Lookup no longer crashes if all type prefixes are bogus.

[zzz-pokedex.git] / pokedex / lookup.py

diff --git a/pokedex/lookup.py b/pokedex/lookup.py
index 28dd4b2..60a6a56 100644 (file)
--- a/pokedex/lookup.py
+++ b/pokedex/lookup.py
@@ -94,6 +94,7 @@ def open_index(directory=None, session=None, recreate=False):
         table=whoosh.fields.ID(stored=True),
         row_id=whoosh.fields.ID(stored=True),
         language=whoosh.fields.STORED,
+        iso3166=whoosh.fields.STORED,
         display_name=whoosh.fields.STORED,  # non-lowercased name
         forme_name=whoosh.fields.ID,
     )
@@ -117,9 +118,10 @@ def open_index(directory=None, session=None, recreate=False):
                            row_id=unicode(row.id),
                            forme_name=u'XXX')
 
-            def add(name, language, score):
+            def add(name, language, iso3166, score):
                 writer.add_document(name=name.lower(), display_name=name,
                                     language=language,
+                                    iso3166=iso3166,
                                     **row_key)
                 speller_entries.append((name.lower(), score))
 
@@ -128,7 +130,7 @@ def open_index(directory=None, session=None, recreate=False):
                 row_key['forme_name'] = row.forme_name
 
             name = row.name
-            add(name, None, 1)
+            add(name, None, u'us', 1)
 
             # Pokemon also get other languages
             for foreign_name in getattr(row, 'foreign_names', []):
@@ -138,12 +140,14 @@ def open_index(directory=None, session=None, recreate=False):
                     # no point and it makes spell results confusing
                     continue
 
-                add(moonspeak, foreign_name.language.name, 3)
+                add(moonspeak, foreign_name.language.name,
+                               foreign_name.language.iso3166,
+                               3)
 
                 # Add Roomaji too
                 if foreign_name.language.name == 'Japanese':
                     roomaji = romanize(foreign_name.name)
-                    add(roomaji, u'Roomaji', 8)
+                    add(roomaji, u'Roomaji', u'jp', 8)
 
     writer.commit()
 
@@ -176,7 +180,7 @@ class LanguageWeighting(whoosh.scoring.Weighting):
 rx_is_number = re.compile('^\d+$')
 
 LookupResult = namedtuple('LookupResult',
-                          ['object', 'name', 'language', 'exact'])
+                          ['object', 'name', 'language', 'iso3166', 'exact'])
 
 def _parse_table_name(name):
     """Takes a singular table name, table name, or table object and returns the
@@ -215,6 +219,7 @@ def _whoosh_records_to_results(records, session, exact=True):
         results.append(LookupResult(object=obj,
                                     name=record['display_name'],
                                     language=record['language'],
+                                    iso3166=record['iso3166'],
                                     exact=exact))
 
     return results
@@ -223,10 +228,11 @@ def _whoosh_records_to_results(records, session, exact=True):
 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.
 
-    Returns a list of named (object, name, language, exact) tuples.  `object`
-    is a database object, `name` is the name under which the object was found,
-    `language` is the name of the language in which the name was found, and
-    `exact` is True iff this was an exact match.
+    Returns a list of named (object, name, language, iso3166, exact) tuples.
+    `object` is a database object, `name` is the name under which the object
+    was found, `language` and `iso3166` are the name and country code of the
+    language in which the name was found, and `exact` is True iff this was an
+    exact match.
 
     This function currently ONLY does fuzzy matching if there are no exact
     matches.
@@ -326,7 +332,9 @@ def lookup(input, valid_types=[], session=None, indices=None, exact_only=False):
     type_terms = []
     for valid_type in valid_types:
         table_name = _parse_table_name(valid_type)
-        type_terms.append(whoosh.query.Term(u'table', table_name))
+        if table_name:
+            # Quietly ignore bogus valid_types; more likely to DTRT
+            type_terms.append(whoosh.query.Term(u'table', table_name))
 
     if type_terms:
         query = query & whoosh.query.Or(type_terms)
@@ -371,11 +379,17 @@ def random_lookup(valid_types=[], session=None, indices=None):
             tables.append(indexed_tables[table_name])
 
     if not tables:
+        # n.b.: It's possible we got a list of valid_types and none of them
+        # were valid, but this function is guaranteed to return *something*, so
+        # it politely selects from the entire index isntead
         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
+    # pick randomly from that, and partition the whole range into chunks.
+    # This also avoids the slight problem that the index contains more rows
+    # (for languages) for some items than others.
+    # XXX ought to cache this (in the index?) if possible
     total = 0
     partitions = []
     for table in tables: