X-Git-Url: http://git.veekun.com/zzz-pokedex.git/blobdiff_plain/c0c111bb674c04565fc86c50435c0ce0fb0ca3f3..16dd9d8ac11b990125eb836d4d3f7b62c4b1c712:/pokedex/db/load.py diff --git a/pokedex/db/load.py b/pokedex/db/load.py index a6e9d4a..f0e4b6d 100644 --- a/pokedex/db/load.py +++ b/pokedex/db/load.py @@ -174,14 +174,10 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s if not safe and session.connection().dialect.name == 'postgresql': """ - Postgres' CSV dialect is nearly the same as ours, except that it - treats completely empty values as NULL, and empty quoted - strings ("") as an empty strings. - Pokedex dump does not quote empty strings. So, both empty strings - and NULLs are read in as NULL. - For an empty string in a NOT NULL column, the load will fail, and - load will fall back to the cross-backend row-by-row loading. And in - nullable columns, we already load empty stings as NULL. + 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] @@ -189,15 +185,15 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s 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}" + command = "COPY %(table_name)s (%(columns)s) FROM '%(csvpath)s' CSV HEADER %(force_not_null)s" 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, - ) + command % dict( + 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 @@ -217,6 +213,8 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s new_rows = [] def insert_and_commit(): + if not new_rows: + return session.connection().execute(insert_stmt, new_rows) session.commit() new_rows[:] = []