| 1 |
<?php
|
| 2 |
/**
|
| 3 |
* @file Main drupal interface to import_html. This file contains Drupal hooks
|
| 4 |
* and some config/preferences. Actual processing functions are in
|
| 5 |
* import_html_process.inc
|
| 6 |
*
|
| 7 |
* Synopsis:
|
| 8 |
*
|
| 9 |
* Facility to import an existing, static HTML site structure into the Drupal
|
| 10 |
* CMS as structured nodes.
|
| 11 |
*
|
| 12 |
* Intent:
|
| 13 |
*
|
| 14 |
* Allow an admin to define a source directory of an existing
|
| 15 |
* traditional static HTML website, and import (as much as possible)
|
| 16 |
* the content and structure into a drupal site.
|
| 17 |
* Source files will be stripped of exisiting chrome and navigation
|
| 18 |
* elements before being inserted as nodes.
|
| 19 |
*
|
| 20 |
* See import_html_help.htm for more
|
| 21 |
*
|
| 22 |
* Main data extraction routines are in import_html_process.inc. Module-specific
|
| 23 |
* data extraction methods have been shifted into import_html_modules.inc UI
|
| 24 |
* forms and themes have been shifted to import_html_ui.inc.
|
| 25 |
* Drupal hooks and some configs remain in this module file.
|
| 26 |
*
|
| 27 |
* import_html libraries can be used by external modules, specifically wrapper.
|
| 28 |
* module (private development) and static.module (experimental)
|
| 29 |
*
|
| 30 |
* @ingroup import_html Import HTML
|
| 31 |
* @author Dan Morrison http://coders.co.nz/
|
| 32 |
* @version $Id: import_html.module,v 1.55.2.2 2009/03/14 03:18:50 dman Exp $
|
| 33 |
*
|
| 34 |
*/
|
| 35 |
|
| 36 |
$import_html_path = drupal_get_path('module', 'import_html');
|
| 37 |
$import_html_path = dirname(__FILE__);
|
| 38 |
set_include_path($import_html_path . PATH_SEPARATOR . get_include_path());
|
| 39 |
$import_html_library_path = $import_html_path .'/coders_php_library';
|
| 40 |
set_include_path($import_html_library_path . PATH_SEPARATOR . get_include_path());
|
| 41 |
|
| 42 |
/**
|
| 43 |
* @name Debug Flag
|
| 44 |
* Used for testing only
|
| 45 |
* @{
|
| 46 |
*/
|
| 47 |
if (! function_exists('debug')) {
|
| 48 |
require_once 'debug.inc';
|
| 49 |
define('DEBUGLEVEL', 0);
|
| 50 |
define('DEBUGBACKTRACE', 8);
|
| 51 |
}
|
| 52 |
#debug_set_level(3);
|
| 53 |
|
| 54 |
/**
|
| 55 |
* @}
|
| 56 |
*/
|
| 57 |
|
| 58 |
// Split into libraries in prep for D6
|
| 59 |
require_once "import_html_process.inc";
|
| 60 |
require_once "import_html_ui.inc";
|
| 61 |
|
| 62 |
require_once "xml-transform.inc";
|
| 63 |
require_once "file-routines.inc";
|
| 64 |
|
| 65 |
|
| 66 |
/**
|
| 67 |
* @name $_import_html_file_classes
|
| 68 |
* poor-mans mime-list
|
| 69 |
*/
|
| 70 |
|
| 71 |
global $_import_html_file_classes;
|
| 72 |
$_import_html_file_classes = array(
|
| 73 |
"htm" => 'html',
|
| 74 |
"html" => 'html',
|
| 75 |
"shtml" => 'html',
|
| 76 |
"php" => 'html',
|
| 77 |
"asp" => 'html',
|
| 78 |
"aspx" => 'html',
|
| 79 |
"jsp" => 'html',
|
| 80 |
|
| 81 |
"gif" => "image",
|
| 82 |
"jpg" => "image",
|
| 83 |
"png" => "image",
|
| 84 |
|
| 85 |
"css" => "resource",
|
| 86 |
"js" => "resource",
|
| 87 |
"swf" => "resource",
|
| 88 |
|
| 89 |
"pdf" => "document",
|
| 90 |
"txt" => "document",
|
| 91 |
"rtf" => "document"
|
| 92 |
);
|
| 93 |
|
| 94 |
// How to handle duplicates and other problems when importing
|
| 95 |
define("IMPORT_HTML_SKIP", 0);
|
| 96 |
define("IMPORT_HTML_MERGE", 1);
|
| 97 |
define("IMPORT_HTML_GUESS", 2);
|
| 98 |
define("IMPORT_HTML_DEFAULT", 4);
|
| 99 |
|
| 100 |
define("IMPORT_HTML_MAX_LABEL_LENGTH", 24);
|
| 101 |
|
| 102 |
define('IMPORT_HTML_ADMIN_PATH', 'admin/build/import_html');
|
| 103 |
|
| 104 |
/**
|
| 105 |
* Implementation of hook_menu() .
|
| 106 |
*
|
| 107 |
* Declares all the menu items relating to this module
|
| 108 |
* And all the public functions it supplies
|
| 109 |
*/
|
| 110 |
function import_html_menu($may_cache) {
|
| 111 |
if ($may_cache) {
|
| 112 |
$items[] = array(
|
| 113 |
'path' => IMPORT_HTML_ADMIN_PATH,
|
| 114 |
'title' => t('Import HTML'),
|
| 115 |
'description' => t("Import/Export an entire site or directory to/from static HTML"),
|
| 116 |
'callback' => 'system_admin_menu_block_page', // Overview menu thing
|
| 117 |
'access' => user_access('access import_html'),
|
| 118 |
'type' => MENU_NORMAL_ITEM,
|
| 119 |
);
|
| 120 |
$items[] = array(
|
| 121 |
'path' => IMPORT_HTML_ADMIN_PATH .'/settings',
|
| 122 |
'title' => t('Import HTML Settings'),
|
| 123 |
'description' => t('Adjust the import_html options and settings.'),
|
| 124 |
'callback' => 'drupal_get_form',
|
| 125 |
'weight' => 1,
|
| 126 |
'callback arguments' => array('import_html_admin_settings'),
|
| 127 |
'access' => user_access('administer site configuration'),
|
| 128 |
'type' => MENU_NORMAL_ITEM,
|
| 129 |
);
|
| 130 |
$items[] = array(
|
| 131 |
'path' => IMPORT_HTML_ADMIN_PATH .'/profile',
|
| 132 |
'title' => t('Import HTML Profiles'),
|
| 133 |
'description' => t('Edit multiple import profiles.'),
|
| 134 |
'callback' => 'import_html_profiles_page',
|
| 135 |
'weight' => 2,
|
| 136 |
'access' => user_access('administer site configuration'),
|
| 137 |
'type' => MENU_NORMAL_ITEM,
|
| 138 |
);
|
| 139 |
$items[] = array(
|
| 140 |
'path' => IMPORT_HTML_ADMIN_PATH .'/import_site',
|
| 141 |
'title' => t('Import HTML Site'),
|
| 142 |
'description' => t("Import/Export an entire site to/from static HTML"),
|
| 143 |
'weight' => -1,
|
| 144 |
'callback' => 'drupal_get_form',
|
| 145 |
'callback arguments' => array('import_html_process_form'),
|
| 146 |
'access' => user_access('access import_html'),
|
| 147 |
'type' => MENU_NORMAL_ITEM,
|
| 148 |
);
|
| 149 |
$items[] = array(
|
| 150 |
'path' => IMPORT_HTML_ADMIN_PATH .'/demo',
|
| 151 |
'title' => t('Demo'),
|
| 152 |
'description' => t('Demonstrate or test HTML Import on one file.'),
|
| 153 |
'callback' => 'drupal_get_form',
|
| 154 |
'callback arguments' => array('import_html_demo_form'),
|
| 155 |
'access' => user_access('access import_html'),
|
| 156 |
'type' => MENU_NORMAL_ITEM,
|
| 157 |
);
|
| 158 |
}
|
| 159 |
return $items ? $items : array();
|
| 160 |
}
|
| 161 |
|
| 162 |
/**
|
| 163 |
* Hook Implimentation
|
| 164 |
*/
|
| 165 |
function import_html_perm() {
|
| 166 |
return array(
|
| 167 |
'access import_html'
|
| 168 |
);
|
| 169 |
}
|
| 170 |
|
| 171 |
/**
|
| 172 |
* Return help text describing this module
|
| 173 |
*
|
| 174 |
* @param $section string Context this help is being called from
|
| 175 |
* @return string
|
| 176 |
*/
|
| 177 |
function import_html_help($section) {
|
| 178 |
switch ($section) {
|
| 179 |
case 'admin/build/modules#description' :
|
| 180 |
return t("Import/Export an entire site to/from static HTML");
|
| 181 |
case 'admin/build/modules/import_html' :
|
| 182 |
return t("Import/Export an entire site to/from static HTML");
|
| 183 |
case 'admin/help#import_html' :
|
| 184 |
return file_get_contents(drupal_get_path("module", "import_html") ."/docs/import_html_help.htm");
|
| 185 |
case IMPORT_HTML_ADMIN_PATH :
|
| 186 |
return l("DO check the help page, this is a complex process", 'admin/help/import_html');
|
| 187 |
case 'admin/settings/import_html' :
|
| 188 |
return l("DO check the help page, this is a complex process", 'admin/help/import_html');
|
| 189 |
case 'admin/build/import_html/import_site' :
|
| 190 |
$output = t("<p>
|
| 191 |
For background, remember to read <a href='!help_link'>the Import Html help page</a>.
|
| 192 |
The <a href='!settings_link'>settings page</a> contains the config options.
|
| 193 |
</p><p>
|
| 194 |
For a quick intro, try <a href='!demo_link'>a quick demo</a>.
|
| 195 |
</p><p>
|
| 196 |
The source website files must be directly available to this server.
|
| 197 |
This process will NOT yet import the structure of a client website to a remote server,
|
| 198 |
or spider all the resources of a remote site.
|
| 199 |
</p><p>
|
| 200 |
<b>Note</b> Big sites <i>will</i> take a long time to process,
|
| 201 |
and processing may timeout.
|
| 202 |
You can either increase the php timeout value and wait around,
|
| 203 |
or just do sections at a time by using the subsection parameter above.
|
| 204 |
</p>",
|
| 205 |
array(
|
| 206 |
'!help_link' => url('admin/help/import_html'),
|
| 207 |
'!settings_link' => url(IMPORT_HTML_ADMIN_PATH .'/settings'),
|
| 208 |
'!demo_link' => url(IMPORT_HTML_ADMIN_PATH .'/demo'),
|
| 209 |
)
|
| 210 |
);
|
| 211 |
return $output;
|
| 212 |
|
| 213 |
case 'admin/build/import_html/demo' :
|
| 214 |
return t("<p>
|
| 215 |
Enter one HTML filepath or URL to process as an import.
|
| 216 |
The retrieved data <em>will not</em> automatically become part of the
|
| 217 |
site unless you confirm it, it's just a demo of what data would be
|
| 218 |
extracted on a simple import.
|
| 219 |
</p><p>
|
| 220 |
The given file will be passed through the currently configured import process
|
| 221 |
and shown as a node-edit form, displaying all the deduced content in the appropriate fields.
|
| 222 |
</p><p>
|
| 223 |
If you see <em>too much</em> of the page in the 'body' area (nested navbars and layout)
|
| 224 |
then the XSL import template or selector needs to be made more specific.
|
| 225 |
If you see none, or not enough content in the body area, the template or
|
| 226 |
selector needs to be adjusted to encompass the text correctly.
|
| 227 |
</p><p>
|
| 228 |
If you get an error or no result, the input HTML is probably too invalid to work with.
|
| 229 |
</p><p>
|
| 230 |
Single demo imports do not have the full context information to work with,
|
| 231 |
so the menu or URL alias (and internal relinking) shown may not be representative of the real result.
|
| 232 |
</p>");
|
| 233 |
|
| 234 |
}
|
| 235 |
return false;
|
| 236 |
}
|
| 237 |
|
| 238 |
|
| 239 |
|
| 240 |
/**
|
| 241 |
* Implementation of hook_form_alter() .
|
| 242 |
*/
|
| 243 |
function import_html_form_alter($form_id, &$form) {
|
| 244 |
// I need to pretend the demo page is a submit page
|
| 245 |
if (($form_id == 'page_node_form') && ($form['#action'] == url(IMPORT_HTML_ADMIN_PATH .'/demo')) ) {
|
| 246 |
$form['#action'] = url('node/add/page');
|
| 247 |
debug("Redirecting the demo form to pretend to be a node_edit form", 2);
|
| 248 |
}
|
| 249 |
}
|
| 250 |
|
| 251 |
////////////////////////////////////////////
|
| 252 |
// End of Drupal core hooks.
|
| 253 |
// Module utilities below.
|
| 254 |
|
| 255 |
|
| 256 |
/**
|
| 257 |
* A wrapper to variable_set, variable_get to encapsulate multiple import 'profiles'
|
| 258 |
*
|
| 259 |
* This natively just returns the settings from the current 'default' profile,
|
| 260 |
* but also allows the settings forms to be extended to other sets.
|
| 261 |
*
|
| 262 |
* Use INSTEAD OF variable_get() and it will return the 'default' or 'active' profile vars.
|
| 263 |
*
|
| 264 |
* @param $var Name of the variable within the currently active profile to retrieve.
|
| 265 |
* @param $val If set, sets this variable within the profile and saves it.
|
| 266 |
* @see import_html_current_profile();
|
| 267 |
*/
|
| 268 |
function import_html_variable($var, $val = NULL) {
|
| 269 |
|
| 270 |
static $import_html_profiles;
|
| 271 |
if (! $import_html_profiles ) {
|
| 272 |
$import_html_profiles = variable_get('import_html_profiles', array());
|
| 273 |
}
|
| 274 |
|
| 275 |
$import_html_current_profile_id = import_html_current_profile_id();
|
| 276 |
|
| 277 |
if (! $import_html_profiles[$import_html_current_profile_id]) {
|
| 278 |
// Fill in defaults (should only be needed first time, if that
|
| 279 |
drupal_set_message('import_html_variable initing profile from nowhere - should this ever happen?');
|
| 280 |
$import_html_profiles[$import_html_current_profile_id] = import_html_profile_defaults();
|
| 281 |
}
|
| 282 |
|
| 283 |
$import_html_profile = &$import_html_profiles[$import_html_current_profile_id];
|
| 284 |
|
| 285 |
if (isset($val)) {
|
| 286 |
$import_html_profile[$var] = $val;
|
| 287 |
#$import_html_profiles[$import_html_current_profile_id] = $import_html_profile;
|
| 288 |
variable_set('import_html_profiles', $import_html_profiles);
|
| 289 |
}
|
| 290 |
|
| 291 |
return $import_html_profile[$var];
|
| 292 |
}
|
| 293 |
|
| 294 |
/**
|
| 295 |
* Accessor for the current profile data
|
| 296 |
*
|
| 297 |
* Returns the data in a named import_html_profile. The currently active one if not explicitly defined.
|
| 298 |
*
|
| 299 |
* @param $profile If set, saves this data back to the saved settings. Pass FALSE to delete it.
|
| 300 |
*/
|
| 301 |
function import_html_profile($profile_id = null, $profile = null) {
|
| 302 |
$import_html_profile_id = import_html_current_profile_id($profile_id);
|
| 303 |
static $import_html_profiles;
|
| 304 |
if (! $import_html_profiles ) {
|
| 305 |
$import_html_profiles = variable_get('import_html_profiles', array());
|
| 306 |
}
|
| 307 |
if (isset($profile)) {
|
| 308 |
$import_html_profiles[$import_html_profile_id] = $profile;
|
| 309 |
// Allow a quiet delete
|
| 310 |
if (!$profile) { unset($import_html_profiles[$import_html_profile_id]); }
|
| 311 |
variable_set('import_html_profiles', $import_html_profiles);
|
| 312 |
}
|
| 313 |
|
| 314 |
return $import_html_profiles[$import_html_profile_id];
|
| 315 |
}
|
| 316 |
|
| 317 |
|
| 318 |
/**
|
| 319 |
* Accessor for a persistant profile switcher.
|
| 320 |
*
|
| 321 |
* @param $profile_id If set, this becomes the active profile for the duration of the request.
|
| 322 |
* @returns the current set profile id. 'default' by default.
|
| 323 |
*/
|
| 324 |
function import_html_current_profile_id($profile_id = null) {
|
| 325 |
static $import_html_current_profile_id;
|
| 326 |
if (! $import_html_current_profile_id) {
|
| 327 |
$import_html_current_profile_id = variable_get('import_html_current_profile', 'default');
|
| 328 |
}
|
| 329 |
if ($profile_id) {
|
| 330 |
$import_html_current_profile_id = $profile_id;
|
| 331 |
}
|
| 332 |
return $import_html_current_profile_id;
|
| 333 |
}
|
| 334 |
|
| 335 |
/**
|
| 336 |
* Return an array containing the set of import_html settings to use as a template
|
| 337 |
*/
|
| 338 |
function import_html_profile_defaults() {
|
| 339 |
return array(
|
| 340 |
'profile_id' => 'default', // required to be filled in by the system
|
| 341 |
# extraction prefs
|
| 342 |
'translation_template' => drupal_get_path('module', 'import_html') .'/templates/html2simplehtml.xsl',
|
| 343 |
'content_tag_id' => 'main',
|
| 344 |
'content_type' => 'page',
|
| 345 |
'preferred_filter' => import_html_get_preferred_filter(),
|
| 346 |
'default_document' => 'index.htm',
|
| 347 |
'file_exclusions' => "^_\n/_\nCVS\n^\.\n/\.",
|
| 348 |
'strip_tables' => FALSE,
|
| 349 |
'strip_scripts' => FALSE,
|
| 350 |
'force_tidy' => TRUE,
|
| 351 |
#replication options
|
| 352 |
'trim_suffixes' => TRUE,
|
| 353 |
'legacy_aliases' => TRUE,
|
| 354 |
'relink_files' => FALSE,
|
| 355 |
'allow_bad_urls' => FALSE,
|
| 356 |
'file_storage_path' => variable_get('file_directory_path', 'files') ."/imported/",
|
| 357 |
'rewrite_links' => TRUE,
|
| 358 |
'import_site_prefix' => 'imported/',
|
| 359 |
#subsettings
|
| 360 |
'import_category' => "",
|
| 361 |
'import_status' => TRUE,
|
| 362 |
'import_promote' => "",
|
| 363 |
'import_user' => $user->name,
|
| 364 |
'create_menus' => TRUE,
|
| 365 |
'menu_parent_id' => 1,
|
| 366 |
#advanced
|
| 367 |
'handle_duplicates' => IMPORT_HTML_MERGE,
|
| 368 |
'handle_no_title' => IMPORT_HTML_GUESS,
|
| 369 |
'debug_level' => 0,
|
| 370 |
'keep_temp_files' => FALSE,
|
| 371 |
|
| 372 |
);
|
| 373 |
}
|
| 374 |
|
| 375 |
/**
|
| 376 |
* Retrieve a default value for the 'input filter' which applies to nodes even after import.
|
| 377 |
*
|
| 378 |
* This is needed otherwise the edit page defaults to 'filtered' and existing
|
| 379 |
* formatting gets badly stripped.
|
| 380 |
* Normally this is available in the variable import_html_preferred_filter
|
| 381 |
* but the very first time this module is called, the setting may not be
|
| 382 |
* initialized yet. Scan the available filters here and return the appropriate
|
| 383 |
* named filter id.
|
| 384 |
*
|
| 385 |
* @param $as_list bool set this to just get an option list of available filters
|
| 386 |
*/
|
| 387 |
function import_html_get_preferred_filter($as_list = FALSE) {
|
| 388 |
static $fid;
|
| 389 |
if ($fid && !$as_list) return $fid;
|
| 390 |
|
| 391 |
if (($preferred_filter = variable_get('import_html_preferred_filter', 0)) && (!$as_list)) {
|
| 392 |
return $preferred_filter;
|
| 393 |
}
|
| 394 |
|
| 395 |
// As we are importing existing html, we don't even need line breaks done for us,
|
| 396 |
// so the preferred filter is NONE AT ALL - Unfiltered HTML
|
| 397 |
|
| 398 |
// If this module is being enabled as part of an install profile - we cannot be sure that other core modules are even available!
|
| 399 |
if(! function_exists('user_access')) return;
|
| 400 |
|
| 401 |
// Detect the ID of 'Unfiltered HTML' format, if available
|
| 402 |
// it's almost always 3, but I guess we must look it up.
|
| 403 |
$ff = filter_formats();
|
| 404 |
|
| 405 |
$formats = array();
|
| 406 |
foreach ($ff as $f) {
|
| 407 |
$formats[$f->format] = $f->name;
|
| 408 |
if ($f->name == 'Unfiltered HTML') {
|
| 409 |
$fid = $f->format;
|
| 410 |
}
|
| 411 |
}
|
| 412 |
|
| 413 |
if ($as_list) {
|
| 414 |
return $formats;
|
| 415 |
}
|
| 416 |
|
| 417 |
if (!$fid) {
|
| 418 |
// No unfiltered HTML option available... but I need it!;
|
| 419 |
drupal_set_message("Creating new filter option - Unfiltered HTML");
|
| 420 |
|
| 421 |
// can't use filter_admin_add(); as it rewrites the whole page
|
| 422 |
// sorry, direct SQL
|
| 423 |
db_query("INSERT INTO {filter_formats} (name) VALUES ('%s')", 'Unfiltered HTML');
|
| 424 |
$fid = db_result(db_query("SELECT format from {filter_formats} where name='%s'", 'Unfiltered HTML'));
|
| 425 |
}
|
| 426 |
variable_set('import_html_preferred_filter', $fid);
|
| 427 |
return $fid;
|
| 428 |
}
|
| 429 |
|
| 430 |
|
| 431 |
/**
|
| 432 |
* Returns what general 'type' a file probably is, based on suffix or mime if
|
| 433 |
* available.
|
| 434 |
*
|
| 435 |
* @returns one of the defined 'file_classes' : page|image|resource|document ... or null if unknown
|
| 436 |
*
|
| 437 |
* This is mainly used for UI coloring, so is not totally cannonic. HTML-or-not
|
| 438 |
* is all that really counts.
|
| 439 |
*
|
| 440 |
* @see $_import_html_file_classes
|
| 441 |
*/
|
| 442 |
function import_html_guess_file_class($filename) {
|
| 443 |
global $_import_html_file_classes;
|
| 444 |
|
| 445 |
// First try mime
|
| 446 |
if ( function_exists ('mime_content_type') ) {
|
| 447 |
$mime = mime_content_type($filename);
|
| 448 |
list($mime_type, $mime_subtype) = split('/', $mime);
|
| 449 |
if ($mime == 'text/html') return $mime_subtype;
|
| 450 |
if ($mime_type == 'image') return $mime_type;
|
| 451 |
if ($mime == 'text/css') return 'resource';
|
| 452 |
if ($mime == 'application/x-shockwave-flash') return 'resource';
|
| 453 |
if ($mime == 'application/x-javascript') return 'resource';
|
| 454 |
if ($mime_type == 'application') return 'document'; // gross generalization
|
| 455 |
# return 'resource';
|
| 456 |
}
|
| 457 |
|
| 458 |
// Some file mirrors (wget or myself) may have produced odd filenames
|
| 459 |
// strip URL args like # and ? off it
|
| 460 |
$filename = preg_replace('|[\?\#].*$|', '', $filename);
|
| 461 |
|
| 462 |
if (count(explode('.', $filename)) == 1) {
|
| 463 |
// assume no suffix at all is a html page
|
| 464 |
return 'html';
|
| 465 |
}
|
| 466 |
return $_import_html_file_classes[strtolower(array_pop(explode('.', $filename)))];
|
| 467 |
}
|
Parent Directory
| 
Revision Log
| 
Revision Graph