feed.py 文件源码

def read_table(zip_file, table_name, columns, dtypes=None):
    if not dtypes:
        dtypes = {}

    with zip_file.open(table_name + '.txt', 'r') as csv_stream:
        hearer = csv_stream.readline().strip()

        names = [
            GET_COLUMN_NAME_REGEX.sub(b'', name).decode('ascii')
            for name in hearer.split(b',')]

        table = pandas.read_csv(
            csv_stream, names=names, quotechar='"', quoting=csv.QUOTE_ALL,
            usecols=[col for col in columns])
        table = [
            numpy.asarray(remove_nans(table[column]), dtype=dtypes.get(column))
            for column in columns]
        return table