lregex.c

/*
*   $Id$
*
*   Copyright (c) 2000-2003, Darren Hiebert
*
*   This source code is released for free distribution under the terms of the
*   GNU General Public License.
*
*   This module contains functions for applying regular expression matching.
*
*   The code for utlizing the Gnu regex package with regards to processing the
*   regex option and checking for regex matches was adapted from routines in
*   Gnu etags.
*/

/*
*   INCLUDE FILES
*/
#include "general.h"  /* must always come first */

#include <string.h>

#ifdef HAVE_REGCOMP
# include <ctype.h>
# include <stddef.h>
# ifdef HAVE_SYS_TYPES_H
#  include <sys/types.h>  /* declare off_t (not known to regex.h on FreeBSD) */
# endif
# include <regex.h>
#endif

#include "debug.h"
#include "entry.h"
#include "flags.h"
#include "parse.h"
#include "read.h"
#include "routines.h"


static boolean regexAvailable = FALSE;

#if defined (HAVE_REGEX) && !defined (REGCOMP_BROKEN)

/*
*   MACROS
*/

/* Back-references \0 through \9 */
#define BACK_REFERENCE_COUNT 10


#define REGEX_NAME "Regex"

/*
*   DATA DECLARATIONS
*/

struct sKind {
	boolean enabled;
	char letter;
	char* name;
	char* description;
};

enum pType { PTRN_TAG, PTRN_CALLBACK };

typedef struct {
	regex_t *pattern;
	enum pType type;
	boolean exclusive;
	union {
		struct {
			char *name_pattern;
			struct sKind kind;
		} tag;
		struct {
			regexCallback function;
		} callback;
	} u;
} regexPattern;


typedef struct {
	regexPattern *patterns;
	unsigned int count;
} patternSet;

#define COUNT(D) (sizeof(D)/sizeof(D[0]))

/*
*   DATA DEFINITIONS
*/

/* Array of pattern sets, indexed by language */
static patternSet* Sets = NULL;
static int SetUpper = -1;  /* upper language index in list */

/*
*   FUNCTION DEFINITIONS
*/


static void clearPatternSet (const langType language)
{
	if (language <= SetUpper)
	{
		patternSet* const set = Sets + language;
		unsigned int i;
		for (i = 0  ;  i < set->count  ;  ++i)
		{
			regexPattern *p = &set->patterns [i];
			regfree (p->pattern);
			eFree (p->pattern);
			p->pattern = NULL;

			if (p->type == PTRN_TAG)
			{
				eFree (p->u.tag.name_pattern);
				p->u.tag.name_pattern = NULL;
				eFree (p->u.tag.kind.name);
				p->u.tag.kind.name = NULL;
				if (p->u.tag.kind.description != NULL)
				{
					eFree (p->u.tag.kind.description);
					p->u.tag.kind.description = NULL;
				}
			}
		}
		if (set->patterns != NULL)
			eFree (set->patterns);
		set->patterns = NULL;
		set->count = 0;
	}
}

/*
*   Regex psuedo-parser
*/

static void makeRegexTag (
		const vString* const name, const struct sKind* const kind)
{
	if (kind->enabled)
	{
		tagEntryInfo e;
		Assert (name != NULL  &&  vStringLength (name) > 0);
		Assert (kind != NULL);
		initTagEntry (&e, vStringValue (name));
		e.kind     = kind->letter;
		e.kindName = kind->name;
		makeTagEntry (&e);
	}
}

/*
*   Regex pattern definition
*/

/* Take a string like "/blah/" and turn it into "blah", making sure
 * that the first and last characters are the same, and handling
 * quoted separator characters.  Actually, stops on the occurrence of
 * an unquoted separator.  Also turns "\t" into a Tab character.
 * Returns pointer to terminating separator.  Works in place.  Null
 * terminates name string.
 */
static char* scanSeparators (char* name)
{
	char sep = name [0];
	char *copyto = name;
	boolean quoted = FALSE;

	for (++name ; *name != '\0' ; ++name)
	{
		if (quoted)
		{
			if (*name == sep)
				*copyto++ = sep;
			else if (*name == 't')
				*copyto++ = '\t';
			else
			{
				/* Something else is quoted, so preserve the quote. */
				*copyto++ = '\\';
				*copyto++ = *name;
			}
			quoted = FALSE;
		}
		else if (*name == '\\')
			quoted = TRUE;
		else if (*name == sep)
		{
			break;
		}
		else
			*copyto++ = *name;
	}
	*copyto = '\0';
	return name;
}

/* Parse `regexp', in form "/regex/name/[k,Kind/]flags" (where the separator
 * character is whatever the first character of `regexp' is), by breaking it
 * up into null terminated strings, removing the separators, and expanding
 * '\t' into tabs. When complete, `regexp' points to the line matching
 * pattern, a pointer to the name matching pattern is written to `name', a
 * pointer to the kinds is written to `kinds' (possibly NULL), and a pointer
 * to the trailing flags is written to `flags'. If the pattern is not in the
 * correct format, a false value is returned.
 */
static boolean parseTagRegex (
		char* const regexp, char** const name,
		char** const kinds, char** const flags)
{
	boolean result = FALSE;
	const int separator = (unsigned char) regexp [0];

	*name = scanSeparators (regexp);
	if (*regexp == '\0')
		error (WARNING, "empty regexp");
	else if (**name != separator)
		error (WARNING, "%s: incomplete regexp", regexp);
	else
	{
		char* const third = scanSeparators (*name);
		if (**name == '\0')
			error (WARNING, "%s: regexp missing name pattern", regexp);
		if ((*name) [strlen (*name) - 1] == '\\')
			error (WARNING, "error in name pattern: \"%s\"", *name);
		if (*third != separator)
			error (WARNING, "%s: regexp missing final separator", regexp);
		else
		{
			char* const fourth = scanSeparators (third);
			if (*fourth == separator)
			{
				*kinds = third;
				scanSeparators (fourth);
				*flags = fourth;
			}
			else
			{
				*flags = third;
				*kinds = NULL;
			}
			result = TRUE;
		}
	}
	return result;
}


static void ptrn_flag_exclusive_short (char c __unused__, void* data)
{
	regexPattern *ptrn = data;
	ptrn->exclusive = TRUE;
}

static void ptrn_flag_exclusive_long (const char* const s __unused__, const char* const unused __unused__, void* data)
{
	ptrn_flag_exclusive_short ('x', data);
}

static flagDefinition ptrnFlagDef[] = {
	{ 'x', "exclusive", ptrn_flag_exclusive_short, ptrn_flag_exclusive_long },
};

static regexPattern* addCompiledTagCommon (const langType language,
					 regex_t* const pattern)
{
	patternSet* set;
	regexPattern *ptrn;
	if (language > SetUpper)
	{
		int i;
		Sets = xRealloc (Sets, (language + 1), patternSet);
		for (i = SetUpper + 1  ;  i <= language  ;  ++i)
		{
			Sets [i].patterns = NULL;
			Sets [i].count = 0;
		}
		SetUpper = language;
	}
	set = Sets + language;
	set->patterns = xRealloc (set->patterns, (set->count + 1), regexPattern);

	ptrn = &set->patterns [set->count];
	ptrn->pattern = pattern;
	ptrn->exclusive = FALSE;

	set->count += 1;
	useRegexMethod(language);
	return ptrn;
}

static void addCompiledTagPattern (
		const langType language, regex_t* const pattern,
		char* const name, const char kind, char* const kindName,
		char *const description, const char* flags)
{
	regexPattern * ptrn;

	ptrn  = addCompiledTagCommon(language, pattern);
	ptrn->type    = PTRN_TAG;
	ptrn->u.tag.name_pattern = name;
	ptrn->u.tag.kind.enabled = TRUE;
	ptrn->u.tag.kind.letter  = kind;
	ptrn->u.tag.kind.name    = kindName;
	ptrn->u.tag.kind.description = description;
	flagsEval (flags, ptrnFlagDef, COUNT(ptrnFlagDef), ptrn);
}

static void addCompiledCallbackPattern (
		const langType language, regex_t* const pattern,
		const regexCallback callback, const char* flags)
{
	regexPattern * ptrn;

	ptrn  = addCompiledTagCommon(language, pattern);
	ptrn->type    = PTRN_CALLBACK;
	ptrn->u.callback.function = callback;
	flagsEval (flags, ptrnFlagDef, COUNT(ptrnFlagDef), ptrn);
}


static void regex_flag_basic_short (char c __unused__, void* data)
{
	int* cflags = data;
	*cflags &= ~REG_EXTENDED;
}
static void regex_flag_basic_long (const char* const s __unused__, const char* const unused __unused__, void* data)
{
	regex_flag_basic_short ('b', data);
}

static void regex_flag_extend_short (char c __unused__, void* data)
{
	int* cflags = data;
	*cflags |= REG_EXTENDED;
}

static void regex_flag_extend_long (const char* const c __unused__, const char* const unused __unused__, void* data)
{
	regex_flag_extend_short('e', data);
}

static void regex_flag_icase_short (char c __unused__, void* data)
{
	int* cflags = data;
	*cflags |= REG_ICASE;
}

static void regex_flag_icase_long (const char* s __unused__, const char* const unused __unused__, void* data)
{
	regex_flag_icase_short ('i', data);
}


static regex_t* compileRegex (const char* const regexp, const char* const flags)
{
	int cflags = REG_EXTENDED | REG_NEWLINE;
	regex_t *result = NULL;
	int errcode;

	flagDefinition regexFlagDefs[] = {
		{ 'b', "basic",  regex_flag_basic_short,  regex_flag_basic_long  },
		{ 'e', "extend", regex_flag_extend_short, regex_flag_extend_long },
		{ 'i', "icase",  regex_flag_icase_short,  regex_flag_icase_long  },
	};
	flagsEval (flags,
		   regexFlagDefs,
		   COUNT(regexFlagDefs),
		   &cflags);

	result = xMalloc (1, regex_t);
	errcode = regcomp (result, regexp, cflags);
	if (errcode != 0)
	{
		char errmsg[256];
		regerror (errcode, result, errmsg, 256);
		error (WARNING, "regcomp %s: %s", regexp, errmsg);
		regfree (result);
		eFree (result);
		result = NULL;
	}
	return result;
}


static void parseKinds (
		const char* const kinds, char* const kind, char** const kindName,
		char **description)
{
	*kind = '\0';
	*kindName = NULL;
	*description = NULL;
	if (kinds == NULL  ||  kinds [0] == '\0')
	{
		*kind = 'r';
		*kindName = eStrdup ("regex");
	}
	else if (kinds [0] != '\0')
	{
		const char* k = kinds;
		if (k [0] != ','  &&  (k [1] == ','  ||  k [1] == '\0'))
			*kind = *k++;
		else
			*kind = 'r';
		if (*k == ',')
			++k;
		if (k [0] == '\0')
			*kindName = eStrdup ("regex");
		else
		{
			const char *const comma = strchr (k, ',');
			if (comma == NULL)
				*kindName = eStrdup (k);
			else
			{
				*kindName = (char*) eMalloc (comma - k + 1);
				strncpy (*kindName, k, comma - k);
				(*kindName) [comma - k] = '\0';
				k = comma + 1;
				if (k [0] != '\0')
					*description = eStrdup (k);
			}
		}
	}
}

static void printRegexKind (const regexPattern *pat, unsigned int i, boolean indent)
{
	const struct sKind *const kind = &pat [i].u.tag.kind;
	const char *const indentation = indent ? "    " : "";
	Assert (pat [i].type == PTRN_TAG);
	printf ("%s%c  %s %s\n", indentation,
			kind->letter != '\0' ? kind->letter : '?',
			kind->description != NULL ? kind->description : kind->name,
			kind->enabled ? "" : " [off]");
}

static void processLanguageRegex (const langType language,
		const char* const parameter)
{
	if (parameter == NULL  ||  parameter [0] == '\0')
		clearPatternSet (language);
	else if (parameter [0] != '@')
		addLanguageRegex (language, parameter);
	else if (! doesFileExist (parameter + 1))
		error (WARNING, "cannot open regex file");
	else
	{
		const char* regexfile = parameter + 1;
		FILE* const fp = fopen (regexfile, "r");
		if (fp == NULL)
			error (WARNING | PERROR, "%s", regexfile);
		else
		{
			vString* const regex = vStringNew ();
			while (readLine (regex, fp))
				addLanguageRegex (language, vStringValue (regex));
			fclose (fp);
			vStringDelete (regex);
		}
	}
}

/*
*   Regex pattern matching
*/


static vString* substitute (
		const char* const in, const char* out,
		const int nmatch, const regmatch_t* const pmatch)
{
	vString* result = vStringNew ();
	const char* p;
	for (p = out  ;  *p != '\0'  ;  p++)
	{
		if (*p == '\\'  &&  isdigit ((int) *++p))
		{
			const int dig = *p - '0';
			if (0 < dig  &&  dig < nmatch  &&  pmatch [dig].rm_so != -1)
			{
				const int diglen = pmatch [dig].rm_eo - pmatch [dig].rm_so;
				vStringNCatS (result, in + pmatch [dig].rm_so, diglen);
			}
		}
		else if (*p != '\n'  &&  *p != '\r')
			vStringPut (result, *p);
	}
	vStringTerminate (result);
	return result;
}

static void matchTagPattern (const vString* const line,
		const regexPattern* const patbuf,
		const regmatch_t* const pmatch)
{
	vString *const name = substitute (vStringValue (line),
			patbuf->u.tag.name_pattern, BACK_REFERENCE_COUNT, pmatch);
	vStringStripLeading (name);
	vStringStripTrailing (name);
	if (vStringLength (name) > 0)
		makeRegexTag (name, &patbuf->u.tag.kind);
	else
		error (WARNING, "%s:%ld: null expansion of name pattern \"%s\"",
			getInputFileName (), getInputLineNumber (),
			patbuf->u.tag.name_pattern);
	vStringDelete (name);
}

static void matchCallbackPattern (
		const vString* const line, const regexPattern* const patbuf,
		const regmatch_t* const pmatch)
{
	regexMatch matches [BACK_REFERENCE_COUNT];
	unsigned int count = 0;
	int i;
	for (i = 0  ;  i < BACK_REFERENCE_COUNT  ;  ++i)
	{
		matches [i].start  = pmatch [i].rm_so;
		matches [i].length = pmatch [i].rm_eo - pmatch [i].rm_so;
		/* a valid match may have both offsets == -1,
		 * e.g. (foo)*(bar) matching "bar" - see CTags bug 271.
		 * As POSIX regex doesn't seem to have a way to count matches,
		 * we return the count up to the last non-empty match. */
		if (pmatch [i].rm_so != -1)
			count = i + 1;
	}
	patbuf->u.callback.function (vStringValue (line), matches, count);
}

static boolean matchRegexPattern (const vString* const line,
		const regexPattern* const patbuf)
{
	boolean result = FALSE;
	regmatch_t pmatch [BACK_REFERENCE_COUNT];
	const int match = regexec (patbuf->pattern, vStringValue (line),
							   BACK_REFERENCE_COUNT, pmatch, 0);
	if (match == 0)
	{
		result = TRUE;
		if (patbuf->type == PTRN_TAG)
			matchTagPattern (line, patbuf, pmatch);
		else if (patbuf->type == PTRN_CALLBACK)
			matchCallbackPattern (line, patbuf, pmatch);
		else
		{
			Assert ("invalid pattern type" == NULL);
			result = FALSE;
		}
	}
	return result;
}


/* PUBLIC INTERFACE */

/* Match against all patterns for specified language. Returns true if at least
 * on pattern matched.
 */
extern boolean matchRegex (const vString* const line, const langType language)
{
	boolean result = FALSE;
	if (language != LANG_IGNORE  &&  language <= SetUpper  &&
		Sets [language].count > 0)
	{
		const patternSet* const set = Sets + language;
		unsigned int i;
		for (i = 0  ;  i < set->count  ;  ++i)
		{
			regexPattern* ptrn = set->patterns + i;
			if (matchRegexPattern (line, ptrn))
			{
				result = TRUE;
				if (ptrn->exclusive)
					break;
			}
		}
	}
	return result;
}

extern void findRegexTags (void)
{
	/* merely read all lines of the file */
	while (fileReadLine () != NULL)
		;
}

#endif  /* HAVE_REGEX */

extern void addTagRegex (
		const langType language __unused__,
		const char* const regex __unused__,
		const char* const name __unused__,
		const char* const kinds __unused__,
		const char* const flags __unused__)
{
#ifdef HAVE_REGEX
	Assert (regex != NULL);
	Assert (name != NULL);
	if (regexAvailable)
	{
		regex_t* const cp = compileRegex (regex, flags);
		if (cp != NULL)
		{
			char kind;
			char* kindName;
			char* description;
			parseKinds (kinds, &kind, &kindName, &description);
			addCompiledTagPattern (language, cp, eStrdup (name),
					kind, kindName, description, flags);
		}
	}
#endif
}

extern void addCallbackRegex (
		const langType language __unused__,
		const char* const regex __unused__,
		const char* const flags __unused__,
		const regexCallback callback __unused__)
{
#ifdef HAVE_REGEX
	Assert (regex != NULL);
	if (regexAvailable)
	{
		regex_t* const cp = compileRegex (regex, flags);
		if (cp != NULL)
			addCompiledCallbackPattern (language, cp, callback, flags);
	}
#endif
}

extern void addLanguageRegex (
		const langType language __unused__, const char* const regex __unused__)
{
#ifdef HAVE_REGEX
	if (regexAvailable)
	{
		char *const regex_pat = eStrdup (regex);
		char *name, *kinds, *flags;
		if (parseTagRegex (regex_pat, &name, &kinds, &flags))
		{
			addTagRegex (language, regex_pat, name, kinds, flags);
			eFree (regex_pat);
		}
	}
#endif
}

/*
*   Regex option parsing
*/

extern boolean processRegexOption (const char *const option,
								   const char *const parameter __unused__)
{
	const char* const dash = strchr (option, '-');
	char* lang;
	langType language;

	if (dash == NULL)
		return FALSE;

	if (strncmp (option, "regex", dash - option) == 0)
		lang = eStrdup (dash + 1);
	else if (strcmp (dash + 1, "regex") == 0)
		lang = eStrndup (option, dash - option);
	else
		return FALSE;

#ifdef HAVE_REGEX
	language = getNamedLanguage (lang);
	if (language == LANG_IGNORE)
		error (WARNING, "unknown language \"%s\" in --%s option", lang, option);
	else
		processLanguageRegex (language, parameter);
#else
	error (WARNING, "regex support not available; required for --%s option",
	       option);
#endif

	eFree (lang);

	return TRUE;
}

extern void resetRegexKinds (const langType language __unused__,
			     boolean mode)
{
#ifdef HAVE_REGEX
	if (language <= SetUpper  &&  Sets [language].count > 0)
	{
		patternSet* const set = Sets + language;
		unsigned int i;
		for (i = 0  ;  i < set->count  ;  ++i)
			if (set->patterns [i].type == PTRN_TAG)
				set->patterns [i].u.tag.kind.enabled = mode;
	}
#endif
}

extern boolean enableRegexKind (
		const langType language __unused__,
		const int kind __unused__, const boolean mode __unused__)
{
	boolean result = FALSE;
#ifdef HAVE_REGEX
	if (language <= SetUpper  &&  Sets [language].count > 0)
	{
		patternSet* const set = Sets + language;
		unsigned int i;
		for (i = 0  ;  i < set->count  ;  ++i)
			if (set->patterns [i].type == PTRN_TAG &&
				set->patterns [i].u.tag.kind.letter == kind)
			{
				set->patterns [i].u.tag.kind.enabled = mode;
				result = TRUE;
			}
	}
#endif
	return result;
}

extern void printRegexKinds (const langType language __unused__, boolean indent __unused__)
{
#ifdef HAVE_REGEX
	if (language <= SetUpper  &&  Sets [language].count > 0)
	{
		patternSet* const set = Sets + language;
		unsigned int i;
		for (i = 0  ;  i < set->count  ;  ++i)
			if (set->patterns [i].type == PTRN_TAG)
				printRegexKind (set->patterns, i, indent);
	}
#endif
}

extern void freeRegexResources (void)
{
#ifdef HAVE_REGEX
	int i;
	for (i = 0  ;  i <= SetUpper  ;  ++i)
		clearPatternSet (i);
	if (Sets != NULL)
		eFree (Sets);
	Sets = NULL;
	SetUpper = -1;
#endif
}

/* Return TRUE if available. */
extern boolean checkRegex (void)
{
#if ! defined (HAVE_REGEX)
	regexAvailable = FALSE;
#elif defined (CHECK_REGCOMP)
	{
		/* Check for broken regcomp() on Cygwin */
		regex_t patbuf;
		int errcode;
		if (regcomp (&patbuf, "/hello/", 0) != 0)
			error (WARNING, "Disabling broken regex");
		else
			regexAvailable = TRUE;
	}
#else
	/* We are using bundled regex engine. */
	regexAvailable = TRUE;
#endif
	return regexAvailable;
}

/* vi:set tabstop=4 shiftwidth=4: */