Merge pull request #2277 from lonvia/update-osm2pgsql

[nominatim.git] / nominatim / tools / special_phrases.py

diff --git a/nominatim/tools/special_phrases.py b/nominatim/tools/special_phrases.py
index 3faeefbd19680df0833fc9ca41b813e3c83c5c04..0c1258fefb0b9cece4c4e7ca88869b39fd97ab0e 100644 (file)
--- a/nominatim/tools/special_phrases.py
+++ b/nominatim/tools/special_phrases.py
@@ -3,6 +3,7 @@
 """
 import logging
 import os
 """
 import logging
 import os

+from pathlib import Path

 import re
 import subprocess
 import json
 import re
 import subprocess
 import json


@@ -10,6 +11,7 @@ from os.path import isfile
 from icu import Transliterator
 from psycopg2.sql import Identifier, Literal, SQL
 from nominatim.tools.exec_utils import get_url
 from icu import Transliterator
 from psycopg2.sql import Identifier, Literal, SQL
 from nominatim.tools.exec_utils import get_url

+from nominatim.errors import UsageError

 
 LOG = logging.getLogger()
 class SpecialPhrasesImporter():
 
 LOG = logging.getLogger()
 class SpecialPhrasesImporter():


@@ -30,6 +32,17 @@ class SpecialPhrasesImporter():
         self.sanity_check_pattern = re.compile(r'^\w+$')
         self.transliterator = Transliterator.createFromRules("special-phrases normalizer",
                                                              self.config.TERM_NORMALIZATION)
         self.sanity_check_pattern = re.compile(r'^\w+$')
         self.transliterator = Transliterator.createFromRules("special-phrases normalizer",
                                                              self.config.TERM_NORMALIZATION)

+        #This set will contain all existing phrases from the word table which
+        #no longer exist on the wiki.
+        #It contain tuples with the following format: (normalized_word, class, type, operator)
+        self.words_phrases_to_delete = set()
+        #This set will contain the phrases which still exist from the wiki.
+        #It is used to prevent duplicates on the wiki by removing them from
+        #the word_phrases_to_delete only at the end.
+        self.words_phrases_still_exist = set()
+        #This set will contain all existing place_classtype tables which doesn't match any
+        #special phrases class/type on the wiki.
+        self.table_phrases_to_delete = set()

 
     def import_from_wiki(self, languages=None):
         """
 
     def import_from_wiki(self, languages=None):
         """


@@ -37,28 +50,68 @@ class SpecialPhrasesImporter():
             extract corresponding special phrases from the wiki.
         """
         if languages is not None and not isinstance(languages, list):
             extract corresponding special phrases from the wiki.
         """
         if languages is not None and not isinstance(languages, list):

-            raise TypeError('languages argument should be of type list')
+            raise TypeError('The \'languages\' argument should be of type list.')
+
+        self._fetch_existing_words_phrases()
+        self._fetch_existing_place_classtype_tables()

 
         #Get all languages to process.
         languages = self._load_languages() if not languages else languages
 
 
         #Get all languages to process.
         languages = self._load_languages() if not languages else languages
 

-        #array for pairs of class/type
+        #Store pairs of class/type for further processing

         class_type_pairs = set()
 
         for lang in languages:
             LOG.warning('Import phrases for lang: %s', lang)
             wiki_page_xml_content = SpecialPhrasesImporter._get_wiki_content(lang)
         class_type_pairs = set()
 
         for lang in languages:
             LOG.warning('Import phrases for lang: %s', lang)
             wiki_page_xml_content = SpecialPhrasesImporter._get_wiki_content(lang)

-            self._process_xml_content(wiki_page_xml_content, lang)
+            class_type_pairs.update(self._process_xml_content(wiki_page_xml_content, lang))

 
         self._create_place_classtype_table_and_indexes(class_type_pairs)
 
         self._create_place_classtype_table_and_indexes(class_type_pairs)

+        self._remove_non_existent_phrases_from_db()

         self.db_connection.commit()
         LOG.warning('Import done.')
 
         self.db_connection.commit()
         LOG.warning('Import done.')
 

+    def _fetch_existing_words_phrases(self):
+        """
+            Fetch existing special phrases from the word table.
+            Fill the word_phrases_to_delete set of the class.
+        """
+        #Only extract special phrases terms:
+        #If class=place and type=house then it is a housenumber term.
+        #If class=place and type=postcode then it is a postcode term.
+        word_query = """
+            SELECT word, class, type, operator FROM word
+            WHERE class != 'place' OR (type != 'house' AND type != 'postcode')
+        """
+        with self.db_connection.cursor() as db_cursor:
+            db_cursor.execute(SQL(word_query))
+            for row in db_cursor:
+                row[3] = '-' if row[3] is None else row[3]
+                self.words_phrases_to_delete.add(
+                    (row[0], row[1], row[2], row[3])
+                )
+
+    def _fetch_existing_place_classtype_tables(self):
+        """
+            Fetch existing place_classtype tables.
+            Fill the table_phrases_to_delete set of the class.
+        """
+        query = """
+            SELECT table_name
+            FROM information_schema.tables
+            WHERE table_schema='public'
+            AND table_name like 'place_classtype_%';
+        """
+        with self.db_connection.cursor() as db_cursor:
+            db_cursor.execute(SQL(query))
+            for row in db_cursor:
+                self.table_phrases_to_delete.add(row[0])
+

     def _load_white_and_black_lists(self):
         """
             Load white and black lists from phrases-settings.json.
         """
     def _load_white_and_black_lists(self):
         """
             Load white and black lists from phrases-settings.json.
         """

-        settings_path = str(self.config.config_dir)+'/phrase-settings.json'
+        settings_path = (self.config.config_dir / 'phrase-settings.json').resolve()

 
         if self.config.PHRASE_CONFIG:
             settings_path = self._convert_php_settings_if_needed(self.config.PHRASE_CONFIG)
 
         if self.config.PHRASE_CONFIG:
             settings_path = self._convert_php_settings_if_needed(self.config.PHRASE_CONFIG)


@@ -78,7 +131,7 @@ class SpecialPhrasesImporter():
             'et', 'eu', 'fa', 'fi', 'fr', 'gl', 'hr', 'hu',
             'ia', 'is', 'it', 'ja', 'mk', 'nl', 'no', 'pl',
             'ps', 'pt', 'ru', 'sk', 'sl', 'sv', 'uk', 'vi']
             'et', 'eu', 'fa', 'fi', 'fr', 'gl', 'hr', 'hu',
             'ia', 'is', 'it', 'ja', 'mk', 'nl', 'no', 'pl',
             'ps', 'pt', 'ru', 'sk', 'sl', 'sv', 'uk', 'vi']

-        return self.config.LANGUAGES or default_languages
+        return self.config.LANGUAGES.split(',') if self.config.LANGUAGES else default_languages

 
     @staticmethod
     def _get_wiki_content(lang):
 
     @staticmethod
     def _get_wiki_content(lang):


@@ -100,11 +153,18 @@ class SpecialPhrasesImporter():
         class_matchs = self.sanity_check_pattern.findall(phrase_class)
 
         if len(class_matchs) < 1 or len(type_matchs) < 1:
         class_matchs = self.sanity_check_pattern.findall(phrase_class)
 
         if len(class_matchs) < 1 or len(type_matchs) < 1:

-            LOG.error("Bad class/type for language %s: %s=%s", lang, phrase_class, phrase_type)
+            raise UsageError("Bad class/type for language {}: {}={}".format(
+                lang, phrase_class, phrase_type))

 
     def _process_xml_content(self, xml_content, lang):
 
     def _process_xml_content(self, xml_content, lang):

+        """
+            Process given xml content by extracting matching patterns.
+            Matching patterns are processed there and returned in a
+            set of class/type pairs.
+        """

         #One match will be of format [label, class, type, operator, plural]
         matches = self.occurence_pattern.findall(xml_content)
         #One match will be of format [label, class, type, operator, plural]
         matches = self.occurence_pattern.findall(xml_content)

+        #Store pairs of class/type for further processing

         class_type_pairs = set()
 
         for match in matches:
         class_type_pairs = set()
 
         for match in matches:


@@ -113,12 +173,11 @@ class SpecialPhrasesImporter():
             phrase_class = match[1].strip()
             phrase_type = match[2].strip()
             phrase_operator = match[3].strip()
             phrase_class = match[1].strip()
             phrase_type = match[2].strip()
             phrase_operator = match[3].strip()

+            #Needed if some operator in the wiki are not written in english
+            phrase_operator = '-' if phrase_operator not in ('near', 'in') else phrase_operator

             #hack around a bug where building=yes was imported with quotes into the wiki
             phrase_type = re.sub(r'\"|&quot;', '', phrase_type)
 
             #hack around a bug where building=yes was imported with quotes into the wiki
             phrase_type = re.sub(r'\"|&quot;', '', phrase_type)
 

-            #sanity check, in case somebody added garbage in the wiki
-            self._check_sanity(lang, phrase_class, phrase_type)
-

             #blacklisting: disallow certain class/type combinations
             if (
                     phrase_class in self.black_list.keys() and
             #blacklisting: disallow certain class/type combinations
             if (
                     phrase_class in self.black_list.keys() and


@@ -132,7 +191,22 @@ class SpecialPhrasesImporter():
             ):
                 continue
 
             ):
                 continue
 

-            #add class/type to the pairs dict
+            #Check if the phrase already exists in the database.
+            if (
+                    (normalized_label, phrase_class, phrase_type, phrase_operator)
+                    in self.words_phrases_to_delete
+            ):
+                #Remove this phrase from the ones to delete as it still exist on the wiki.
+                self.words_phrases_still_exist.add(
+                    (normalized_label, phrase_class, phrase_type, phrase_operator)
+                )
+                class_type_pairs.add((phrase_class, phrase_type))
+                #Dont need to add this phrase as it already exists in the word table.
+                continue
+
+            #sanity check, in case somebody added garbage in the wiki
+            self._check_sanity(lang, phrase_class, phrase_type)
+

             class_type_pairs.add((phrase_class, phrase_type))
 
             self._process_amenity(
             class_type_pairs.add((phrase_class, phrase_type))
 
             self._process_amenity(


@@ -182,6 +256,15 @@ class SpecialPhrasesImporter():
             phrase_class = pair[0]
             phrase_type = pair[1]
 
             phrase_class = pair[0]
             phrase_type = pair[1]
 

+            table_name = 'place_classtype_{}_{}'.format(phrase_class, phrase_type)
+
+            if table_name in self.table_phrases_to_delete:
+                #Remove this table from the ones to delete as it match a class/type
+                #still existing on the special phrases of the wiki.
+                self.table_phrases_to_delete.remove(table_name)
+                #So dont need to create the table and indexes.
+                continue
+

             #Table creation
             self._create_place_classtype_table(sql_tablespace, phrase_class, phrase_type)
 
             #Table creation
             self._create_place_classtype_table(sql_tablespace, phrase_class, phrase_type)
 


@@ -242,16 +325,58 @@ class SpecialPhrasesImporter():
                               .format(Identifier(table_name),
                                       Identifier(self.config.DATABASE_WEBUSER)))
 
                               .format(Identifier(table_name),
                                       Identifier(self.config.DATABASE_WEBUSER)))
 

+    def _remove_non_existent_phrases_from_db(self):
+        """
+            Remove special phrases which doesn't exist on the wiki anymore.
+            Delete from the word table and delete the place_classtype tables.
+        """
+        LOG.warning('Cleaning database...')
+        self.words_phrases_to_delete = self.words_phrases_to_delete - self.words_phrases_still_exist
+        #Array containing all queries to execute. Contain tuples of format (query, parameters)
+        queries_parameters = []
+
+        #Delete phrases from the word table which are not on the wiki anymore.
+        for phrase_to_delete in self.words_phrases_to_delete:
+            if phrase_to_delete[3] == '-':
+                query = """
+                    DELETE FROM word WHERE word = %s AND class = %s AND type = %s AND operator IS null
+                """
+                parameters = (phrase_to_delete[0], phrase_to_delete[1], phrase_to_delete[2], )
+                queries_parameters.append((query, parameters))
+            else:
+                query = """
+                    DELETE FROM word WHERE word = %s AND class = %s AND type = %s AND operator = %s
+                """
+                parameters = (phrase_to_delete[0], phrase_to_delete[1],
+                              phrase_to_delete[2], phrase_to_delete[3], )
+                queries_parameters.append((query, parameters))
+
+        #Delete place_classtype tables corresponding to class/type which are not on the wiki anymore
+        for table in self.table_phrases_to_delete:
+            query = SQL('DROP TABLE IF EXISTS {}').format(Identifier(table))
+            queries_parameters.append((query, ()))
+
+        with self.db_connection.cursor() as db_cursor:
+            for query, parameters in queries_parameters:
+                db_cursor.execute(query, parameters)
+

     def _convert_php_settings_if_needed(self, file_path):
         """
             Convert php settings file of special phrases to json file if it is still in php format.
         """
     def _convert_php_settings_if_needed(self, file_path):
         """
             Convert php settings file of special phrases to json file if it is still in php format.
         """

+        if not isfile(file_path):
+            raise UsageError(str(file_path) + ' is not a valid file.')
+

         file, extension = os.path.splitext(file_path)
         file, extension = os.path.splitext(file_path)

-        json_file_path = file + '.json'
+        json_file_path = Path(file + '.json').resolve()
+
+        if extension not in('.php', '.json'):
+            raise UsageError('The custom NOMINATIM_PHRASE_CONFIG file has not a valid extension.')
+

         if extension == '.php' and not isfile(json_file_path):
             try:
                 subprocess.run(['/usr/bin/env', 'php', '-Cq',
         if extension == '.php' and not isfile(json_file_path):
             try:
                 subprocess.run(['/usr/bin/env', 'php', '-Cq',

-                                self.phplib_dir / 'migration/phraseSettingsToJson.php',
+                                (self.phplib_dir / 'migration/PhraseSettingsToJson.php').resolve(),

                                 file_path], check=True)
                 LOG.warning('special_phrase configuration file has been converted to json.')
                 return json_file_path
                                 file_path], check=True)
                 LOG.warning('special_phrase configuration file has been converted to json.')
                 return json_file_path