Method test_row_number_offset_in_errors

python/pyarrow/tests/test_csv.py:621–723

(self)

    @pytest.mark.numpy
    def test_row_number_offset_in_errors(self):
        # Row numbers are only correctly counted in serial reads
        def format_msg(msg_format, row, *args):
            if self.use_threads:
                row_info = ""
            else:
                row_info = f"Row #{row}: "
            return msg_format.format(row_info, *args)

        csv, _ = make_random_csv(4, 100, write_names=True)

        read_options = ReadOptions()
        read_options.block_size = len(csv) / 3
        convert_options = ConvertOptions()
        convert_options.column_types = {"a": pa.int32()}

        # Test without skip_rows and column names in the csv
        csv_bad_columns = csv + b"1,2\r\n"
        message_columns = format_msg("{}Expected 4 columns, got 2", 102)
        with pytest.raises(pa.ArrowInvalid, match=message_columns):
            self.read_bytes(csv_bad_columns,
                            read_options=read_options,
                            convert_options=convert_options)

        csv_bad_type = csv + b"a,b,c,d\r\n"
        message_value = format_msg(
            "In CSV column #0: {}"
            "CSV conversion error to int32: invalid value 'a'",
            102, csv)
        with pytest.raises(pa.ArrowInvalid, match=message_value):
            self.read_bytes(csv_bad_type,
                            read_options=read_options,
                            convert_options=convert_options)

        long_row = (b"this is a long row" * 15) + b",3\r\n"
        csv_bad_columns_long = csv + long_row
        message_long = format_msg("{}Expected 4 columns, got 2: {} ...", 102,
                                  long_row[0:96].decode("utf-8"))
        with pytest.raises(pa.ArrowInvalid, match=message_long):
            self.read_bytes(csv_bad_columns_long,
                            read_options=read_options,
                            convert_options=convert_options)

        # Test skipping rows after the names
        read_options.skip_rows_after_names = 47

        with pytest.raises(pa.ArrowInvalid, match=message_columns):
            self.read_bytes(csv_bad_columns,
                            read_options=read_options,
                            convert_options=convert_options)

        with pytest.raises(pa.ArrowInvalid, match=message_value):
            self.read_bytes(csv_bad_type,
                            read_options=read_options,
                            convert_options=convert_options)

        with pytest.raises(pa.ArrowInvalid, match=message_long):
            self.read_bytes(csv_bad_columns_long,