AMK's latest
diff --git a/Modules/pcre-int.h b/Modules/pcre-int.h
index 2c34dfe..07aeb84 100644
--- a/Modules/pcre-int.h
+++ b/Modules/pcre-int.h
@@ -3,7 +3,7 @@
*************************************************/
-#define PCRE_VERSION "1.04 22-Dec-1997"
+#define PCRE_VERSION "1.07 16-Feb-1998"
/* This is a library of functions to support regular expressions whose syntax
@@ -12,7 +12,7 @@
Written by: Philip Hazel <ph10@cam.ac.uk>
- Copyright (c) 1997 University of Cambridge
+ Copyright (c) 1998 University of Cambridge
-----------------------------------------------------------------------------
Permission is granted to anyone to use this software for any purpose on any
@@ -192,6 +192,7 @@
OP_CRMINRANGE,
OP_CLASS, /* Match a character class */
+ OP_NEGCLASS, /* Match a character class, specified negatively */
OP_CLASS_L, /* Match a character class */
OP_REF, /* Match a back reference */
diff --git a/Modules/pcre.h b/Modules/pcre.h
index 4a01bd2..06768a9 100644
--- a/Modules/pcre.h
+++ b/Modules/pcre.h
@@ -2,7 +2,7 @@
* Perl-Compatible Regular Expressions *
*************************************************/
-/* Copyright (c) 1997 University of Cambridge */
+/* Copyright (c) 1998 University of Cambridge */
#ifndef _PCRE_H
#define _PCRE_H
@@ -17,6 +17,12 @@
#include <sys/types.h>
#include <stdlib.h>
+/* Allow for C++ users */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
/* Options */
#define PCRE_CASELESS 0x0001
@@ -68,4 +74,8 @@
extern pcre_extra *pcre_study(const pcre *, int, const char **);
extern const char *pcre_version(void);
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
#endif /* End of pcre.h */
diff --git a/Modules/pcremodule.c b/Modules/pcremodule.c
index c4890df..07a36fa 100644
--- a/Modules/pcremodule.c
+++ b/Modules/pcremodule.c
@@ -72,7 +72,7 @@
#define NOT_WORD_BOUNDARY 6
#define BEGINNING_OF_BUFFER 7
#define END_OF_BUFFER 8
-
+#define STRING 9
static PcreObject *
newPcreObject(arg)
@@ -191,49 +191,20 @@
{
PcreObject *rv;
PyObject *dictionary;
- char *pattern, *newpattern;
+ char *pattern;
const char *error;
int num_zeros, i, j;
- int patternlen, options, erroroffset;
- if (!PyArg_ParseTuple(args, "s#iO!", &pattern, &patternlen, &options,
+ int options, erroroffset;
+ if (!PyArg_ParseTuple(args, "siO!", &pattern, &options,
&PyDict_Type, &dictionary))
return NULL;
rv = newPcreObject(args);
if ( rv == NULL )
return NULL;
- /* PCRE doesn't like having null bytes in its pattern, so we have to replace
- any zeros in the string with the characters '\000'. This increases the size
- of the string by 3*num_zeros, plus 1 byte for the terminating \0. */
- num_zeros=1; /* Start at 1; this will give 3 extra bytes of leeway */
- for(i=0; i<patternlen; i++) {
- if (pattern[i]==0) num_zeros++;
- }
- newpattern=malloc(patternlen + num_zeros*3 + 4);
- if (newpattern==NULL) {
- PyErr_SetString(PyExc_MemoryError, "can't allocate memory for new pattern");
- return NULL;
- }
- for (i=j=0; i<patternlen; i++, j++)
- {
- if (pattern[i]!=0) newpattern[j]=pattern[i];
- else {
- newpattern[j++] ='\\';
- newpattern[j++] = '0';
- newpattern[j++] = '0';
- newpattern[j ] = '0';
- }
- }
- /* Keep purify happy; for pcre, one null byte is enough! */
- newpattern[j++]='\0';
- newpattern[j++]='\0';
- newpattern[j++]='\0';
- newpattern[j]='\0';
-
- rv->regex = pcre_compile((char*)newpattern, options,
+ rv->regex = pcre_compile((char*)pattern, options,
&error, &erroroffset, dictionary);
- free(newpattern);
if (rv->regex==NULL)
{
PyMem_DEL(rv);
@@ -312,6 +283,10 @@
*indexptr=index;
return Py_BuildValue("c", (char)8);
break;
+ case('\\'):
+ *indexptr=index;
+ return Py_BuildValue("c", '\\');
+ break;
case('x'):
{
@@ -348,6 +323,8 @@
case('g'):
{
int end, i;
+ int group_num = 0, is_number=0;
+
if (pattern_len<=index)
{
PyErr_SetString(ErrorObject, "unfinished symbolic reference");
@@ -374,16 +351,22 @@
PyErr_SetString(ErrorObject, "zero-length symbolic reference");
return NULL;
}
- if (!(pcre_ctypes[pattern[index]] & ctype_word) /* First char. not alphanumeric */
- || (pcre_ctypes[pattern[index]] & ctype_digit) ) /* First char. a digit */
+ if ((pcre_ctypes[pattern[index]] & ctype_digit)) /* First char. a digit */
{
- /* XXX should include the text of the reference */
- PyErr_SetString(ErrorObject, "first character of symbolic reference not a letter or _");
- return NULL;
+ is_number = 1;
+ group_num = pattern[index] - '0';
}
for(i=index+1; i<end; i++)
{
+ if (is_number &&
+ !(pcre_ctypes[pattern[i]] & ctype_digit) )
+ {
+ /* XXX should include the text of the reference */
+ PyErr_SetString(ErrorObject, "illegal non-digit character in \\g<...> starting with digit");
+ return NULL;
+ }
+ else {group_num = group_num * 10 + pattern[i] - '0';}
if (!(pcre_ctypes[pattern[i]] & ctype_word) )
{
/* XXX should include the text of the reference */
@@ -394,6 +377,9 @@
*typeptr = MEMORY_REFERENCE;
*indexptr = end+1;
+ /* If it's a number, return the integer value of the group */
+ if (is_number) return Py_BuildValue("i", group_num);
+ /* Otherwise, return a string containing the group name */
return Py_BuildValue("s#", pattern+index, end-index);
}
break;
@@ -478,8 +464,11 @@
break;
default:
+ /* It's some unknown escape like \s, so return a string containing
+ \s */
+ *typeptr = STRING;
*indexptr = index;
- return Py_BuildValue("c", c);
+ return Py_BuildValue("s#", pattern+index-2, 2);
break;
}
}
@@ -571,6 +560,12 @@
Py_DECREF(result);
}
break;
+ case(STRING):
+ {
+ PyList_Append(results, value);
+ total_len += PyString_Size(value);
+ break;
+ }
default:
Py_DECREF(results);
PyErr_SetString(ErrorObject,
diff --git a/Modules/pypcre.c b/Modules/pypcre.c
index 796f3b4..bec9197 100644
--- a/Modules/pypcre.c
+++ b/Modules/pypcre.c
@@ -211,7 +211,7 @@
Written by: Philip Hazel <ph10@cam.ac.uk>
- Copyright (c) 1997 University of Cambridge
+ Copyright (c) 1998 University of Cambridge
-----------------------------------------------------------------------------
Permission is granted to anyone to use this software for any purpose on any
@@ -409,6 +409,7 @@
according to the repeat count. */
case OP_CLASS:
+ case OP_NEGCLASS:
{
tcode++;
for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
@@ -547,7 +548,7 @@
Written by: Philip Hazel <ph10@cam.ac.uk>
- Copyright (c) 1997 University of Cambridge
+ Copyright (c) 1998 University of Cambridge
-----------------------------------------------------------------------------
Permission is granted to anyone to use this software for any purpose on any
@@ -586,18 +587,26 @@
+
#ifndef Py_eval_input
/* For Python 1.4, graminit.h has to be explicitly included */
#define Py_eval_input eval_input
#endif /* FOR_PYTHON */
+/* Allow compilation as C++ source code, should anybody want to do that. */
+
+#ifdef __cplusplus
+#define class pcre_class
+#endif
+
+
/* Min and max values for the common repeats; for the maxima, 0 => infinity */
-static char rep_min[] = { 0, 0, 1, 1, 0, 0 };
-static char rep_max[] = { 0, 0, 0, 0, 1, 1 };
+static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
+static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
-/* Text forms of OP_ values and things, for debugging */
+/* Text forms of OP_ values and things, for debugging (not all used) */
#ifdef DEBUG
static const char *OP_names[] = {
@@ -610,7 +619,7 @@
"*", "*?", "+", "+?", "?", "??", "{", "{", "{",
"*", "*?", "+", "+?", "?", "??", "{", "{", "{",
"*", "*?", "+", "+?", "?", "??", "{", "{",
- "class", "classL", "Ref",
+ "class", "negclass", "classL", "Ref",
"Alt", "Ket", "KetRmax", "KetRmin", "Assert", "Assert not", "Once",
"Brazero", "Braminzero", "Bra"
};
@@ -621,7 +630,7 @@
on. Zero means further processing is needed (for things like \x), or the escape
is invalid. */
-static short int escapes[] = {
+static const short int escapes[] = {
0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 7 */
0, 0, ':', ';', '<', '=', '>', '?', /* 8 - ? */
'@', -ESC_A, -ESC_B, 0, -ESC_D, 0, 0, 0, /* @ - G */
@@ -636,8 +645,9 @@
/* Definition to allow mutual recursion */
-static BOOL compile_regex(int, int *, uschar **, const uschar **,
- const char **, PyObject *);
+static BOOL
+compile_regex(int, int *, uschar **, const uschar **, const char **,
+ PyObject *);
/* Structure for passing "static" information around between the functions
doing the matching, so that they are thread-safe. */
@@ -866,12 +876,13 @@
/* Check a class or a back reference for a zero minimum */
case OP_CLASS:
+ case OP_NEGCLASS:
case OP_REF:
case OP_CLASS_L:
switch(*cc)
{
case (OP_REF): cc += 2; break;
- case (OP_CLASS): cc += 1+32; break;
+ case (OP_CLASS): case (OP_NEGCLASS): cc += 1+32; break;
case (OP_CLASS_L): cc += 1+1+32; break;
}
@@ -1017,15 +1028,17 @@
{
/* PYTHON: Try to compute an octal value for a character */
- for(c=0, i=0; c!=-1 && ptr[i]!=0 && i<3; i++)
+ for(c=0, i=0; ptr[i]!=0 && i<3; i++)
{
if (( pcre_ctypes[ ptr[i] ] & ctype_odigit) != 0)
c = c * 8 + ptr[i]-'0';
else
- c = -1; /* Non-octal character */
+ break; /* Non-octal character--break out of the loop */
}
- /* Aha! There were 3 octal digits, so it must be a character */
- if (c != -1 && i == 3)
+ /* It's a character if there were exactly 3 octal digits, or if
+ we're inside a character class and there was at least one
+ octal digit. */
+ if ( (i == 3) || (isclass && i!=0) )
{
ptr += i-1;
break;
@@ -1278,11 +1291,14 @@
class_flag = NULL;
}
- /* If the first character is '^', set the negation flag */
+ /* If the first character is '^', set the negation flag, and use a
+ different opcode. This only matters if caseless matching is specified at
+ runtime. */
if ((c = *(++ptr)) == '^')
{
negate_class = TRUE;
+ if (*(code-1)==OP_CLASS) *(code-1) = OP_NEGCLASS;
c = *(++ptr);
}
else negate_class = FALSE;
@@ -1648,7 +1664,8 @@
/* If previous was a character class or a back reference, we put the repeat
stuff after it. */
- else if (*previous == OP_CLASS || *previous==OP_CLASS_L || *previous == OP_REF)
+ else if (*previous == OP_CLASS || *previous == OP_NEGCLASS ||
+ *previous==OP_CLASS_L || *previous == OP_REF)
{
if (repeat_min == 0 && repeat_max == -1)
*code++ = OP_CRSTAR + repeat_type;
@@ -2003,7 +2020,7 @@
the next state. */
previous[1] = length;
- ptr--;
+ if (length < 255) ptr--;
break;
}
} /* end of big loop */
@@ -2832,6 +2849,7 @@
goto CLASS_REF_REPEAT;
case OP_CLASS:
+ case OP_NEGCLASS:
case OP_CLASS_L:
{
int i, min, max;
@@ -2840,11 +2858,14 @@
{
code++;
printf("Locflag = %i ", *code++);
+ printf(" [");
}
else
- code++;
-
- printf(" [");
+ {
+ if (*code++ == OP_CLASS) printf(" [");
+ else printf(" ^[");
+ }
+
for (i = 0; i < 256; i++)
{
@@ -3601,10 +3622,14 @@
item to see if there is repeat information following. Then obey similar
code to character type repeats - written out again for speed. If caseless
matching was set at runtime but not at compile time, we have to check both
- versions of a character. */
+ versions of a character, and we have to behave differently for positive and
+ negative classes. This is the only time where OP_CLASS and OP_NEGCLASS are
+ treated differently. */
case OP_CLASS:
+ case OP_NEGCLASS:
{
+ BOOL nasty_case = *ecode == OP_NEGCLASS && md->runtime_caseless;
const uschar *data = ecode + 1; /* Save for matching */
ecode += 33; /* Advance past the item */
@@ -3633,15 +3658,8 @@
break;
default: /* No repeat follows */
- if (eptr >= md->end_subject) FAIL;
- c = *eptr++;
- if ((data[c/8] & (1 << (c&7))) != 0) continue; /* With main loop */
- if (md->runtime_caseless)
- {
- c = pcre_fcc[c];
- if ((data[c/8] & (1 << (c&7))) != 0) continue; /* With main loop */
- }
- FAIL;
+ min = max = 1;
+ break;
}
/* First, ensure the minimum number of matches are present. */
@@ -3650,13 +3668,31 @@
{
if (eptr >= md->end_subject) FAIL;
c = *eptr++;
- if ((data[c/8] & (1 << (c&7))) != 0) continue;
- if (md->runtime_caseless)
+
+ /* Either not runtime caseless, or it was a positive class. For
+ runtime caseless, continue if either case is in the map. */
+
+ if (!nasty_case)
{
- c = pcre_fcc[c];
if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ if (md->runtime_caseless)
+ {
+ c = pcre_fcc[c];
+ if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ }
}
- FAIL;
+
+ /* Runtime caseless and it was a negative class. Continue only if
+ both cases are in the map. */
+
+ else
+ {
+ if ((data[c/8] & (1 << (c&7))) == 0) FAIL;
+ c = pcre_fcc[c];
+ if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ }
+
+ FAIL;
}
/* If max == min we can continue with the main loop without the
@@ -3674,12 +3710,30 @@
if (match(eptr, ecode, offset_top, md)) SUCCEED;
if (i >= max || eptr >= md->end_subject) FAIL;
c = *eptr++;
- if ((data[c/8] & (1 << (c&7))) != 0) continue;
- if (md->runtime_caseless)
+
+ /* Either not runtime caseless, or it was a positive class. For
+ runtime caseless, continue if either case is in the map. */
+
+ if (!nasty_case)
{
- c = pcre_fcc[c];
if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ if (md->runtime_caseless)
+ {
+ c = pcre_fcc[c];
+ if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ }
}
+
+ /* Runtime caseless and it was a negative class. Continue only if
+ both cases are in the map. */
+
+ else
+ {
+ if ((data[c/8] & (1 << (c&7))) == 0) return FALSE;
+ c = pcre_fcc[c];
+ if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ }
+
FAIL;
}
/* Control never gets here */
@@ -3694,12 +3748,30 @@
{
if (eptr >= md->end_subject) break;
c = *eptr;
- if ((data[c/8] & (1 << (c&7))) != 0) continue;
- if (md->runtime_caseless)
+
+ /* Either not runtime caseless, or it was a positive class. For
+ runtime caseless, continue if either case is in the map. */
+
+ if (!nasty_case)
{
+ if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ if (md->runtime_caseless)
+ {
+ c = pcre_fcc[c];
+ if ((data[c/8] & (1 << (c&7))) != 0) continue;
+ }
+ }
+
+ /* Runtime caseless and it was a negative class. Continue only if
+ both cases are in the map. */
+
+ else
+ {
+ if ((data[c/8] & (1 << (c&7))) == 0) break;
c = pcre_fcc[c];
if ((data[c/8] & (1 << (c&7))) != 0) continue;
}
+
break;
}
@@ -4430,17 +4502,17 @@
/* The "volatile" directives are to make gcc -Wall stop complaining
that these variables can be clobbered by the longjmp. Hopefully
they won't cost too much performance. */
-int resetcount, ocount;
-int first_char = -1;
+volatile int resetcount, ocount;
+volatile int first_char = -1;
match_data match_block;
const uschar *start_bits = NULL;
const uschar *start_match = (const uschar *)subject + start_pos;
const uschar *end_subject;
const real_pcre *re = (const real_pcre *)external_re;
const real_pcre_extra *extra = (const real_pcre_extra *)external_extra;
-BOOL using_temporary_offsets = FALSE;
-BOOL anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
-BOOL startline = (re->options & PCRE_STARTLINE) != 0;
+volatile BOOL using_temporary_offsets = FALSE;
+volatile BOOL anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
+volatile BOOL startline = (re->options & PCRE_STARTLINE) != 0;
if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
@@ -4480,7 +4552,7 @@
if (re->top_backref > 0 && re->top_backref >= ocount/2)
{
ocount = re->top_backref * 2 + 2;
- match_block.offset_vector = (pcre_malloc)(ocount * sizeof(int));
+ match_block.offset_vector = (int *)(pcre_malloc)(ocount * sizeof(int));
if (match_block.offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
using_temporary_offsets = TRUE;
DPRINTF(("Got memory to hold back references\n"));
@@ -4639,10 +4711,10 @@
free_stack(&match_block);
return rc;
} /* End of (if setjmp(match_block.error_env)...) */
+ free_stack(&match_block);
+
/* Return an error code; pcremodule.c will preserve the exception */
if (PyErr_Occurred()) return PCRE_ERROR_NOMEMORY;
-
- free_stack(&match_block);
}
while (!anchored &&
match_block.errorcode == PCRE_ERROR_NOMATCH &&