diff options
Diffstat (limited to 'camel/url-util.c')
-rw-r--r-- | camel/url-util.c | 418 |
1 files changed, 0 insertions, 418 deletions
diff --git a/camel/url-util.c b/camel/url-util.c deleted file mode 100644 index 7fd63d7b6d..0000000000 --- a/camel/url-util.c +++ /dev/null @@ -1,418 +0,0 @@ -/* -*- Mode: C; tab-width: 8; indent-tabs-mode: t; c-basic-offset: 8 -*- */ -/* url-util.c : utility functions to parse URLs */ - - -/* - * Author : - * Bertrand Guiheneuf <Bertrand.Guiheneuf@aful.org> - * - * Copyright 1999 International GNOME Support (http://www.gnome-support.com) - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public License as - * published by the Free Software Foundation; either version 2 of the - * License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 - * USA - */ - - - -/* - Here we deal with URL following the general scheme: - protocol://user:password@host:port/name - where name is a path-like string (ie dir1/dir2/....) - See rfc1738 for the complete description of - Uniform Ressource Locators - - Bertrand. */ -/* - XXX TODO: recover the words between #'s or ?'s after the path */ - -#include <config.h> -#include "url-util.h" - -/* general item finder */ -/* it begins the search at position @position in @url, - returns true when the item is found, amd set position after the item */ -typedef gboolean find_item_func(const gchar *url, gchar **item, guint *position, gboolean *error); - -/* used to find one item (protocol, then user .... */ -typedef struct { - char *item_name; /* item name : for debug only */ - gchar **item_value; /* where the item value will go */ - find_item_func *find_func; /* item finder */ -} FindStepStruct; - -static gboolean _find_protocol (const gchar *url, gchar **item, guint *position, gboolean *error); -static gboolean _find_user (const gchar *url, gchar **item, guint *position, gboolean *error); -static gboolean _find_passwd (const gchar *url, gchar **item, guint *position, gboolean *error); -static gboolean _find_host (const gchar *url, gchar **item, guint *position, gboolean *error); -static gboolean _find_port (const gchar *url, gchar **item, guint *position, gboolean *error); -static gboolean _find_path (const gchar *url, gchar **item, guint *position, gboolean *error); - - - -/** - * g_url_new: create an Gurl object from a string - * - * @url_string: The string containing the URL to scan - * - * This routine takes a gchar and parses it as an - * URL of the form: - * protocol://user:password@host:port/path - * there is no test on the values. For example, - * "port" can be a string, not only a number ! - * The Gurl structure fields ar filled with - * the scan results. When a member of the - * general URL can not be found, the corresponding - * Gurl member is NULL - * Fields filled in the Gurl structure are allocated - * and url_string is not modified. - * - * Return value: a Gurl structure containing the URL items. - **/ -Gurl *g_url_new (const gchar* url_string) -{ - Gurl *g_url; - - gchar *protocol; - gchar *user; - gchar *passwd; - gchar *host; - gchar *port; - gchar *path; - - guint position = 0; - gboolean error; - gboolean found; - guint i; - - g_url = g_new (Gurl,1); - -#define NB_STEP_URL 6 - { - FindStepStruct step[NB_STEP_URL] = { - { "protocol", &(g_url->protocol), _find_protocol}, - { "user", &(g_url->user), _find_user}, - { "password", &(g_url->passwd), _find_passwd}, - { "host", &(g_url->host), _find_host}, - { "port", &(g_url->port), _find_port}, - { "path", &(g_url->path), _find_path} - }; - - for (i = 0; i < NB_STEP_URL; i++) { - found = step[i].find_func (url_string, - step[i].item_value, - &position, - &error); - } - } - - return g_url; -} - - - -void -g_url_free (Gurl *url) -{ - g_assert (url); - - g_free (url->protocol); - g_free (url->user); - g_free (url->passwd); - g_free (url->host); - g_free (url->port); - g_free (url->path); - - g_free (url); - -} - -/**** PARSING FUNCTIONS ****/ - -/* So, yes, I must admit there would have been more elegant - ways to do this, but it works, and quite well :) */ - - -static gboolean -_find_protocol (const gchar *url, gchar **item, guint *position, gboolean *error) -{ - - guint i; - gint len_url; - - g_assert (url); - g_assert (item); - g_assert (position); - - len_url = strlen (url); - - *item = NULL; - *error = FALSE; - i = *position; - - /* find a ':' */ - while ((i < len_url) && (url[i] != ':')) i++; - - if (i == len_url) return FALSE; - i++; - - /* check if it is followed by a "//" */ - if ((i < len_url) && (url[i++] == '/')) - if ((i < len_url) && (url[i++] == '/')) - { - *item = g_strndup (url, i-3); - *position = i; - return TRUE; - } - - return FALSE; -} - - - - -static gboolean -_find_user (const gchar *url, gchar **item, guint *position, gboolean *error) -{ - guint i; - guint at_pos; - gint len_url; - - g_assert (url); - g_assert (item); - g_assert (position); - - len_url = strlen (url); - *item = NULL; - i = *position; - - /* find a '@' */ - while ((i < len_url) && (url[i] != '@')) i++; - - if (i == len_url) return FALSE; - at_pos = i; - i = *position; - - /* find a ':' */ - while ((i < at_pos) && (url[i] != ':')) i++; - - /* now if i has not been incremented at all, there is no user */ - if (i == *position) { - (*position)++; - return FALSE; - } - - *item = g_strndup (url+ *position, i - *position); - if (i < at_pos) *position = i + 1; /* there was a ':', skip it */ - else *position = i; - - return TRUE; -} - -static gboolean -_find_passwd (const gchar *url, gchar **item, guint *position, gboolean *error) -{ - guint i; - gint len_url; - gchar *str_passwd; - - g_assert (url); - g_assert (item); - g_assert (position); - - len_url = strlen (url); - *item = NULL; - i = *position; - - /* find a '@' */ - while ((i < len_url) && (url[i] != '@')) i++; - - if (i == len_url) return FALSE; - /*i has not been incremented at all, there is no passwd */ - if (i == *position) { - *position = i + 1; - return FALSE; - } - - *item = g_strndup (url + *position, i - *position); - *position = i + 1; /* skip it the '@' */ - - return TRUE; -} - - - -static gboolean -_find_host (const gchar *url, gchar **item, guint *position, gboolean *error) -{ - guint i; - guint slash_pos; - gint len_url; - - g_assert (url); - g_assert (item); - g_assert (position); - - len_url = strlen (url); - *item = NULL; - i = *position; - - /* find a '/' */ - while ((i < len_url) && (url[i] != '/')) i++; - - slash_pos = i; - i = *position; - - /* find a ':' */ - while ( (i < slash_pos) && (url[i] != ':') ) i++; - - /* at this point if i has not been incremented at all, - there is no host */ - if (i == *position) { - /* if we have not met / or \0, we have : and must skip it */ - if (i < slash_pos) (*position)++; - return FALSE; - } - - *item = g_strndup (url + *position, i - *position); - if (i < slash_pos) *position = i + 1; /* there was a ':', skip it */ - else *position=i; - - return TRUE; -} - - -static gboolean -_find_port (const gchar *url, gchar **item, guint *position, gboolean *error) -{ - guint i; - guint slash_pos; - gint len_url; - - g_assert (url); - g_assert (item); - g_assert (position); - - len_url = strlen (url); - *item = NULL; - i=*position; - - /* find a '/' */ - while ((i < len_url) && (url[i] != '/')) i++; - - slash_pos = i; - i = *position; - - /* find a ':' */ - while ((i < slash_pos) && (url[i] != ':')) i++; - - /* at this point if i has not been incremented at all, */ - /* there is no port */ - if (i == *position) return FALSE; - - *item = g_strndup (url+ *position, i - *position); - *position = i; - return TRUE; -} - - -static gboolean -_find_path (const gchar *url, gchar **item, guint *position, gboolean *error) -{ - guint i; - gint len_url; - - g_assert (url); - g_assert (item); - g_assert (position); - - len_url = strlen (url); - *item = NULL; - i = *position; - - - /* find a '#' */ - while ((i < len_url) && (url[i] != '#') && (url[i] != '?')) i++; - - /*i has not been incremented at all, there is no path */ - if (i == *position) return FALSE; - - *item = g_strndup (url + *position, i - *position); - *position=i; - - return TRUE; -} - - - - - -/**** TEST ROUTINE - NOT COMPILED BY DEFAULT ****/ - -/* to tests this file : - gcc -o test_url_util `glib-config --cflags` -I.. -DTEST_URL_UTIL url-util.c `glib-config --libs` - ./test_url_util URL -*/ -#ifdef TEST_URL_UTIL - - - -int -main (int argc, char **argv) -{ - - gchar *url; - gchar *protocol; - gchar *user; - gchar *passwd; - gchar *host; - gchar *port; - gchar *path; - guint position=0; - gboolean error; - gboolean found; - guint i; - guint i_pos; - -#define NB_STEP_TEST 6 - FindStepStruct test_step[NB_STEP_TEST] = { - { "protocol", &protocol, _find_protocol}, - { "user", &user, _find_user}, - { "password", &passwd, _find_passwd}, - { "host", &host, _find_host}, - { "port", &port, _find_port}, - { "path", &path, _find_path} - }; - url = argv[1]; - printf("URL to test : %s\n\n", url); - for (i=0; i<NB_STEP_TEST; i++) { - found = test_step[i].find_func (url, - test_step[i].item_value, - &position, - &error); - if (found) { - printf ("\t\t\t\t** %s found : %s\n", - test_step[i].item_name, - *(test_step[i].item_value)); - } else printf ("** %s not found in URL\n", test_step[i].item_name); - printf ("next item position:\n"); - printf ("%s\n", url); - for (i_pos = 0; i_pos < position; i_pos++) printf (" "); - printf ("^\n"); - - } - -} - -#endif /* TEST_URL_UTIL */ |