local_str == ')' || *local_str == '(' ||
- *local_str == '}' || *local_str == '{') {
- local_str++;
- }else {
- for (; *local_str != ' '
- && *local_str != '\n'
- && *local_str != '\t'
- && *local_str != '{'
- && *local_str != '}'
- && *local_str != '('
- && *local_str != ')'; local_str++, (*length)++);
- (*length)--;
- }
- return(ret_string);
+ *ptr = '\0';
+ return result;
}
+#define RIGHT_PAREN (1000000 + 1)
+#define LEFT_PAREN (1000000 + 2)
+#define LEFT_BRACE (1000000 + 3)
+#define OTHER_TOKEN (1000000 + 4)
+
/*
- * This guy does all the reading.
+ * nodeTokenType -
+ * returns the type of the node token contained in token.
+ * It returns one of the following valid NodeTags:
+ * T_Integer, T_Float, T_String, T_BitString
+ * and some of its own:
+ * RIGHT_PAREN, LEFT_PAREN, LEFT_BRACE, OTHER_TOKEN
*
- * Secrets: He assumes that lsptok already has the string (see below).
- * Any callers should set read_car_only to true.
+ * Assumption: the ascii representation is legal
*/
-void *
-nodeRead(bool read_car_only)
+static NodeTag
+nodeTokenType(char *token, int length)
{
- char *token;
- NodeTag type;
- Node *this_value = NULL, *return_value = NULL;
- int tok_len;
- char tmp;
- bool make_dotted_pair_cell = false;
-
- token = lsptok(NULL, &tok_len);
-
- if (token == NULL) return(NULL);
-
- type = nodeTokenType(token, tok_len);
-
- switch(type) {
- case PLAN_SYM:
- this_value = parsePlanString();
- token = lsptok(NULL, &tok_len);
- if (token[0] != '}') return(NULL);
-
- if (!read_car_only)
- make_dotted_pair_cell = true;
- else
- make_dotted_pair_cell = false;
- break;
- case LEFT_PAREN:
- if (!read_car_only) {
- List *l = makeNode(List);
-
- lfirst(l) = nodeRead(false);
- lnext(l) = nodeRead(false);
- this_value = (Node*)l;
- }else {
- this_value = nodeRead(false);
+ NodeTag retval;
+ char *numptr;
+ int numlen;
+
+ /*
+ * Check if the token is a number
+ */
+ numptr = token;
+ numlen = length;
+ if (*numptr == '+' || *numptr == '-')
+ numptr++, numlen--;
+ if ((numlen > 0 && isdigit((unsigned char) *numptr)) ||
+ (numlen > 1 && *numptr == '.' && isdigit((unsigned char) numptr[1])))
+ {
+ /*
+ * Yes. Figure out whether it is integral or float; this requires
+ * both a syntax check and a range check. strtol() can do both for us.
+ * We know the token will end at a character that strtol will stop at,
+ * so we do not need to modify the string.
+ */
+ long val;
+ char *endptr;
+
+ errno = 0;
+ val = strtol(token, &endptr, 10);
+ if (endptr != token + length || errno == ERANGE ||
+ /* check for overflow of int */
+ val != (int) val)
+ return T_Float;
+ return T_Integer;
}
- break;
- case RIGHT_PAREN:
- this_value = NULL;
- break;
- case AT_SYMBOL:
- break;
- case ATOM_TOKEN:
- if (!strncmp(token, "nil", 3)) {
- this_value = NULL;
- /*
- * It might be "nil" but it is an atom!
- */
- if (read_car_only) {
- make_dotted_pair_cell = false;
- } else {
- make_dotted_pair_cell = true;
- }
- }else {
- tmp = token[tok_len];
- token[tok_len] = '\0';
- this_value = (Node*)pstrdup(token); /* !attention! not a Node.
- use with caution */
- token[tok_len] = tmp;
- make_dotted_pair_cell = true;
+
+ /*
+ * these three cases do not need length checks, since pg_strtok() will
+ * always treat them as single-byte tokens
+ */
+ else if (*token == '(')
+ retval = LEFT_PAREN;
+ else if (*token == ')')
+ retval = RIGHT_PAREN;
+ else if (*token == '{')
+ retval = LEFT_BRACE;
+ else if (*token == '"' && length > 1 && token[length - 1] == '"')
+ retval = T_String;
+ else if (*token == 'b')
+ retval = T_BitString;
+ else
+ retval = OTHER_TOKEN;
+ return retval;
+}
+
+/*
+ * nodeRead -
+ * Slightly higher-level reader.
+ *
+ * This routine applies some semantic knowledge on top of the purely
+ * lexical tokenizer pg_strtok(). It can read
+ * * Value token nodes (integers, floats, or strings);
+ * * General nodes (via parseNodeString() from readfuncs.c);
+ * * Lists of the above;
+ * * Lists of integers or OIDs.
+ * The return value is declared void *, not Node *, to avoid having to
+ * cast it explicitly in callers that assign to fields of different types.
+ *
+ * External callers should always pass NULL/0 for the arguments. Internally
+ * a non-NULL token may be passed when the upper recursion level has already
+ * scanned the first token of a node's representation.
+ *
+ * We assume pg_strtok is already initialized with a string to read (hence
+ * this should only be invoked from within a stringToNode operation).
+ */
+void *
+nodeRead(char *token, int tok_len)
+{
+ Node *result;
+ NodeTag type;
+
+ if (token == NULL) /* need to read a token? */
+ {
+ token = pg_strtok(&tok_len);
+
+ if (token == NULL) /* end of input */
+ return NULL;
}
- break;
- case T_Float:
- tmp = token[tok_len];
- token[tok_len] = '\0';
- this_value = (Node*)makeFloat(atof(token));
- token[tok_len] = tmp;
- make_dotted_pair_cell = true;
- break;
- case T_Integer:
- tmp = token[tok_len];
- token[tok_len] = '\0';
- this_value = (Node*)makeInteger(atoi(token));
- token[tok_len] = tmp;
- make_dotted_pair_cell = true;
- break;
- case T_String:
- tmp = token[tok_len - 1];
- token[tok_len - 1] = '\0';
- token++;
- this_value = (Node*)makeString(token); /* !! not strdup'd */
- token[tok_len - 2] = tmp;
- make_dotted_pair_cell = true;
- break;
- default:
- elog(WARN, "nodeRead: Bad type %d", type);
- break;
- }
- if (make_dotted_pair_cell) {
- List *l = makeNode(List);
-
- lfirst(l) = this_value;
- if (!read_car_only) {
- lnext(l) = nodeRead(false);
- }else {
- lnext(l) = NULL;
+
+ type = nodeTokenType(token, tok_len);
+
+ switch ((int) type)
+ {
+ case LEFT_BRACE:
+ result = parseNodeString();
+ token = pg_strtok(&tok_len);
+ if (token == NULL || token[0] != '}')
+ elog(ERROR, "did not find '}' at end of input node");
+ break;
+ case LEFT_PAREN:
+ {
+ List *l = NIL;
+
+ /*----------
+ * Could be an integer list: (i int int ...)
+ * or an OID list: (o int int ...)
+ * or a list of nodes/values: (node node ...)
+ *----------
+ */
+ token = pg_strtok(&tok_len);
+ if (token == NULL)
+ elog(ERROR, "unterminated List structure");
+ if (tok_len == 1 && token[0] == 'i')
+ {
+ /* List of integers */
+ for (;;)
+ {
+ int val;
+ char *endptr;
+
+ token = pg_strtok(&tok_len);
+ if (token == NULL)
+ elog(ERROR, "unterminated List structure");
+ if (token[0] == ')')
+ break;
+ val = (int) strtol(token, &endptr, 10);
+ if (endptr != token + tok_len)
+ elog(ERROR, "unrecognized integer: \"%.*s\"",
+ tok_len, token);
+ l = lappend_int(l, val);
+ }
+ }
+ else if (tok_len == 1 && token[0] == 'o')
+ {
+ /* List of OIDs */
+ for (;;)
+ {
+ Oid val;
+ char *endptr;
+
+ token = pg_strtok(&tok_len);
+ if (token == NULL)
+ elog(ERROR, "unterminated List structure");
+ if (token[0] == ')')
+ break;
+ val = (Oid) strtoul(token, &endptr, 10);
+ if (endptr != token + tok_len)
+ elog(ERROR, "unrecognized OID: \"%.*s\"",
+ tok_len, token);
+ l = lappend_oid(l, val);
+ }
+ }
+ else
+ {
+ /* List of other node types */
+ for (;;)
+ {
+ /* We have already scanned next token... */
+ if (token[0] == ')')
+ break;
+ l = lappend(l, nodeRead(token, tok_len));
+ token = pg_strtok(&tok_len);
+ if (token == NULL)
+ elog(ERROR, "unterminated List structure");
+ }
+ }
+ result = (Node *) l;
+ break;
+ }
+ case RIGHT_PAREN:
+ elog(ERROR, "unexpected right parenthesis");
+ result = NULL; /* keep compiler happy */
+ break;
+ case OTHER_TOKEN:
+ if (tok_len == 0)
+ {
+ /* must be "<>" --- represents a null pointer */
+ result = NULL;
+ }
+ else
+ {
+ elog(ERROR, "unrecognized token: \"%.*s\"", tok_len, token);
+ result = NULL; /* keep compiler happy */
+ }
+ break;
+ case T_Integer:
+
+ /*
+ * we know that the token terminates on a char atoi will stop at
+ */
+ result = (Node *) makeInteger(atoi(token));
+ break;
+ case T_Float:
+ {
+ char *fval = (char *) palloc(tok_len + 1);
+
+ memcpy(fval, token, tok_len);
+ fval[tok_len] = '\0';
+ result = (Node *) makeFloat(fval);
+ }
+ break;
+ case T_String:
+ /* need to remove leading and trailing quotes, and backslashes */
+ result = (Node *) makeString(debackslash(token + 1, tok_len - 2));
+ break;
+ case T_BitString:
+ {
+ char *val = palloc(tok_len);
+
+ /* skip leading 'b' */
+ memcpy(val, token + 1, tok_len - 1);
+ val[tok_len - 1] = '\0';
+ result = (Node *) makeBitString(val);
+ break;
+ }
+ default:
+ elog(ERROR, "unrecognized node type: %d", (int) type);
+ result = NULL; /* keep compiler happy */
+ break;
}
- return_value = (Node*)l;
- }else {
- return_value = this_value;
- }
- return(return_value);
-}
+ return (void *) result;
+}