X-Git-Url: http://git.veekun.com/zzz-pokedex.git/blobdiff_plain/4a32d3d0331bfb9cfb3be8d7016859e860bf453e..8cb267bd53921cdcde82e7a8c9ae302d21ef8e26:/pokedex/lookup.py diff --git a/pokedex/lookup.py b/pokedex/lookup.py index cb6d5f6..1ba3fd1 100644 --- a/pokedex/lookup.py +++ b/pokedex/lookup.py @@ -2,6 +2,7 @@ from collections import namedtuple import os, os.path import pkg_resources +import random import re import shutil @@ -18,7 +19,10 @@ from pokedex.db import connect import pokedex.db.tables as tables from pokedex.roomaji import romanize -__all__ = ['open_index', 'lookup'] +__all__ = ['open_index', 'lookup', 'random_lookup'] + +INTERMEDIATE_LOOKUP_RESULTS = 25 +MAX_LOOKUP_RESULTS = 10 # Dictionary of table name => table class. # Need the table name so we can get the class from the table name after we @@ -173,6 +177,24 @@ rx_is_number = re.compile('^\d+$') LookupResult = namedtuple('LookupResult', ['object', 'name', 'language', 'exact']) + +def _parse_table_name(name): + """Takes a singular table name, table name, or table object and returns the + table name. + + Returns None for a bogus name. + """ + if hasattr(name, '__tablename__'): + return getattr(name, '__tablename__') + elif name in indexed_tables: + return name + elif name + 's' in indexed_tables: + return name + 's' + else: + # Bogus. Be nice and return dummy + return None + + def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): """Attempts to find some sort of object, given a database session and name. @@ -184,7 +206,10 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): This function currently ONLY does fuzzy matching if there are no exact matches. - Formes are not returned; "Shaymin" will return only grass Shaymin. + Formes are not returned unless requested; "Shaymin" will return only grass + Shaymin. + + Extraneous whitespace is removed with extreme prejudice. Recognizes: - Names: "Eevee", "Surf", "Run Away", "Payapa Berry", etc. @@ -229,23 +254,34 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): else: index, speller = open_index() - name = unicode(input).lower() + name = unicode(input).strip().lower() exact = True form = None # Remove any type prefix (pokemon:133) before constructing a query if ':' in name: - prefix_chunk, name = name.split(':', 2) - prefixes = prefix_chunk.split(',') + prefix_chunk, name = name.split(':', 1) + name = name.strip() + if not valid_types: # Only use types from the query string if none were explicitly # provided - valid_types = prefixes - - # If the input provided is a number, match it as an id. Otherwise, name. - # Term objects do an exact match, so we don't have to worry about a query - # parser tripping on weird characters in the input - if rx_is_number.match(name): + prefixes = prefix_chunk.split(',') + valid_types = [_.strip() for _ in prefixes] + + # Random lookup + if name == 'random': + return random_lookup(indices=(index, speller), + session=session, + valid_types=valid_types) + + # Do different things depending what the query looks like + # Note: Term objects do an exact match, so we don't have to worry about a + # query parser tripping on weird characters in the input + if '*' in name or '?' in name: + exact_only = True + query = whoosh.query.Wildcard(u'name', name) + elif rx_is_number.match(name): # Don't spell-check numbers! exact_only = True query = whoosh.query.Term(u'row_id', name) @@ -256,7 +292,7 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): # If there's a space in the input, this might be a form if ' ' in name: - form, formless_name = name.split(' ', 2) + form, formless_name = name.split(' ', 1) form_query = whoosh.query.Term(u'name', formless_name) \ & whoosh.query.Term(u'forme_name', form) query = query | form_query @@ -264,16 +300,7 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): ### Filter by type of object type_terms = [] for valid_type in valid_types: - if hasattr(valid_type, '__tablename__'): - table_name = getattr(valid_type, '__tablename__') - elif valid_type in indexed_tables: - table_name = valid_type - elif valid_type + 's' in indexed_tables: - table_name = valid_type + 's' - else: - # Bogus. Be nice and ignore it - continue - + table_name = _parse_table_name(valid_type) type_terms.append(whoosh.query.Term(u'table', table_name)) if type_terms: @@ -285,14 +312,14 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): searcher.weighting = LanguageWeighting() # XXX kosher? docs say search() # takes a weighting kw but it # certainly does not - results = searcher.search(query) + results = searcher.search(query, limit=INTERMEDIATE_LOOKUP_RESULTS) # Look for some fuzzy matches if necessary if not exact_only and not results: exact = False results = [] - for suggestion in speller.suggest(name, 25): + for suggestion in speller.suggest(name, INTERMEDIATE_LOOKUP_RESULTS): query = whoosh.query.Term('name', suggestion) results.extend(searcher.search(query)) @@ -319,4 +346,37 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False): # should have more than 10 here and lost a few. The speller returns 25 to # give us some padding, and should avoid that problem. Not a big deal if # we lose the 25th-most-likely match anyway. - return objects[:10] + return objects[:MAX_LOOKUP_RESULTS] + + +def random_lookup(valid_types=[], session=None, indices=None): + """Takes similar arguments as `lookup()`, but returns a random lookup + result from one of the provided `valid_types`. + """ + + tables = [] + for valid_type in valid_types: + table_name = _parse_table_name(valid_type) + if table_name: + tables.append(indexed_tables[table_name]) + + if not tables: + tables = indexed_tables.values() + + # Rather than create an array of many hundred items and pick randomly from + # it, just pick a number up to the total number of potential items, then + # pick randomly from that, and partition the whole range into chunks + total = 0 + partitions = [] + for table in tables: + count = session.query(table).count() + total += count + partitions.append((table, count)) + + n = random.randint(1, total) + while n > partitions[0][1]: + n -= partitions[0][1] + partitions.pop(0) + + return lookup(unicode(n), valid_types=[ partitions[0][0] ], + indices=indices, session=session)