emkornfield commented on a change in pull request #9504: URL: https://github.com/apache/arrow/pull/9504#discussion_r577320416
########## File path: cpp/src/arrow/csv/writer.cc ########## @@ -0,0 +1,398 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +#include "arrow/csv/writer.h" +#include "arrow/array.h" +#include "arrow/compute/cast.h" +#include "arrow/io/interfaces.h" +#include "arrow/record_batch.h" +#include "arrow/result.h" +#include "arrow/result_internal.h" +#include "arrow/stl_allocator.h" +#include "arrow/util/make_unique.h" + +#include "arrow/visitor_inline.h" + +namespace arrow { +namespace csv { +// This implementation is intentionally light on configurability to minimize the size of +// the initial PR. Aditional features can be added as there is demand and interest to +// implement them. +// +// The algorithm used here at a high level is to break RecordBatches/Tables into slices +// and convert each slice independently. A slice is then converted to CSV by first +// scanning each column to determine the size of its contents when rendered as a string in +// CSV. For non-string types this requires casting the value to string (which is cached). +// This data is used to understand the precise length of each row and a single allocation +// for the final CSV data buffer. Once the final size is known each column is then +// iterated over again to place its contents into the CSV data buffer. The rationale for +// choosing this approach is it allows for reuse of the cast functionality in the compute +// module // and inline data visiting functionality in the core library. A performance +// comparison has not been done using a naive single-pass approach. This approach might +// still be competitive due to reduction in the number of per row branches necessary with +// a single pass approach. Profiling would likely yield further opportunities for +// optimization with this approach. Review comment: yep :) ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
