I’m reading data from the BigQuery as TableRows, then convert them to Rows. This step takes a lot of time. Is it possible to read Rows from the beginning? Or any other ideas of how to make it faster?

p.apply("BQ " + tableName, BigQueryIO.readTableRows().from(table))
                .apply(ParDo.of(new TableRowToRowConverter(BQSchema)))
                .setRowSchema(BQSchema);

public class TableRowToRowConverter extends DoFn<TableRow, Row>  implements Serializable {

    private static final long serialVersionUID = 1L;
    
    private final Schema schema;

    public TableRowToRowConverter(Schema schema) {
        this.schema = schema;
    }

    @ProcessElement
    public void processElement(ProcessContext c) {
        TableRow tableRow = c.element();
        Map<String, Object> fieldValues = new HashMap<>();
        assert tableRow != null;
        schema.getFieldNames().forEach(e -> {
            try {
                fieldValues.put(e, tableRow.get(e)!= null ? tableRow.get(e) : "");
            } catch (Exception e1) {
            }
        });
        Row modifiedRow = Row.withSchema(schema).withFieldValues(fieldValues).build();

        assert tableRow != null;
        c.output(modifiedRow);
    }
}