# encoding: utf8
import sys
+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:
from sqlalchemy.orm.attributes import instrumentation_registry
- session = connect(engine_uri)
+ # Use autocommit in case rows fail due to foreign key incest
+ session = connect(engine_uri, autocommit=True, autoflush=False)
metadata.create_all()
- # Oh, mysql-chan.
- # TODO try to insert data in preorder so we don't need this hack and won't
- # break similarly on other engines
- if 'mysql' in engine_uri:
- session.execute('SET FOREIGN_KEY_CHECKS = 0')
-
# SQLAlchemy is retarded and there is no way for me to get a list of ORM
# classes besides to inspect the module they all happen to live in for
# things that look right.
table_base = tables_module.TableBase
- orm_classes = {}
+ orm_classes = {} # table object => table class
for name in dir(tables_module):
# dir() returns strings! How /convenient/.
continue
# thingy is definitely a table class! Hallelujah.
- orm_classes[thingy.__table__.name] = thingy
+ orm_classes[thingy.__table__] = thingy
# Okay, run through the tables and actually load the data now
- for table_name, table in sorted(orm_classes.items()):
+ for table_obj in metadata.sorted_tables:
+ table_class = orm_classes[table_obj]
+ table_name = table_obj.name
+
# Print the table name but leave the cursor in a fixed column
print table_name + '...', ' ' * (40 - len(table_name)),
+ sys.stdout.flush()
try:
csvfile = open("%s/%s.csv" % (directory, table_name), 'rb')
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 = []
+
for csvs in reader:
- row = table()
+ row = table_class()
for column_name, value in zip(column_names, csvs):
- column = table.__table__.c[column_name]
+ column = table_obj.c[column_name]
if column.nullable and value == '':
# Empty string in a nullable column really means NULL
value = None
setattr(row, column_name, value)
- session.add(row)
-
- session.commit()
- print 'loaded'
-
- # Shouldn't matter since this is usually the end of the program and thus
- # the connection too, but let's change this back just in case
- if 'mysql' in engine_uri:
- session.execute('SET FOREIGN_KEY_CHECKS = 1')
-
+ 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='.'):
import csv
writer.writerow(csvs)
+def 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():
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
projects / zzz-pokedex.git / blobdiff