glib/grefstring.c

   1 /* grefstring.c: Reference counted strings
   2  *
   3  * Copyright 2018  Emmanuele Bassi
   4  *
   5  * This library is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU Lesser General Public
   7  * License as published by the Free Software Foundation; either
   8  * version 2.1 of the License, or (at your option) any later version.
   9  *
  10  * This library is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * Lesser General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU Lesser General Public
  16  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  17  */
  18
  19 /**
  20  * SECTION:refstring
  21  * @Title: Reference counted strings
  22  * @Short_description: Strings with reference counted memory management
  23  *
  24  * Reference counted strings are normal C strings that have been augmented
  25  * with a reference counter to manage their resources. You allocate a new
  26  * reference counted string and acquire and release references as needed,
  27  * instead of copying the string among callers; when the last reference on
  28  * the string is released, the resources allocated for it are freed.
  29  *
  30  * Typically, reference counted strings can be used when parsing data from
  31  * files and storing them into data structures that are passed to various
  32  * callers:
  33  *
  34  * |[<!-- language="C" -->
  35  * PersonDetails *
  36  * person_details_from_data (const char *data)
  37  * {
  38  *   // Use g_autoptr() to simplify error cases
  39  *   g_autoptr(GRefString) full_name = NULL;
  40  *   g_autoptr(GRefString) address =  NULL;
  41  *   g_autoptr(GRefString) city = NULL;
  42  *   g_autoptr(GRefString) state = NULL;
  43  *   g_autoptr(GRefString) zip_code = NULL;
  44  *
  45  *   // parse_person_details() is defined elsewhere; returns refcounted strings
  46  *   if (!parse_person_details (data, &full_name, &address, &city, &state, &zip_code))
  47  *     return NULL;
  48  *
  49  *   if (!validate_zip_code (zip_code))
  50  *     return NULL;
  51  *
  52  *   // add_address_to_cache() and add_full_name_to_cache() are defined
  53  *   // elsewhere; they add strings to various caches, using refcounted
  54  *   // strings to avoid copying data over and over again
  55  *   add_address_to_cache (address, city, state, zip_code);
  56  *   add_full_name_to_cache (full_name);
  57  *
  58  *   // person_details_new() is defined elsewhere; it takes a reference
  59  *   // on each string
  60  *   PersonDetails *res = person_details_new (full_name,
  61  *                                            address,
  62  *                                            city,
  63  *                                            state,
  64  *                                            zip_code);
  65  *
  66  *   return res;
  67  * }
  68  * ]|
  69  *
  70  * In the example above, we have multiple functions taking the same strings
  71  * for different uses; with typical C strings, we'd have to copy the strings
  72  * every time the life time rules of the data differ from the life time of
  73  * the string parsed from the original buffer. With reference counted strings,
  74  * each caller can take a reference on the data, and keep it as long as it
  75  * needs to own the string.
  76  *
  77  * Reference counted strings can also be "interned" inside a global table
  78  * owned by GLib; while an interned string has at least a reference, creating
  79  * a new interned reference counted string with the same contents will return
  80  * a reference to the existing string instead of creating a new reference
  81  * counted string instance. Once the string loses its last reference, it will
  82  * be automatically removed from the global interned strings table.
  83  *
  84  * Since: 2.58
  85  */
  86
  87 #include "config.h"
  88
  89 #include "grefstring.h"
  90
  91 #include "ghash.h"
  92 #include "gmessages.h"
  93 #include "grcbox.h"
  94 #include "gthread.h"
  95
  96 #include <string.h>
  97
  98 /* A global table of refcounted strings; the hash table does not own
  99  * the strings, just a pointer to them. Strings are interned as long
 100  * as they are alive; once their reference count drops to zero, they
 101  * are removed from the table
 102  */
 103 G_LOCK_DEFINE_STATIC (interned_ref_strings);
 104 static GHashTable *interned_ref_strings;
 105
 106 /**
 107  * g_ref_string_new:
 108  * @str: (not nullable): a NUL-terminated string
 109  *
 110  * Creates a new reference counted string and copies the contents of @str
 111  * into it.
 112  *
 113  * Returns: (transfer full) (not nullable): the newly created reference counted string
 114  *
 115  * Since: 2.58
 116  */
 117 char *
 118 g_ref_string_new (const char *str)
 119 {
 120   char *res;
 121   gsize len;
 122
 123   g_return_val_if_fail (str != NULL, NULL);
 124
 125   len = strlen (str);
 126
 127   res = (char *) g_atomic_rc_box_dup (sizeof (char) * len + 1, str);
 128
 129   return res;
 130 }
 131
 132 /**
 133  * g_ref_string_new_len:
 134  * @str: (not nullable): a string
 135  * @len: length of @str to use, or -1 if @str is nul-terminated
 136  *
 137  * Creates a new reference counted string and copies the contents of @str
 138  * into it, up to @len bytes.
 139  *
 140  * Since this function does not stop at nul bytes, it is the caller's
 141  * responsibility to ensure that @str has at least @len addressable bytes.
 142  *
 143  * Returns: (transfer full) (not nullable): the newly created reference counted string
 144  *
 145  * Since: 2.58
 146  */
 147 char *
 148 g_ref_string_new_len (const char *str, gssize len)
 149 {
 150   char *res;
 151
 152   g_return_val_if_fail (str != NULL, NULL);
 153
 154   if (len < 0)
 155     return g_ref_string_new (str);
 156
 157   /* allocate then copy as str[len] may not be readable */
 158   res = (char *) g_atomic_rc_box_alloc ((gsize) len + 1);
 159   memcpy (res, str, len);
 160   res[len] = '\0';
 161
 162   return res;
 163 }
 164
 165 /* interned_str_equal: variant of g_str_equal() that compares
 166  * pointers as well as contents; this avoids running strcmp()
 167  * on arbitrarily long strings, as it's more likely to have
 168  * g_ref_string_new_intern() being called on the same refcounted
 169  * string instance, than on a different string with the same
 170  * contents
 171  */
 172 static gboolean
 173 interned_str_equal (gconstpointer v1,
 174                     gconstpointer v2)
 175 {
 176   const char *str1 = v1;
 177   const char *str2 = v2;
 178
 179   if (v1 == v2)
 180     return TRUE;
 181
 182   return strcmp (str1, str2) == 0;
 183 }
 184
 185 /**
 186  * g_ref_string_new_intern:
 187  * @str: (not nullable): a NUL-terminated string
 188  *
 189  * Creates a new reference counted string and copies the content of @str
 190  * into it.
 191  *
 192  * If you call this function multiple times with the same @str, or with
 193  * the same contents of @str, it will return a new reference, instead of
 194  * creating a new string.
 195  *
 196  * Returns: (transfer full) (not nullable): the newly created reference
 197  *   counted string, or a new reference to an existing string
 198  *
 199  * Since: 2.58
 200  */
 201 char *
 202 g_ref_string_new_intern (const char *str)
 203 {
 204   char *res;
 205
 206   g_return_val_if_fail (str != NULL, NULL);
 207
 208   G_LOCK (interned_ref_strings);
 209
 210   if (G_UNLIKELY (interned_ref_strings == NULL))
 211     interned_ref_strings = g_hash_table_new (g_str_hash, interned_str_equal);
 212
 213   res = g_hash_table_lookup (interned_ref_strings, str);
 214   if (res != NULL)
 215     {
 216       /* We acquire the reference while holding the lock, to
 217        * avoid a potential race between releasing the lock on
 218        * the hash table and another thread releasing the reference
 219        * on the same string
 220        */
 221       g_atomic_rc_box_acquire (res);
 222       G_UNLOCK (interned_ref_strings);
 223       return res;
 224     }
 225
 226   res = g_ref_string_new (str);
 227   g_hash_table_add (interned_ref_strings, res);
 228   G_UNLOCK (interned_ref_strings);
 229
 230   return res;
 231 }
 232
 233 /**
 234  * g_ref_string_acquire:
 235  * @str: a reference counted string
 236  *
 237  * Acquires a reference on a string.
 238  *
 239  * Returns: the given string, with its reference count increased
 240  *
 241  * Since: 2.58
 242  */
 243 char *
 244 g_ref_string_acquire (char *str)
 245 {
 246   g_return_val_if_fail (str != NULL, NULL);
 247
 248   return g_atomic_rc_box_acquire (str);
 249 }
 250
 251 static void
 252 remove_if_interned (gpointer data)
 253 {
 254   char *str = data;
 255
 256   G_LOCK (interned_ref_strings);
 257
 258   if (G_LIKELY (interned_ref_strings != NULL))
 259     {
 260       g_hash_table_remove (interned_ref_strings, str);
 261
 262       if (g_hash_table_size (interned_ref_strings) == 0)
 263         g_clear_pointer (&interned_ref_strings, g_hash_table_destroy);
 264     }
 265
 266   G_UNLOCK (interned_ref_strings);
 267 }
 268
 269 /**
 270  * g_ref_string_release:
 271  * @str: a reference counted string
 272  *
 273  * Releases a reference on a string; if it was the last reference, the
 274  * resources allocated by the string are freed as well.
 275  *
 276  * Since: 2.58
 277  */
 278 void
 279 g_ref_string_release (char *str)
 280 {
 281   g_return_if_fail (str != NULL);
 282
 283   g_atomic_rc_box_release_full (str, remove_if_interned);
 284 }
 285
 286 /**
 287  * g_ref_string_length:
 288  * @str: a reference counted string
 289  *
 290  * Retrieves the length of @str.
 291  *
 292  * Returns: the length of the given string, in bytes
 293  *
 294  * Since: 2.58
 295  */
 296 gsize
 297 g_ref_string_length (char *str)
 298 {
 299   g_return_val_if_fail (str != NULL, 0);
 300
 301   return g_atomic_rc_box_get_size (str) - 1;
 302 }