Changeset 5417 for sites/trunk/api.wordpress.org/public_html/events/1.0/index.php

Timestamp:

04/27/2017 06:11:16 PM (7 years ago)

Author:

iandunn

Message:

Events: Refactor the ideographic fallback for use with ASCII queries too

This was originally intended for ideographic languages, but there are new edge cases where it is helpful for ASCII queries as well.

See ​https://github.com/coreymckrill/nearby-wordpress-events/issues/37

File:

• sites/trunk/api.wordpress.org/public_html/events/1.0/index.php (modified) (8 diffs)

sites/trunk/api.wordpress.org/public_html/events/1.0/index.php

@@ -212 +212 @@
 
     if ( ! is_a( $row, 'stdClass' ) && 'ASCII' !== mb_detect_encoding( $location_name ) ) {
-        $row = guess_ideographic_location_from_geonames( $location_name, $country, $timezone );
+        $row = guess_location_from_geonames_fallback( $location_name, $country, $timezone, 'exact', 'ideographic' );
     }
 
@@ -219 +219 @@
 
 /**
- * Look for the given ideographic location in the Geonames database
+ * Look for the given location in the Geonames database using a LIKE query
  *
  * This is a fallback for situations where the full-text search in `guess_location_from_geonames()` resulted
- * in a false-negative. MySQL < 5.7.6 doesn't support full-text searches on ideographic languages, because
- * it cannot determine where the word boundaries are.
- *
+ * in a false-negative.
+ *
+ * One situation where this happens is with queries in ideographic languages, because MySQL < 5.7.6 doesn't
+ * support full-text searches for them, because it can't determine where the word boundaries are.
  * See https://dev.mysql.com/doc/refman/5.7/en/fulltext-restrictions.html
+ *
+ * There are also edge cases where the exact query doesn't exist in the database, but a loose LIKE query will find
+ * a similar alternate, like `Osakashi`.
  *
  * @param string $location_name
@@ -232 +236 @@
  * @param string $mode          'exact' to only return exact matches from the database;
  *                              'loose' to return any match. This has a high chance of false positives.
+ * @param string $restrict_counties 'ideographic' to only search in countries where ideographic languages are common;
+ *                                  'none' to search all countries
  *
  * @return stdClass|null
  */
-function guess_ideographic_location_from_geonames( $location_name, $country, $timezone, $mode = 'exact' ) {
+function guess_location_from_geonames_fallback( $location_name, $country, $timezone, $mode = 'exact', $restrict_counties = 'ideographic' ) {
     global $wpdb;
 
-    $ideographic_countries            = get_ideographic_counties();
-    $ideographic_country_placeholders = get_prepare_placeholders( count( $ideographic_countries ), '%s' );
+    $where = $ideographic_countries = $ideographic_country_placeholders = '';
 
     /*
@@ -256 +261 @@
     );
 
+    $prepare_args = array( $escaped_location_name, $country, $timezone );
+
+    if ( 'ideographic' == $restrict_counties ) {
+        $ideographic_countries            = get_ideographic_counties();
+        $ideographic_country_placeholders = get_prepare_placeholders( count( $ideographic_countries ), '%s' );
+
+        $where .= "country IN ( $ideographic_country_placeholders ) AND";
+
+        $prepare_args = array_merge( $ideographic_countries, $prepare_args );
+    }
+
     /*
      * REPLACE() is used because sometimes the `alternatenames` column contains entries where the `asciiname` is
@@ -270 +286 @@
         FROM `geoname`
         WHERE
-            country IN ( $ideographic_country_placeholders ) AND
+            $where
             REPLACE( alternatenames, CONCAT( asciiname, ' - ' ), '' ) LIKE %s
         ORDER BY
@@ -278 +294 @@
         LIMIT 1";
 
-    $prepared_query = $wpdb->prepare(
-        $query,
-        array_merge( $ideographic_countries, array( $escaped_location_name, $country, $timezone ) )
-    );
+    $prepared_query = $wpdb->prepare( $query, $prepare_args );
 
     return $wpdb->get_row( $prepared_query );
@@ -444 +457 @@
 
     /*
-     * If all else fails for a non-ASCII request, cast a wide net and try to find something before giving up, even
+     * If all else fails, cast a wide net and try to find something before giving up, even
      * if the chance of success if lower than normal. Returning false is guaranteed failure, so this improves things
      * even if it only works 10% of the time.
@@ -450 +463 @@
      * This must be done as the very last thing before giving up, because the likelihood of false positives is high.
      */
-    if ( ! $location && isset( $args['location_name'] ) && 'ASCII' !== mb_detect_encoding( $args['location_name'] ) ) {
-        $guess = guess_ideographic_location_from_geonames( $args['location_name'], $country_code, $args['timezone'] ?? '', 'loose' );
+    if ( ! $location && isset( $args['location_name'] ) ) {
+        if ( 'ASCII' === mb_detect_encoding( $args['location_name'] ) ) {
+            $guess = guess_location_from_geonames_fallback( $args['location_name'], $country_code, $args['timezone'] ?? '', 'loose', 'none' );
+        } else {
+            $guess = guess_location_from_geonames_fallback( $args['location_name'], $country_code, $args['timezone'] ?? '', 'loose', 'ideographic' );
+        }
 
         if ( $guess ) {