Thanks for opening a pull request!

If this is not a minor PR. Could you open an issue for this pull request on JIRA? https://issues.apache.org/jira/browse/ARROW

Opening JIRAs ahead of time contributes to the Openness of the Apache Arrow project.

Then could you also rename pull request title in the following format?

ARROW-${JIRA_ID}: [${COMPONENT}] ${SUMMARY}

or

MINOR: [${COMPONENT}] ${SUMMARY}

See also:

• Other pull requests
• Contribution Guidelines - How to contribute patches

https://issues.apache.org/jira/browse/ARROW-13005

ARROW-13005: [C++] Implementation of take function for dense union data type.

@ZMZ91 Thanks for tackling this slightly delicate issue. I've posted two comments about the chosen approach.

Hi @pitrou and @bkietz, just update the pr due to your comments. Please help check the latest code. Thanks a lot.

Thank you @bkietz. All comments resolved. Please help check.

Hi @bkietz and @pitrou, just a question not quite related to this pr and about dense union array. I've seen the limitation for array data is under 2GB. Does it apply to nested arrays or just the sub arrays inside? For example, dense union array, it has child id array, value offset array and a bunch of child arrays. Each of them should be under 2GB. And is it restricted that the sum under 2GB?

The 2GB limitation applies to buffers and is type specific. It derives from usage of signed 32 bit integers to encode offsets, so for example a StringArray cannot have more than 2GB total character data because otherwise we could not express the final offset. IIUC it's allowed for a child of the union to be an Int8Array with 2**33 elements but a dense union's offsets could not point to most of those elements.

In short, this is not a constraint you need to enforce in this PR.

Hi @bkietz and @pitrou, could you help check what else to update in this pr? Or we may merge it? Thanks.