X-Git-Url: http://git.veekun.com/zzz-pokedex.git/blobdiff_plain/6af0c42ed6234d8db295b8bf3811fbacac75dbc6..9d6aa28bca0bae50c58993a55f5b586dbadff55e:/pokedex/db/load.py diff --git a/pokedex/db/load.py b/pokedex/db/load.py index 1cff39d..2b14883 100644 --- a/pokedex/db/load.py +++ b/pokedex/db/load.py @@ -10,7 +10,41 @@ from pokedex.db import metadata import pokedex.db.tables as tables -def load(session, directory=None, drop_tables=False): +def _get_verbose_prints(verbose): + """If `verbose` is true, returns two functions: one for printing a starting + message, and the other for printing a success or failure message when + finished. + + If `verbose` is false, returns two no-op functions. + """ + + if verbose: + import sys + def print_start(thing): + # Truncate to 66 characters, leaving 10 characters for a success + # or failure message + truncated_thing = thing[0:66] + + # Also, space-pad to keep the cursor in a known column + num_spaces = 66 - len(truncated_thing) + + print "%s...%s" % (truncated_thing, ' ' * num_spaces), + sys.stdout.flush() + + def print_done(msg='ok'): + print msg + sys.stdout.flush() + + return print_start, print_done + + # Not verbose; return dummies + def dummy(*args, **kwargs): + pass + + return dummy, dummy + + +def load(session, directory=None, drop_tables=False, verbose=False): """Load data from CSV files into the given database session. Tables are created automatically. @@ -24,55 +58,39 @@ def load(session, directory=None, drop_tables=False): `drop_tables` If set to True, existing `pokedex`-related tables will be dropped. + + `verbose` + If set to True, status messages will be printed to stdout. """ + # First take care of verbosity + print_start, print_done = _get_verbose_prints(verbose) + + if not directory: directory = pkg_resources.resource_filename('pokedex', 'data/csv') # Drop all tables if requested - if options.drop_tables: - print 'Dropping tables...' + if drop_tables: + print_start('Dropping tables') metadata.drop_all() + print_done() metadata.create_all() - - # 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.TableBase - orm_classes = {} # table object => table class - - for name in dir(tables): - # dir() returns strings! How /convenient/. - thingy = getattr(tables, name) - - if not isinstance(thingy, type): - # Not a class; bail - continue - elif not issubclass(thingy, table_base): - # Not a declarative table; bail - continue - elif thingy == table_base: - # Declarative table base, so not a real table; bail - continue - - # thingy is definitely a table class! Hallelujah. - orm_classes[thingy.__table__] = thingy + connection = session.connection() # Okay, run through the tables and actually load the data now for table_obj in metadata.sorted_tables: - table_class = orm_classes[table_obj] table_name = table_obj.name + insert_stmt = table_obj.insert() - # Print the table name but leave the cursor in a fixed column - print table_name + '...', ' ' * (40 - len(table_name)), - sys.stdout.flush() + print_start(table_name) try: csvfile = open("%s/%s.csv" % (directory, table_name), 'rb') except IOError: # File doesn't exist; don't load anything! - print 'no data!' + print_done('missing?') continue reader = csv.reader(csvfile, lineterminator='\n') @@ -91,8 +109,14 @@ def load(session, directory=None, drop_tables=False): if any(_.references(table_obj) for _ in column.foreign_keys): self_ref_columns.append(column) + new_rows = [] + def insert_and_commit(): + session.connection().execute(insert_stmt, new_rows) + session.commit() + new_rows[:] = [] + for csvs in reader: - row = table_class() + row_data = {} for column_name, value in zip(column_names, csvs): column = table_obj.c[column_name] @@ -110,50 +134,59 @@ def load(session, directory=None, drop_tables=False): # Otherwise, unflatten from bytes value = value.decode('utf-8') - setattr(row, column_name, value) + # nb: Dictionaries flattened with ** have to have string keys + row_data[ str(column_name) ] = value # 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 = [row_data[_.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 + seen_ids[row_data['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 + insert_and_commit() + seen_ids[row_data['id']] = 1 else: # Non-NULL future id. Save this and insert it later! - deferred_rows.append((row, foreign_ids)) + deferred_rows.append((row_data, foreign_ids)) continue - session.add(row) + # Insert row! + new_rows.append(row_data) - session.commit() + # Remembering some zillion rows in the session consumes a lot of + # RAM. Let's not do that. Commit every 1000 rows + if len(new_rows) > 1000: + insert_and_commit() + + insert_and_commit() # Attempt to add any spare rows we've collected - for row, foreign_ids in deferred_rows: + for row_data, 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__)) + "Row was: " + str(row)) - session.add(row) - seen_ids[row.id] = 1 - session.commit() + session.connection().execute( + insert_stmt.values(**row_data) + ) + seen_ids[row_data['id']] = 1 + session.commit() - print 'loaded' + print_done() -def dump(session, directory=None): +def dump(session, directory=None, verbose=False): """Dumps the contents of a database to a set of CSV files. Probably not useful to anyone besides a developer. @@ -163,13 +196,20 @@ def dump(session, directory=None): `directory` Directory the CSV files should be put in. Defaults to the `pokedex` data directory. + + `verbose` + If set to True, status messages will be printed to stdout. """ + # First take care of verbosity + print_start, print_done = _get_verbose_prints(verbose) + + if not directory: directory = pkg_resources.resource_filename('pokedex', 'data/csv') for table_name in sorted(metadata.tables.keys()): - print table_name + print_start(table_name) table = metadata.tables[table_name] writer = csv.writer(open("%s/%s.csv" % (directory, table_name), 'wb'), @@ -195,3 +235,5 @@ def dump(session, directory=None): csvs.append(val) writer.writerow(csvs) + + print_done()