+from pokedex.roomaji import romanize
+
+# Dictionary of table name => table class.
+# Need the table name so we can get the class from the table name after we
+# retrieve something from the index
+indexed_tables = {}
+for cls in [
+ tables.Ability,
+ tables.Item,
+ tables.Move,
+ tables.Pokemon,
+ tables.Type,
+ ]:
+ indexed_tables[cls.__tablename__] = cls
+
+def open_index(directory=None, session=None, recreate=False):
+ """Opens the whoosh index stored in the named directory and returns (index,
+ speller). If the index doesn't already exist, it will be created.
+
+ `directory`
+ Directory containing the index. Defaults to a location within the
+ `pokedex` egg directory.
+
+ `session`
+ If the index needs to be created, this database session will be used.
+ Defaults to an attempt to connect to the default SQLite database
+ installed by `pokedex setup`.
+
+ `recreate`
+ If set to True, the whoosh index will be created even if it already
+ exists.
+ """
+
+ # Defaults
+ if not directory:
+ directory = pkg_resources.resource_filename('pokedex',
+ 'data/whoosh_index')
+
+ if not session:
+ session = connect()
+
+ # Attempt to open or create the index
+ directory_exists = os.path.exists(directory)
+ if directory_exists and not recreate:
+ # Already exists; should be an index!
+ try:
+ index = whoosh.index.open_dir(directory, indexname='MAIN')
+ spell_store = whoosh.filedb.filestore.FileStorage(directory)
+ speller = whoosh.spelling.SpellChecker(spell_store)
+ return index, speller
+ except whoosh.index.EmptyIndexError as e:
+ # Apparently not a real index. Fall out of the if and create it
+ pass
+
+ if not directory_exists:
+ os.mkdir(directory)
+
+
+ # Create index
+ schema = whoosh.fields.Schema(
+ name=whoosh.fields.ID(stored=True),
+ table=whoosh.fields.STORED,
+ row_id=whoosh.fields.ID(stored=True),
+ language=whoosh.fields.STORED,
+ )
+
+ index = whoosh.index.create_in(directory, schema=schema, indexname='MAIN')
+ writer = index.writer()
+
+ # Index every name in all our tables of interest
+ # speller_entries becomes a list of (word, score) tuples; the score is 2
+ # for English names, 1.5 for Roomaji, and 1 for everything else. I think
+ # this biases the results in the direction most people expect, especially
+ # when e.g. German names are very similar to English names
+ speller_entries = []
+ for cls in indexed_tables.values():
+ q = session.query(cls)
+
+ # Only index base Pokémon formes
+ if hasattr(cls, 'forme_base_pokemon_id'):
+ q = q.filter_by(forme_base_pokemon_id=None)
+
+ for row in q.yield_per(5):
+ row_key = dict(table=cls.__tablename__, row_id=unicode(row.id))
+
+ name = row.name.lower()
+ writer.add_document(name=name, **row_key)
+ speller_entries.append((name, 1))
+
+ # Pokemon also get other languages
+ for foreign_name in getattr(row, 'foreign_names', []):
+ moonspeak = foreign_name.name.lower()
+ if name == moonspeak:
+ # Don't add the English name again as a different language;
+ # no point and it makes spell results confusing
+ continue
+
+ writer.add_document(name=moonspeak,
+ language=foreign_name.language.name,
+ **row_key)
+ speller_entries.append((moonspeak, 3))
projects / zzz-pokedex.git / blobdiff