X-Git-Url: http://git.veekun.com/zzz-pokedex.git/blobdiff_plain/53b6b2d57f58830d7aa3d416f2fa9192f6ff6c99..a04865183fefdd58b0de1cea7c69a9b04a408e53:/pokedex/db/load.py?ds=sidebyside diff --git a/pokedex/db/load.py b/pokedex/db/load.py index df0d0f1..6a548b8 100644 --- a/pokedex/db/load.py +++ b/pokedex/db/load.py @@ -95,7 +95,7 @@ def _get_verbose_prints(verbose): return print_start, print_status, print_done -def load(session, tables=[], directory=None, drop_tables=False, verbose=False): +def load(session, tables=[], directory=None, drop_tables=False, verbose=False, safe=True): """Load data from CSV files into the given database session. Tables are created automatically. @@ -115,6 +115,10 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False): `verbose` If set to True, status messages will be printed to stdout. + + `safe` + If set to False, load can be faster, but can corrupt the database if + it crashes or is interrupted. """ # First take care of verbosity @@ -128,16 +132,24 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False): table_objs = [metadata.tables[name] for name in table_names] table_objs = sqlalchemy.sql.util.sort_tables(table_objs) + # SQLite speed tweaks + if not safe and session.connection().dialect.name == 'sqlite': + session.connection().execute("PRAGMA synchronous=OFF") + session.connection().execute("PRAGMA journal_mode=OFF") # Drop all tables if requested if drop_tables: print_start('Dropping tables') - for table in reversed(table_objs): + for n, table in enumerate(reversed(table_objs)): table.drop(checkfirst=True) + print_status('%s/%s' % (n, len(table_objs))) print_done() - for table in table_objs: + print_start('Creating tables') + for n, table in enumerate(table_objs): table.create() + print_status('%s/%s' % (n, len(table_objs))) + print_done() connection = session.connection() # Okay, run through the tables and actually load the data now @@ -160,6 +172,32 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False): reader = csv.reader(csvfile, lineterminator='\n') column_names = [unicode(column) for column in reader.next()] + if not safe and session.connection().dialect.name == 'postgresql': + """ + Postgres' CSV dialect works with our data, if we mark the not-null + columns with FORCE NOT NULL. + COPY is only allowed for DB superusers. If you're not one, use safe + loading (pokedex load -S). + """ + session.commit() + not_null_cols = [c for c in column_names if not table_obj.c[c].nullable] + if not_null_cols: + force_not_null = 'FORCE NOT NULL ' + ','.join('"%s"' % c for c in not_null_cols) + else: + force_not_null = '' + command = "COPY {table_name} ({columns}) FROM '{csvpath}' CSV HEADER {force_not_null}" + session.connection().execute( + command.format( + table_name=table_name, + csvpath=csvpath, + columns=','.join('"%s"' % c for c in column_names), + force_not_null=force_not_null, + ) + ) + session.commit() + print_done() + continue + # 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 @@ -175,6 +213,8 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False): new_rows = [] def insert_and_commit(): + if not new_rows: + return session.connection().execute(insert_stmt, new_rows) session.commit() new_rows[:] = [] @@ -251,6 +291,10 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False): print_done() + # SQLite check + if session.connection().dialect.name == 'sqlite': + session.connection().execute("PRAGMA integrity_check") + def dump(session, tables=[], directory=None, verbose=False):