src/tagmanager/tm_tag.c

/*
*
*   Copyright (c) 2001-2002, Biswapesh Chattopadhyay
*   Copyright 2005 The Geany contributors
*
*   This source code is released for free distribution under the terms of the
*   GNU General Public License.
*
*/

#include <stdlib.h>
#include <string.h>
#include <glib-object.h>

#include "tm_tag.h"
#include "ctags-api.h"


#define TAG_NEW(T)	((T) = g_slice_new0(TMTag))
#define TAG_FREE(T)	g_slice_free(TMTag, (T))


#ifdef DEBUG_TAG_REFS

static GHashTable *alive_tags = NULL;

static void foreach_tags_log(gpointer key, gpointer value, gpointer data)
{
	gsize *ref_count = data;
	const TMTag *tag = value;

	*ref_count += (gsize) tag->refcount;
	g_debug("Leaked TMTag (%d refs): %s", tag->refcount, tag->name);
}

static void log_refs_at_exit(void)
{
	gsize ref_count = 0;

	g_hash_table_foreach(alive_tags, foreach_tags_log, &ref_count);
	g_debug("TMTag references left at exit: %lu", ref_count);
}

static TMTag *log_tag_new(void)
{
	TMTag *tag;

	if (! alive_tags)
	{
		alive_tags = g_hash_table_new(g_direct_hash, g_direct_equal);
		atexit(log_refs_at_exit);
	}
	TAG_NEW(tag);
	g_hash_table_insert(alive_tags, tag, tag);

	return tag;
}

static void log_tag_free(TMTag *tag)
{
	g_return_if_fail(alive_tags != NULL);

	if (! g_hash_table_remove(alive_tags, tag)) {
		g_critical("Freeing invalid TMTag pointer %p", (void *) tag);
	} else {
		TAG_FREE(tag);
	}
}

#undef TAG_NEW
#undef TAG_FREE
#define TAG_NEW(T)	((T) = log_tag_new())
#define TAG_FREE(T)	log_tag_free(T)

#endif /* DEBUG_TAG_REFS */


typedef struct
{
	guint *sort_attrs;
	gboolean partial;
	const GPtrArray *tags_array;
	gboolean first;
} TMSortOptions;

/** Gets the GType for a TMTag.
 *
 * @return TMTag type
 * @since 1.32 (API 233) */
GEANY_API_SYMBOL
GType tm_tag_get_type(void)
{
	static GType gtype = 0;
	if (G_UNLIKELY (gtype == 0))
	{
		gtype = g_boxed_type_register_static("TMTag", (GBoxedCopyFunc)tm_tag_ref,
											 (GBoxedFreeFunc)tm_tag_unref);
	}
	return gtype;
}

/*
 Creates a new tag structure and returns a pointer to it.
 @return the new TMTag structure. This should be free()-ed using tm_tag_free()
*/
TMTag *tm_tag_new(void)
{
	TMTag *tag;

	TAG_NEW(tag);
	tag->refcount = 1;

	return tag;
}

/*
 Destroys a TMTag structure, i.e. frees all elements except the tag itself.
 @param tag The TMTag structure to destroy
 @see tm_tag_free()
*/
static void tm_tag_destroy(TMTag *tag)
{
	g_free(tag->name);
	g_free(tag->arglist);
	g_free(tag->scope);
	g_free(tag->inheritance);
	g_free(tag->var_type);
}


/*
 Drops a reference from a TMTag. If the reference count reaches 0, this function
 destroys all data in the tag and frees the tag structure as well.
 @param tag Pointer to a TMTag structure
*/
void tm_tag_unref(TMTag *tag)
{
	/* be NULL-proof because tm_tag_free() was NULL-proof and we indent to be a
	 * drop-in replacment of it */
	if (NULL != tag && g_atomic_int_dec_and_test(&tag->refcount))
	{
		tm_tag_destroy(tag);
		TAG_FREE(tag);
	}
}

/*
 Adds a reference to a TMTag.
 @param tag Pointer to a TMTag structure
 @return the passed-in TMTag
*/
TMTag *tm_tag_ref(TMTag *tag)
{
	g_atomic_int_inc(&tag->refcount);
	return tag;
}

/*
 Inbuilt tag comparison function.
*/
static gint tm_tag_compare(gconstpointer ptr1, gconstpointer ptr2, gpointer user_data)
{
	unsigned int *sort_attr;
	int returnval = 0;
	TMTag *t1 = *((TMTag **) ptr1);
	TMTag *t2 = *((TMTag **) ptr2);
	TMSortOptions *sort_options = user_data;

	if ((NULL == t1) || (NULL == t2))
	{
		g_warning("Found NULL tag");
		return t2 - t1;
	}
	if (NULL == sort_options->sort_attrs)
	{
		if (sort_options->partial)
			return strncmp(FALLBACK(t1->name, ""), FALLBACK(t2->name, ""), strlen(FALLBACK(t1->name, "")));
		else
			return strcmp(FALLBACK(t1->name, ""), FALLBACK(t2->name, ""));
	}

	for (sort_attr = sort_options->sort_attrs; returnval == 0 && *sort_attr != tm_tag_attr_none_t; ++ sort_attr)
	{
		switch (*sort_attr)
		{
			case tm_tag_attr_name_t:
				if (sort_options->partial)
					returnval = strncmp(FALLBACK(t1->name, ""), FALLBACK(t2->name, ""), strlen(FALLBACK(t1->name, "")));
				else
					returnval = strcmp(FALLBACK(t1->name, ""), FALLBACK(t2->name, ""));
				break;
			case tm_tag_attr_file_t:
				returnval = t1->file - t2->file;
				break;
			case tm_tag_attr_line_t:
				returnval = t1->line - t2->line;
				break;
			case tm_tag_attr_type_t:
				returnval = t1->type - t2->type;
				break;
			case tm_tag_attr_scope_t:
				returnval = strcmp(FALLBACK(t1->scope, ""), FALLBACK(t2->scope, ""));
				break;
			case tm_tag_attr_arglist_t:
				returnval = strcmp(FALLBACK(t1->arglist, ""), FALLBACK(t2->arglist, ""));
				if (returnval != 0)
				{
					int line_diff = (t1->line - t2->line);

					returnval = line_diff ? line_diff : returnval;
				}
				break;
			case tm_tag_attr_vartype_t:
				returnval = strcmp(FALLBACK(t1->var_type, ""), FALLBACK(t2->var_type, ""));
				break;
		}
	}
	return returnval;
}

gboolean tm_tags_equal(const TMTag *a, const TMTag *b)
{
	if (a == b)
		return TRUE;

	return (a->line == b->line &&
			a->file == b->file /* ptr comparison */ &&
			strcmp(FALLBACK(a->name, ""), FALLBACK(b->name, "")) == 0 &&
			a->type == b->type &&
			a->local == b->local &&
			a->pointerOrder == b->pointerOrder &&
			a->access == b->access &&
			a->impl == b->impl &&
			a->lang == b->lang &&
			strcmp(FALLBACK(a->scope, ""), FALLBACK(b->scope, "")) == 0 &&
			strcmp(FALLBACK(a->arglist, ""), FALLBACK(b->arglist, "")) == 0 &&
			strcmp(FALLBACK(a->inheritance, ""), FALLBACK(b->inheritance, "")) == 0 &&
			strcmp(FALLBACK(a->var_type, ""), FALLBACK(b->var_type, "")) == 0);
}

/*
 Removes NULL tag entries from an array of tags. Called after tm_tags_dedup() since
 this function substitutes duplicate entries with NULL
 @param tags_array Array of tags to dedup
*/
void tm_tags_prune(GPtrArray *tags_array)
{
	guint i, count;

	g_return_if_fail(tags_array);

	for (i=0, count = 0; i < tags_array->len; ++i)
	{
		if (NULL != tags_array->pdata[i])
			tags_array->pdata[count++] = tags_array->pdata[i];
	}
	tags_array->len = count;
}

/*
 Deduplicates an array on tags using the inbuilt comparison function based on
 the attributes specified. Called by tm_tags_sort() when dedup is TRUE.
 @param tags_array Array of tags to dedup.
 @param sort_attributes Attributes the array is sorted on. They will be deduped
 on the same criteria.
*/
void tm_tags_dedup(GPtrArray *tags_array, TMTagAttrType *sort_attributes, gboolean unref_duplicates)
{
	TMSortOptions sort_options;
	guint i;

	g_return_if_fail(tags_array);
	if (tags_array->len < 2)
		return;

	sort_options.sort_attrs = sort_attributes;
	sort_options.partial = FALSE;
	for (i = 1; i < tags_array->len; ++i)
	{
		if (0 == tm_tag_compare(&(tags_array->pdata[i - 1]), &(tags_array->pdata[i]), &sort_options))
		{
			if (unref_duplicates)
				tm_tag_unref(tags_array->pdata[i-1]);
			tags_array->pdata[i-1] = NULL;
		}
	}
	tm_tags_prune(tags_array);
}

/*
 Sort an array of tags on the specified attribuites using the inbuilt comparison
 function.
 @param tags_array The array of tags to be sorted
 @param sort_attributes Attributes to be sorted on (int array terminated by 0)
 @param dedup Whether to deduplicate the sorted array
*/
void tm_tags_sort(GPtrArray *tags_array, TMTagAttrType *sort_attributes,
	gboolean dedup, gboolean unref_duplicates)
{
	TMSortOptions sort_options;

	g_return_if_fail(tags_array);

	sort_options.sort_attrs = sort_attributes;
	sort_options.partial = FALSE;
	g_ptr_array_sort_with_data(tags_array, tm_tag_compare, &sort_options);
	if (dedup)
		tm_tags_dedup(tags_array, sort_attributes, unref_duplicates);
}

void tm_tags_remove_file_tags(TMSourceFile *source_file, GPtrArray *tags_array)
{
	guint i;

	/* Now we choose between an algorithm with complexity O(tags_array->len) and
	 * O(source_file->tags_array->len * log(tags_array->len)). The latter algorithm
	 * is better when tags_array contains many times more tags than
	 * source_file->tags_array so instead of trying to find the removed tags
	 * linearly, binary search is used. The constant 20 is more or less random
	 * but seems to work well. It's exact value isn't so critical because it's
	 * the extremes where the difference is the biggest: when
	 * source_file->tags_array->len == tags_array->len (single file open) and
	 * source_file->tags_array->len << tags_array->len (the number of tags
	 * from the file is a small fraction of all tags).
	 */
	if (source_file->tags_array->len != 0 &&
		tags_array->len / source_file->tags_array->len < 20)
	{
		for (i = 0; i < tags_array->len; i++)
		{
			TMTag *tag = tags_array->pdata[i];

			if (tag->file == source_file)
				tags_array->pdata[i] = NULL;
		}
	}
	else
	{
		GPtrArray *to_delete = g_ptr_array_sized_new(source_file->tags_array->len);

		for (i = 0; i < source_file->tags_array->len; i++)
		{
			guint j;
			guint tag_count;
			TMTag **found;
			TMTag *tag = source_file->tags_array->pdata[i];

			found = tm_tags_find(tags_array, tag->name, FALSE, &tag_count);

			for (j = 0; j < tag_count; j++)
			{
				if (*found != NULL && (*found)->file == source_file)
				{
					/* we cannot set the pointer to NULL now because the search wouldn't work */
					g_ptr_array_add(to_delete, found);
					/* no break - if there are multiple tags of the same name, we would
					 * always find the first instance and wouldn't remove others; duplicates
					 * in the to_delete list aren't a problem */
				}
				found++;
			}
		}

		for (i = 0; i < to_delete->len; i++)
		{
			TMTag **tag = to_delete->pdata[i];
			*tag = NULL;
		}
		g_ptr_array_free(to_delete, TRUE);
	}

	tm_tags_prune(tags_array);
}

/* Optimized merge sort for merging sorted values from one array to another
 * where one of the arrays is much smaller than the other.
 * The merge complexity depends mostly on the size of the small array
 * and is almost independent of the size of the big array.
 * In addition, get rid of the duplicates (if both big_array and small_array are duplicate-free). */
static GPtrArray *merge(GPtrArray *big_array, GPtrArray *small_array,
	TMSortOptions *sort_options, gboolean unref_duplicates) {
	guint i1 = 0;  /* index to big_array */
	guint i2 = 0;  /* index to small_array */
	guint initial_step;
	guint step;
	GPtrArray *res_array = g_ptr_array_sized_new(big_array->len + small_array->len);
#ifdef TM_DEBUG
	guint cmpnum = 0;
#endif

	/* swap the arrays if len(small) > len(big) */
	if (small_array->len > big_array->len)
	{
		GPtrArray *tmp = small_array;
		small_array = big_array;
		big_array = tmp;
	}

	/* on average, we are merging a value from small_array every
	 * len(big_array) / len(small_array) values - good approximation for fast jump
	 * step size */
	initial_step = (small_array->len > 0) ? big_array->len / small_array->len : 1;
	initial_step = initial_step > 4 ? initial_step : 1;
	step = initial_step;

	while (i1 < big_array->len && i2 < small_array->len)
	{
		gpointer val1;
		gpointer val2 = small_array->pdata[i2];

		if (step > 4)  /* fast path start */
		{
			guint j1 = (i1 + step < big_array->len) ? i1 + step : big_array->len - 1;

			val1 = big_array->pdata[j1];
#ifdef TM_DEBUG
			cmpnum++;
#endif
			/* if the value in big_array after making the big step is still smaller
			 * than the value in small_array, we can copy all the values inbetween
			 * into the result without making expensive string comparisons */
			if (tm_tag_compare(&val1, &val2, sort_options) < 0)
			{
				while (i1 <= j1)
				{
					val1 = big_array->pdata[i1];
					g_ptr_array_add(res_array, val1);
					i1++;
				}
			}
			else
			{
				/* lower the step and try again */
				step /= 2;
			}
		}  /* fast path end */
		else
		{
			gint cmpval;

#ifdef TM_DEBUG
			cmpnum++;
#endif
			val1 = big_array->pdata[i1];
			cmpval = tm_tag_compare(&val1, &val2, sort_options);
			if (cmpval < 0)
			{
				g_ptr_array_add(res_array, val1);
				i1++;
			}
			else
			{
				g_ptr_array_add(res_array, val2);
				i2++;
				/* value from small_array gets merged - reset the step size */
				step = initial_step;
				if (cmpval == 0)
				{
					i1++;  /* remove the duplicate, keep just the newly merged value */
					if (unref_duplicates)
						tm_tag_unref(val1);
				}
			}
		}
	}

	/* end of one of the arrays reached - copy the rest from the other array */
	while (i1 < big_array->len)
		g_ptr_array_add(res_array, big_array->pdata[i1++]);
	while (i2 < small_array->len)
		g_ptr_array_add(res_array, small_array->pdata[i2++]);

#ifdef TM_DEBUG
	printf("cmpnums: %u\n", cmpnum);
	printf("total tags: %u\n", big_array->len);
	printf("merged tags: %u\n\n", small_array->len);
#endif

	return res_array;
}

GPtrArray *tm_tags_merge(GPtrArray *big_array, GPtrArray *small_array,
	TMTagAttrType *sort_attributes, gboolean unref_duplicates)
{
	GPtrArray *res_array;
	TMSortOptions sort_options;

	sort_options.sort_attrs = sort_attributes;
	sort_options.partial = FALSE;
	res_array = merge(big_array, small_array, &sort_options, unref_duplicates);
	return res_array;
}

/*
 This function will extract the tags of the specified types from an array of tags.
 The returned value is a GPtrArray which should be free-d with a call to
 g_ptr_array_free(array, TRUE). However, do not free the tags themselves since they
 are not duplicated.
 @param tags_array The original array of tags
 @param tag_types - The tag types to extract. Can be a bitmask. For example, passing
 (tm_tag_typedef_t | tm_tag_struct_t) will extract all typedefs and structures from
 the original array.
 @return an array of tags (NULL on failure)
*/
GPtrArray *tm_tags_extract(GPtrArray *tags_array, TMTagType tag_types)
{
	GPtrArray *new_tags;
	guint i;

	g_return_val_if_fail(tags_array, NULL);

	new_tags = g_ptr_array_new();
	for (i=0; i < tags_array->len; ++i)
	{
		if (NULL != tags_array->pdata[i])
		{
			if (tag_types & (((TMTag *) tags_array->pdata[i])->type))
				g_ptr_array_add(new_tags, tags_array->pdata[i]);
		}
	}
	return new_tags;
}

/*
 Completely frees an array of tags.
 @param tags_array Array of tags to be freed.
 @param free_array Whether the GptrArray is to be freed as well.
*/
void tm_tags_array_free(GPtrArray *tags_array, gboolean free_all)
{
	if (tags_array)
	{
		guint i;
		for (i = 0; i < tags_array->len; ++i)
			tm_tag_unref(tags_array->pdata[i]);
		if (free_all)
			g_ptr_array_free(tags_array, TRUE);
		else
			g_ptr_array_set_size(tags_array, 0);
	}
}

/* copy/pasted bsearch() from libc extended with user_data for comparison function
 * and using glib types */
static gpointer binary_search(gpointer key, gpointer base, size_t nmemb,
	GCompareDataFunc compar, gpointer user_data)
{
	gsize l, u, idx;
	gpointer p;
	gint comparison;

	l = 0;
	u = nmemb;
	while (l < u)
	{
		idx = (l + u) / 2;
		p = (gpointer) (((const gchar *) base) + (idx * sizeof(gpointer)));
		comparison = (*compar) (key, p, user_data);
		if (comparison < 0)
			u = idx;
		else if (comparison > 0)
			l = idx + 1;
		else
			return (gpointer) p;
	}

	return NULL;
}

static gint tag_search_cmp(gconstpointer ptr1, gconstpointer ptr2, gpointer user_data)
{
	gint res = tm_tag_compare(ptr1, ptr2, user_data);

	if (res == 0)
	{
		TMSortOptions *sort_options = user_data;
		const GPtrArray *tags_array = sort_options->tags_array;
		TMTag **tag = (TMTag **) ptr2;

		/* if previous/next (depending on sort options) tag equal, we haven't
		 * found the first/last tag in a sequence of equal tags yet */
		if (sort_options->first && ptr2 != &tags_array->pdata[0]) {
			if (tm_tag_compare(ptr1, tag - 1, user_data) == 0)
				return -1;
		}
		else if (!sort_options->first && ptr2 != &tags_array->pdata[tags_array->len-1])
		{
			if (tm_tag_compare(ptr1, tag + 1, user_data) == 0)
				return 1;
		}
	}
	return res;
}

/*
 Returns a pointer to the position of the first matching tag in a (sorted) tags array.
 The passed array of tags must be already sorted by name (searched with binary search).
 @param tags_array Tag array (sorted on name)
 @param name Name of the tag to locate.
 @param partial If TRUE, matches the first part of the name instead of doing exact match.
 @param tagCount Return location of the matched tags.
*/
TMTag **tm_tags_find(const GPtrArray *tags_array, const char *name,
		gboolean partial, guint *tagCount)
{
	TMTag *tag, **first;
	TMSortOptions sort_options;

	*tagCount = 0;
	if (!tags_array || !tags_array->len)
		return NULL;

	tag = g_new0(TMTag, 1);
	tag->name = (char *) name;

	sort_options.sort_attrs = NULL;
	sort_options.partial = partial;
	sort_options.tags_array = tags_array;
	sort_options.first = TRUE;
	first = (TMTag **)binary_search(&tag, tags_array->pdata, tags_array->len,
			tag_search_cmp, &sort_options);

	if (first)
	{
		TMTag **last;
		unsigned first_pos;

		sort_options.first = FALSE;
		first_pos = first - (TMTag **)tags_array->pdata;
		/* search between the first element and end */
		last = (TMTag **)binary_search(&tag, first, tags_array->len - first_pos,
				tag_search_cmp, &sort_options);
		*tagCount = last - first + 1;
	}

	g_free(tag);
	return (TMTag **) first;
}

/* Returns TMTag which "own" given line
 @param line Current line in edited file.
 @param file_tags A GPtrArray of edited file TMTag pointers.
 @param tag_types the tag types to include in the match
 @return TMTag pointers to owner tag. */
const TMTag *
tm_get_current_tag (GPtrArray * file_tags, const gulong line, const TMTagType tag_types)
{
	TMTag *matching_tag = NULL;
	if (file_tags && file_tags->len)
	{
		guint i;
		gulong matching_line = 0;

		for (i = 0; (i < file_tags->len); ++i)
		{
			TMTag *tag = TM_TAG (file_tags->pdata[i]);
			if (tag && tag->type & tag_types &&
				tag->line <= line && tag->line > matching_line)
			{
				matching_tag = tag;
				matching_line = tag->line;
			}
		}
	}
	return matching_tag;
}

gboolean tm_tag_is_anon(const TMTag *tag)
{
	guint i;
	char dummy;

	if (tag->lang == TM_PARSER_C || tag->lang == TM_PARSER_CPP)
		return sscanf(tag->name, "anon_%*[a-z]_%u%c", &i, &dummy) == 1;
	else if (tag->lang == TM_PARSER_FORTRAN || tag->lang == TM_PARSER_F77)
		return sscanf(tag->name, "Structure#%u%c", &i, &dummy) == 1 ||
			sscanf(tag->name, "Interface#%u%c", &i, &dummy) == 1 ||
			sscanf(tag->name, "Enum#%u%c", &i, &dummy) == 1;
	return FALSE;
}


#ifdef TM_DEBUG /* various debugging functions */

/*
 Returns the type of tag as a string
 @param tag The tag whose type is required
*/
const char *tm_tag_type_name(const TMTag *tag)
{
	g_return_val_if_fail(tag, NULL);

	return tm_ctags_get_kind_name(tm_parser_get_tag_kind(tag->type, tag->lang), tag->lang);
}

/*
 Returns the TMTagType given the name of the type. Reverse of tm_tag_type_name.
 @param tag_name Name of the tag type
*/
TMTagType tm_tag_name_type(const char* tag_name, TMParserType lang)
{
	g_return_val_if_fail(tag_name, tm_tag_undef_t);

	return tm_parser_get_tag_type(tm_ctags_get_kind_from_name(tag_name, lang), lang);
}

static const char *tm_tag_impl_name(TMTag *tag)
{
	g_return_val_if_fail(tag, NULL);
	if (TAG_IMPL_VIRTUAL == tag->impl)
		return "virtual";
	else
		return NULL;
}

static const char *tm_tag_access_name(TMTag *tag)
{
	g_return_val_if_fail(tag, NULL);
	if (TAG_ACCESS_PUBLIC == tag->access)
		return "public";
	else if (TAG_ACCESS_PROTECTED == tag->access)
		return "protected";
	else if (TAG_ACCESS_PRIVATE == tag->access)
		return "private";
	else
		return NULL;
}

/*
  Prints information about a tag to the given file pointer.
  @param tag The tag whose info is required.
  @param fp The file pointer of the file to print the info to.
*/
void tm_tag_print(TMTag *tag, FILE *fp)
{
	const char *laccess, *impl, *type;
	if (!tag || !fp)
		return;
	laccess = tm_tag_access_name(tag);
	impl = tm_tag_impl_name(tag);
	type = tm_tag_type_name(tag);
	if (laccess)
		fprintf(fp, "%s ", laccess);
	if (impl)
		fprintf(fp, "%s ", impl);
	if (type)
		fprintf(fp, "%s ", type);
	if (tag->var_type)
		fprintf(fp, "%s ", tag->var_type);
	if (tag->scope)
		fprintf(fp, "%s::", tag->scope);
	fprintf(fp, "%s", tag->name);
	if (tag->arglist)
		fprintf(fp, "%s", tag->arglist);
	if (tag->inheritance)
		fprintf(fp, " : from %s", tag->inheritance);
	if ((tag->file) && (tag->line > 0))
		fprintf(fp, "[%s:%ld]", tag->file->file_name
		  , tag->line);
	fprintf(fp, "\n");
}

/*
  Prints info about all tags in the array to the given file pointer.
*/
void tm_tags_array_print(GPtrArray *tags, FILE *fp)
{
	guint i;
	TMTag *tag;
	if (!(tags && (tags->len > 0) && fp))
		return;
	for (i = 0; i < tags->len; ++i)
	{
		tag = TM_TAG(tags->pdata[i]);
		tm_tag_print(tag, fp);
	}
}

/*
  Returns the depth of tag scope (useful for finding tag hierarchy
*/
gint tm_tag_scope_depth(const TMTag *t)
{
	const gchar *context_sep = tm_parser_context_separator(t->lang);
	gint depth;
	char *s;
	if(!(t && t->scope))
		return 0;
	for (s = t->scope, depth = 0; s; s = strstr(s, context_sep))
	{
		++ depth;
		++ s;
	}
	return depth;
}

#endif /* TM_DEBUG */