X-Git-Url: http://git.veekun.com/zzz-pokedex.git/blobdiff_plain/2c0980e4dcadaa31d7ea7ebab174a6fd81723090..ac52b197ed6dfeb5d7ad8cb2e8f0d5b9f1cdd1d4:/pokedex/__init__.py

diff --git a/pokedex/__init__.py b/pokedex/__init__.py
index 02b8ce5..1e6c6d1 100644
--- a/pokedex/__init__.py
+++ b/pokedex/__init__.py
@@ -5,6 +5,7 @@ from sqlalchemy.exc import IntegrityError
 import sqlalchemy.types
 
 from .db import connect, metadata, tables as tables_module
+from pokedex.lookup import lookup as pokedex_lookup
 
 def main():
     if len(sys.argv) <= 1:
@@ -14,20 +15,19 @@ def main():
     args = sys.argv[2:]
 
     # Find the command as a function in this file
-    func = globals().get(command, None)
-    if func and callable(func) and command != 'main':
+    func = globals().get("command_%s" % command, None)
+    if func:
         func(*args)
     else:
-        help()
+        command_help()
 
 
-def csvimport(engine_uri, directory='.'):
+def command_csvimport(engine_uri, directory='.'):
     import csv
 
     from sqlalchemy.orm.attributes import instrumentation_registry
 
-    # Use autocommit in case rows fail due to foreign key incest
-    session = connect(engine_uri, autocommit=True, autoflush=False)
+    session = connect(engine_uri)
 
     metadata.create_all()
 
@@ -73,11 +73,18 @@ def csvimport(engine_uri, directory='.'):
         reader = csv.reader(csvfile, lineterminator='\n')
         column_names = [unicode(column) for column in reader.next()]
 
-        # Self-referential tables may contain rows with foreign keys of
-        # other rows in the same table that do not yet exist.  We'll keep
-        # a running list of these and try inserting them again after the
-        # rest are done
-        failed_rows = []
+        # Self-referential tables may contain rows with foreign keys of other
+        # rows in the same table that do not yet exist.  Pull these out and add
+        # them to the session last
+        # ASSUMPTION: Self-referential tables have a single PK called "id"
+        deferred_rows = []  # ( row referring to id, [foreign ids we need] )
+        seen_ids = {}       # primary key we've seen => 1
+
+        # Fetch foreign key columns that point at this table, if any
+        self_ref_columns = []
+        for column in table_obj.c:
+            if any(_.references(table_obj) for _ in column.foreign_keys):
+                self_ref_columns.append(column)
 
         for csvs in reader:
             row = table_class()
@@ -100,35 +107,46 @@ def csvimport(engine_uri, directory='.'):
 
                 setattr(row, column_name, value)
 
-            try:
-                session.add(row)
-                session.flush()
-            except IntegrityError as e:
-                failed_rows.append(row)
-
-        # Loop over the failed rows and keep trying to insert them.  If a loop
-        # doesn't manage to insert any rows, bail.
-        do_another_loop = True
-        while failed_rows and do_another_loop:
-            do_another_loop = False
-
-            for i, row in enumerate(failed_rows):
-                try:
-                    session.add(row)
-                    session.flush()
-
-                    # Success!
-                    del failed_rows[i]
-                    do_another_loop = True
-                except IntegrityError as e:
-                    pass
-
-        if failed_rows:
-            print len(failed_rows), "rows failed"
-        else:
-            print 'loaded'
-
-def csvexport(engine_uri, directory='.'):
+            # May need to stash this row and add it later if it refers to a
+            # later row in this table
+            if self_ref_columns:
+                foreign_ids = [getattr(row, _.name) for _ in self_ref_columns]
+                foreign_ids = [_ for _ in foreign_ids if _]  # remove NULL ids
+
+                if not foreign_ids:
+                    # NULL key.  Remember this row and add as usual.
+                    seen_ids[row.id] = 1
+
+                elif all(_ in seen_ids for _ in foreign_ids):
+                    # Non-NULL key we've already seen.  Remember it and commit
+                    # so we know the old row exists when we add the new one
+                    session.commit()
+                    seen_ids[row.id] = 1
+
+                else:
+                    # Non-NULL future id.  Save this and insert it later!
+                    deferred_rows.append((row, foreign_ids))
+                    continue
+
+            session.add(row)
+
+        session.commit()
+
+        # Attempt to add any spare rows we've collected
+        for row, foreign_ids in deferred_rows:
+            if not all(_ in seen_ids for _ in foreign_ids):
+                # Could happen if row A refers to B which refers to C.
+                # This is ridiculous and doesn't happen in my data so far
+                raise ValueError("Too many levels of self-reference!  "
+                                 "Row was: " + str(row.__dict__))
+
+            session.add(row)
+            seen_ids[row.id] = 1
+            session.commit()
+
+        print 'loaded'
+
+def command_csvexport(engine_uri, directory='.'):
     import csv
     session = connect(engine_uri)
 
@@ -141,7 +159,8 @@ def csvexport(engine_uri, directory='.'):
         columns = [col.name for col in table.columns]
         writer.writerow(columns)
 
-        for row in session.query(table).all():
+        primary_key = table.primary_key
+        for row in session.query(table).order_by(*primary_key).all():
             csvs = []
             for col in columns:
                 # Convert Pythony values to something more universal
@@ -159,11 +178,25 @@ def csvexport(engine_uri, directory='.'):
 
             writer.writerow(csvs)
 
+def command_lookup(engine_uri, name):
+    # XXX don't require uri!  somehow
+    session = connect(engine_uri)
+
+    results, exact = pokedex_lookup(session, name)
+    if exact:
+        print "Matched:"
+    else:
+        print "Fuzzy-matched:"
+
+    for object in results:
+        print object.__tablename__, object.name
+
 
-def help():
+def command_help():
     print u"""pokedex -- a command-line Pokédex interface
 
     help                        Displays this message.
+    lookup {uri} [name]         Look up something in the Pokédex.
 
   These commands are only useful for developers:
     csvimport {uri} [dir]       Import data from a set of CSVs to the database
@@ -171,6 +204,6 @@ def help():
     csvexport {uri} [dir]       Export data from the database given by the URI
                                   to a set of CSVs.
                                 Directory defaults to cwd.
-"""
+""".encode(sys.getdefaultencoding(), 'replace')
 
     sys.exit(0)