Merge remote-tracking branch 'gvdb/master' into update-gvdb
Minor merge conflicts in gvdb-reader.c with the documentation.
diff --git a/gio/gvdb/gvdb-builder.c b/gio/gvdb/gvdb-builder.c
index 8b9baa0..2383e60 100644
--- a/gio/gvdb/gvdb-builder.c
+++ b/gio/gvdb/gvdb-builder.c
@@ -339,6 +339,13 @@
#undef chunk
memset (*bloom_filter, 0, n_bloom_words * sizeof (guint32_le));
+
+ /* NOTE - the code to actually fill in the bloom filter here is missing.
+ * Patches welcome!
+ *
+ * http://en.wikipedia.org/wiki/Bloom_filter
+ * http://0pointer.de/blog/projects/bloom.html
+ */
}
static void
diff --git a/gio/gvdb/gvdb-reader.c b/gio/gvdb/gvdb-reader.c
index 510eba2..aa3154f 100644
--- a/gio/gvdb/gvdb-reader.c
+++ b/gio/gvdb/gvdb-reader.c
@@ -23,15 +23,11 @@
#include <string.h>
struct _GvdbTable {
- gint ref_count;
+ GBytes *bytes;
const gchar *data;
gsize size;
- gpointer user_data;
- GvdbRefFunc ref_user_data;
- GDestroyNotify unref_user_data;
-
gboolean byteswapped;
gboolean trusted;
@@ -124,80 +120,81 @@
file->n_hash_items = size / sizeof (struct gvdb_hash_item);
}
-static GvdbTable *
-new_from_data (const void *data,
- gsize data_len,
- gboolean trusted,
- gpointer user_data,
- GvdbRefFunc ref,
- GDestroyNotify unref,
- const char *filename,
- GError **error)
+/**
+ * gvdb_table_new_from_bytes:
+ * @bytes: the #GBytes with the data
+ * @trusted: if the contents of @bytes are trusted
+ * @error: %NULL, or a pointer to a %NULL #GError
+ *
+ * Creates a new #GvdbTable from the contents of @bytes.
+ *
+ * This call can fail if the header contained in @bytes is invalid or if @bytes
+ * is empty; if so, %G_FILE_ERROR_INVAL will be returned.
+ *
+ * You should call gvdb_table_free() on the return result when you no
+ * longer require it.
+ *
+ * Returns: a new #GvdbTable
+ **/
+GvdbTable *
+gvdb_table_new_from_bytes (GBytes *bytes,
+ gboolean trusted,
+ GError **error)
{
+ const struct gvdb_header *header;
GvdbTable *file;
file = g_slice_new0 (GvdbTable);
- file->data = data;
- file->size = data_len;
+ file->bytes = g_bytes_ref (bytes);
+ file->data = g_bytes_get_data (bytes, &file->size);
file->trusted = trusted;
- file->ref_count = 1;
- file->ref_user_data = ref;
- file->unref_user_data = unref;
- file->user_data = user_data;
- if (sizeof (struct gvdb_header) <= file->size)
- {
- const struct gvdb_header *header = (gpointer) file->data;
+ if (file->size < sizeof (struct gvdb_header))
+ goto invalid;
- if (header->signature[0] == GVDB_SIGNATURE0 &&
- header->signature[1] == GVDB_SIGNATURE1 &&
- guint32_from_le (header->version) == 0)
- file->byteswapped = FALSE;
+ header = (gpointer) file->data;
- else if (header->signature[0] == GVDB_SWAPPED_SIGNATURE0 &&
- header->signature[1] == GVDB_SWAPPED_SIGNATURE1 &&
- guint32_from_le (header->version) == 0)
- file->byteswapped = TRUE;
+ if (header->signature[0] == GVDB_SIGNATURE0 &&
+ header->signature[1] == GVDB_SIGNATURE1 &&
+ guint32_from_le (header->version) == 0)
+ file->byteswapped = FALSE;
- else
- {
- if (filename)
- g_set_error (error, G_FILE_ERROR, G_FILE_ERROR_INVAL,
- "%s: invalid header", filename);
- else
- g_set_error (error, G_FILE_ERROR, G_FILE_ERROR_INVAL,
- "invalid gvdb header");
- g_slice_free (GvdbTable, file);
- if (unref)
- unref (user_data);
+ else if (header->signature[0] == GVDB_SWAPPED_SIGNATURE0 &&
+ header->signature[1] == GVDB_SWAPPED_SIGNATURE1 &&
+ guint32_from_le (header->version) == 0)
+ file->byteswapped = TRUE;
- return NULL;
- }
+ else
+ goto invalid;
- gvdb_table_setup_root (file, &header->root);
- }
+ gvdb_table_setup_root (file, &header->root);
return file;
+
+invalid:
+ g_set_error_literal (error, G_FILE_ERROR, G_FILE_ERROR_INVAL, "invalid gvdb header");
+
+ g_bytes_unref (file->bytes);
+
+ g_slice_free (GvdbTable, file);
+
+ return NULL;
}
/**
* gvdb_table_new:
- * @filename: the path to the hash file
- * @trusted: if the contents of @filename are trusted
+ * @filename: a filename
+ * @trusted: if the contents of @bytes are trusted
* @error: %NULL, or a pointer to a %NULL #GError
*
- * Creates a new #GvdbTable from the contents of the file found at
- * @filename.
+ * Creates a new #GvdbTable using the #GMappedFile for @filename as the
+ * #GBytes.
*
- * The only time this function fails is if the file cannot be opened.
+ * This function will fail if the file cannot be opened.
* In that case, the #GError that is returned will be an error from
* g_mapped_file_new().
*
- * An empty or otherwise corrupted file is considered to be a valid
- * #GvdbTable with no entries.
- *
- * You should call gvdb_table_unref() on the return result when you no
- * longer require it.
+ * An empty or corrupt file will result in %G_FILE_ERROR_INVAL.
*
* Returns: a new #GvdbTable
**/
@@ -207,53 +204,21 @@
GError **error)
{
GMappedFile *mapped;
+ GvdbTable *table;
+ GBytes *bytes;
- if ((mapped = g_mapped_file_new (filename, FALSE, error)) == NULL)
+ mapped = g_mapped_file_new (filename, FALSE, error);
+ if (!mapped)
return NULL;
- return new_from_data (g_mapped_file_get_contents (mapped),
- g_mapped_file_get_length (mapped),
- trusted,
- mapped,
- (GvdbRefFunc)g_mapped_file_ref,
- (GDestroyNotify)g_mapped_file_unref,
- filename,
- error);
-}
+ bytes = g_mapped_file_get_bytes (mapped);
+ table = gvdb_table_new_from_bytes (bytes, trusted, error);
+ g_mapped_file_unref (mapped);
+ g_bytes_unref (bytes);
-/**
- * gvdb_table_new_from_data:
- * @data: the data
- * @data_len: the length of @data in bytes
- * @trusted: if the contents of @data are trusted
- * @user_data: User supplied data that owns @data
- * @ref: Ref function for @user_data
- * @unref: Unref function for @user_data
- *
- * Creates a new #GvdbTable from the data in @data.
- *
- * An empty or otherwise corrupted data is considered to be a valid
- * #GvdbTable with no entries.
- *
- * You should call gvdb_table_unref() on the return result when you no
- * longer require it.
- *
- * Returns: a new #GvdbTable
- **/
-GvdbTable *
-gvdb_table_new_from_data (const void *data,
- gsize data_len,
- gboolean trusted,
- gpointer user_data,
- GvdbRefFunc ref,
- GDestroyNotify unref,
- GError **error)
-{
- return new_from_data (data, data_len,
- trusted,
- user_data, ref, unref,
- NULL,
- error);
+ g_prefix_error (error, "%s: ", filename);
+
+ return table;
}
static gboolean
@@ -346,18 +311,6 @@
return NULL;
}
-static const struct gvdb_hash_item *
-gvdb_table_get_item (GvdbTable *table,
- guint32_le item_no)
-{
- guint32 item_no_native = guint32_from_le (item_no);
-
- if G_LIKELY (item_no_native < table->n_hash_items)
- return table->hash_items + item_no_native;
-
- return NULL;
-}
-
static gboolean
gvdb_table_list_from_item (GvdbTable *table,
const struct gvdb_hash_item *item,
@@ -376,6 +329,155 @@
return TRUE;
}
+/**
+ * gvdb_table_get_names:
+ * @table: a #GvdbTable
+ * @length: the number of items returned, or %NULL
+ *
+ * Gets a list of all names contained in @table.
+ *
+ * No call to gvdb_table_get_table(), gvdb_table_list() or
+ * gvdb_table_get_value() will succeed unless it is for one of the
+ * names returned by this function.
+ *
+ * Note that some names that are returned may still fail for all of the
+ * above calls in the case of the corrupted file. Note also that the
+ * returned strings may not be utf8.
+ *
+ * Returns: a %NULL-terminated list of strings, of length @length
+ **/
+gchar **
+gvdb_table_get_names (GvdbTable *table,
+ gint *length)
+{
+ gchar **names;
+ gint n_names;
+ gint filled;
+ gint total;
+ gint i;
+
+ /* We generally proceed by iterating over the list of items in the
+ * hash table (in order of appearance) recording them into an array.
+ *
+ * Each item has a parent item (except root items). The parent item
+ * forms part of the name of the item. We could go fetching the
+ * parent item chain at the point that we encounter each item but then
+ * we would need to implement some sort of recursion along with checks
+ * for self-referential items.
+ *
+ * Instead, we do a number of passes. Each pass will build up one
+ * level of names (starting from the root). We continue to do passes
+ * until no more items are left. The first pass will only add root
+ * items and each further pass will only add items whose direct parent
+ * is an item added in the immediately previous pass. It's also
+ * possible that items get filled if they follow their parent within a
+ * particular pass.
+ *
+ * At most we will have a number of passes equal to the depth of the
+ * tree. Self-referential items will never be filled in (since their
+ * parent will have never been filled in). We continue until we have
+ * a pass that fills in no additional items.
+ *
+ * This takes an O(n) algorithm and turns it into O(n*m) where m is
+ * the depth of the tree, but in all sane cases the tree won't be very
+ * deep and the constant factor of this algorithm is lower (and the
+ * complexity of coding it, as well).
+ */
+
+ n_names = table->n_hash_items;
+ names = g_new0 (gchar *, n_names + 1);
+
+ /* 'names' starts out all-NULL. On each pass we record the number
+ * of items changed from NULL to non-NULL in 'filled' so we know if we
+ * should repeat the loop. 'total' counts the total number of items
+ * filled. If 'total' ends up equal to 'n_names' then we know that
+ * 'names' has been completely filled.
+ */
+
+ total = 0;
+ do
+ {
+ /* Loop until we have filled no more entries */
+ filled = 0;
+
+ for (i = 0; i < n_names; i++)
+ {
+ const struct gvdb_hash_item *item = &table->hash_items[i];
+ const gchar *name;
+ gsize name_length;
+ guint32 parent;
+
+ /* already got it on a previous pass */
+ if (names[i] != NULL)
+ continue;
+
+ parent = guint32_from_le (item->parent);
+
+ if (parent == 0xffffffffu)
+ {
+ /* it's a root item */
+ name = gvdb_table_item_get_key (table, item, &name_length);
+
+ if (name != NULL)
+ {
+ names[i] = g_strndup (name, name_length);
+ filled++;
+ }
+ }
+
+ else if (parent < n_names && names[parent] != NULL)
+ {
+ /* It's a non-root item whose parent was filled in already.
+ *
+ * Calculate the name of this item by combining it with
+ * its parent name.
+ */
+ name = gvdb_table_item_get_key (table, item, &name_length);
+
+ if (name != NULL)
+ {
+ const gchar *parent_name = names[parent];
+ gsize parent_length;
+ gchar *fullname;
+
+ parent_length = strlen (parent_name);
+ fullname = g_malloc (parent_length + name_length + 1);
+ memcpy (fullname, parent_name, parent_length);
+ memcpy (fullname + parent_length, name, name_length);
+ fullname[parent_length + name_length] = '\0';
+ names[i] = fullname;
+ filled++;
+ }
+ }
+ }
+
+ total += filled;
+ }
+ while (filled && total < n_names);
+
+ /* If the table was corrupted then 'names' may have holes in it.
+ * Collapse those.
+ */
+ if G_UNLIKELY (total != n_names)
+ {
+ GPtrArray *fixed_names;
+
+ fixed_names = g_ptr_array_new ();
+ for (i = 0; i < n_names; i++)
+ if (names[i] != NULL)
+ g_ptr_array_add (fixed_names, names[i]);
+
+ g_free (names);
+ n_names = fixed_names->len;
+ g_ptr_array_add (fixed_names, NULL);
+ names = (gchar **) g_ptr_array_free (fixed_names, FALSE);
+ }
+
+ if (length)
+ *length = n_names;
+
+ return names;
+}
/**
* gvdb_table_list:
@@ -457,7 +559,15 @@
gvdb_table_has_value (GvdbTable *file,
const gchar *key)
{
- return gvdb_table_lookup (file, key, 'v') != NULL;
+ static const struct gvdb_hash_item *item;
+ gsize size;
+
+ item = gvdb_table_lookup (file, key, 'v');
+
+ if (item == NULL)
+ return FALSE;
+
+ return gvdb_table_dereference (file, &item->value.pointer, 8, &size) != NULL;
}
static GVariant *
@@ -466,6 +576,7 @@
{
GVariant *variant, *value;
gconstpointer data;
+ GBytes *bytes;
gsize size;
data = gvdb_table_dereference (table, &item->value.pointer, 8, &size);
@@ -473,12 +584,11 @@
if G_UNLIKELY (data == NULL)
return NULL;
- variant = g_variant_new_from_data (G_VARIANT_TYPE_VARIANT,
- data, size, table->trusted,
- table->unref_user_data,
- table->ref_user_data ? table->ref_user_data (table->user_data) : table->user_data);
+ bytes = g_bytes_new_from_bytes (table->bytes, ((gchar *) data) - table->data, size);
+ variant = g_variant_new_from_bytes (G_VARIANT_TYPE_VARIANT, bytes, table->trusted);
value = g_variant_get_variant (variant);
g_variant_unref (variant);
+ g_bytes_unref (bytes);
return value;
}
@@ -562,7 +672,7 @@
* contained in the file. This newly-created #GvdbTable does not depend
* on the continued existence of @file.
*
- * You should call gvdb_table_unref() on the return result when you no
+ * You should call gvdb_table_free() on the return result when you no
* longer require it.
*
* Returns: a new #GvdbTable, or %NULL
@@ -580,14 +690,11 @@
return NULL;
new = g_slice_new0 (GvdbTable);
- new->user_data = file->ref_user_data ? file->ref_user_data (file->user_data) : file->user_data;
- new->ref_user_data = file->ref_user_data;
- new->unref_user_data = file->unref_user_data;
+ new->bytes = g_bytes_ref (file->bytes);
new->byteswapped = file->byteswapped;
new->trusted = file->trusted;
new->data = file->data;
new->size = file->size;
- new->ref_count = 1;
gvdb_table_setup_root (new, &item->value.pointer);
@@ -595,38 +702,16 @@
}
/**
- * gvdb_table_ref:
+ * gvdb_table_free:
* @file: a #GvdbTable
*
- * Increases the reference count on @file.
- *
- * Returns: a new reference on @file
- **/
-GvdbTable *
-gvdb_table_ref (GvdbTable *file)
-{
- g_atomic_int_inc (&file->ref_count);
-
- return file;
-}
-
-/**
- * gvdb_table_unref:
- * @file: a #GvdbTable
- *
- * Decreases the reference count on @file, possibly freeing it.
- *
- * Since: 2.26
+ * Frees @file.
**/
void
-gvdb_table_unref (GvdbTable *file)
+gvdb_table_free (GvdbTable *file)
{
- if (g_atomic_int_dec_and_test (&file->ref_count))
- {
- if (file->unref_user_data)
- file->unref_user_data (file->user_data);
- g_slice_free (GvdbTable, file);
- }
+ g_bytes_unref (file->bytes);
+ g_slice_free (GvdbTable, file);
}
/**
@@ -646,105 +731,3 @@
{
return !!*table->data;
}
-
-/**
- * gvdb_table_walk:
- * @table: a #GvdbTable
- * @key: a key corresponding to a list
- * @open_func: the #GvdbWalkOpenFunc
- * @value_func: the #GvdbWalkValueFunc
- * @close_func: the #GvdbWalkCloseFunc
- * @user_data: data to pass to the callbacks
- *
- * Looks up the list at @key and iterate over the items in it.
- *
- * First, @open_func is called to signal that we are starting to iterate over
- * the list. Then the list is iterated. When all items in the list have been
- * iterated over, the @close_func is called.
- *
- * When iterating, if a given item in the list is a value then @value_func is
- * called.
- *
- * If a given item in the list is itself a list then @open_func is called. If
- * that function returns %TRUE then the walk begins iterating the items in the
- * sublist, until there are no more items, at which point a matching
- * @close_func call is made. If @open_func returns %FALSE then no iteration of
- * the sublist occurs and no corresponding @close_func call is made.
- **/
-void
-gvdb_table_walk (GvdbTable *table,
- const gchar *key,
- GvdbWalkOpenFunc open_func,
- GvdbWalkValueFunc value_func,
- GvdbWalkCloseFunc close_func,
- gpointer user_data)
-{
- const struct gvdb_hash_item *item;
- const guint32_le *pointers[64];
- const guint32_le *enders[64];
- gsize name_lengths[64];
- gint index = 0;
-
- item = gvdb_table_lookup (table, key, 'L');
- name_lengths[0] = 0;
- pointers[0] = NULL;
- enders[0] = NULL;
- goto start_here;
-
- while (index)
- {
- close_func (name_lengths[index], user_data);
- index--;
-
- while (pointers[index] < enders[index])
- {
- const gchar *name;
- gsize name_len;
-
- item = gvdb_table_get_item (table, *pointers[index]++);
- start_here:
-
- if (item != NULL &&
- (name = gvdb_table_item_get_key (table, item, &name_len)))
- {
- if (item->type == 'L')
- {
- if (open_func (name, name_len, user_data))
- {
- guint length = 0;
-
- index++;
- g_assert (index < 64);
-
- gvdb_table_list_from_item (table, item,
- &pointers[index],
- &length);
- enders[index] = pointers[index] + length;
- name_lengths[index] = name_len;
- }
- }
- else if (item->type == 'v')
- {
- GVariant *value;
-
- value = gvdb_table_value_from_item (table, item);
-
- if (value != NULL)
- {
- if (table->byteswapped)
- {
- GVariant *tmp;
-
- tmp = g_variant_byteswap (value);
- g_variant_unref (value);
- value = tmp;
- }
-
- value_func (name, name_len, value, user_data);
- g_variant_unref (value);
- }
- }
- }
- }
- }
-}
diff --git a/gio/gvdb/gvdb-reader.h b/gio/gvdb/gvdb-reader.h
index 449241e..3982773 100644
--- a/gio/gvdb/gvdb-reader.h
+++ b/gio/gvdb/gvdb-reader.h
@@ -24,27 +24,21 @@
typedef struct _GvdbTable GvdbTable;
-typedef gpointer (*GvdbRefFunc) (gpointer data);
-
G_BEGIN_DECLS
G_GNUC_INTERNAL
+GvdbTable * gvdb_table_new_from_bytes (GBytes *bytes,
+ gboolean trusted,
+ GError **error);
+G_GNUC_INTERNAL
GvdbTable * gvdb_table_new (const gchar *filename,
gboolean trusted,
GError **error);
G_GNUC_INTERNAL
-GvdbTable * gvdb_table_new_from_data (const void *data,
- gsize data_len,
- gboolean trusted,
- gpointer user_data,
- GvdbRefFunc ref,
- GDestroyNotify unref,
- GError **error);
+void gvdb_table_free (GvdbTable *table);
G_GNUC_INTERNAL
-GvdbTable * gvdb_table_ref (GvdbTable *table);
-G_GNUC_INTERNAL
-void gvdb_table_unref (GvdbTable *table);
-
+gchar ** gvdb_table_get_names (GvdbTable *table,
+ gint *length);
G_GNUC_INTERNAL
gchar ** gvdb_table_list (GvdbTable *table,
const gchar *key);
@@ -61,28 +55,9 @@
G_GNUC_INTERNAL
gboolean gvdb_table_has_value (GvdbTable *table,
const gchar *key);
-
G_GNUC_INTERNAL
gboolean gvdb_table_is_valid (GvdbTable *table);
-typedef void (*GvdbWalkValueFunc) (const gchar *name,
- gsize name_len,
- GVariant *value,
- gpointer user_data);
-typedef gboolean (*GvdbWalkOpenFunc) (const gchar *name,
- gsize name_len,
- gpointer user_data);
-typedef void (*GvdbWalkCloseFunc) (gsize name_len,
- gpointer user_data);
-
-G_GNUC_INTERNAL
-void gvdb_table_walk (GvdbTable *table,
- const gchar *key,
- GvdbWalkOpenFunc open_func,
- GvdbWalkValueFunc value_func,
- GvdbWalkCloseFunc close_func,
- gpointer user_data);
-
G_END_DECLS
#endif /* __gvdb_reader_h__ */
diff --git a/gio/gvdb/gvdb.doap b/gio/gvdb/gvdb.doap
index b4ae60c..8c5f3e8 100644
--- a/gio/gvdb/gvdb.doap
+++ b/gio/gvdb/gvdb.doap
@@ -23,9 +23,34 @@
<maintainer>
<foaf:Person>
- <foaf:name>Ryan Lortie</foaf:name>
- <foaf:mbox rdf:resource='mailto:desrt@desrt.ca'/>
- <gnome:userid>ryanl</gnome:userid>
+ <foaf:name>Matthias Clasen</foaf:name>
+ <foaf:mbox rdf:resource="mailto:mclasen@redhat.com"/>
+ <gnome:userid>matthiasc</gnome:userid>
+ </foaf:Person>
+ </maintainer>
+
+ <maintainer>
+ <foaf:Person>
+ <foaf:name>Allison Ryan Lortie</foaf:name>
+ <foaf:mbox rdf:resource="mailto:desrt@desrt.ca"/>
+ <gnome:userid>desrt</gnome:userid>
+ </foaf:Person>
+ </maintainer>
+
+ <maintainer>
+ <foaf:Person>
+ <foaf:name>Philip Withnall</foaf:name>
+ <foaf:mbox rdf:resource="mailto:philip@tecnocode.co.uk"/>
+ <foaf:mbox rdf:resource="mailto:withnall@endlessm.com"/>
+ <gnome:userid>pwithnall</gnome:userid>
+ </foaf:Person>
+ </maintainer>
+
+ <maintainer>
+ <foaf:Person>
+ <foaf:name>Emmanuele Bassi</foaf:name>
+ <foaf:mbox rdf:resource="mailto:ebassi@gnome.org"/>
+ <gnome:userid>ebassi</gnome:userid>
</foaf:Person>
</maintainer>
