GH-51100: [C++][Format] Deprecate IPC tensor/sparse tensor messages

[pitrou]

Rationale for this change

Deprecate Tensor and SparseTensor messages in the IPC specification.

Deprecate C++ APIs for reading and writing Tensor and SparseTensor IPC messages.

Deprecating IPC support for Tensor and SparseTensor was decided in https://lists.apache.org/thread/w7nvhjyks3p3smw5kq4ylv0nkwxq2s2q

Are these changes tested?

By existing CI tests.

Are there any user-facing changes?

No.

• GitHub Issue: [C++] Deprecate IPC tensor/sparse tensor messages #51100

[github-actions]

⚠️ GitHub issue #51100 has been automatically assigned in GitHub to PR creator.

[pitrou]

Let's keep this PR draft until final consensus is reached on the ML.

[rok]

@pitrou I'm very much on board with the sparse tensors. I so wonder about tensors, are we sure this is not being used? I'm ok to add a deprecation and wait for a release to see if we hear back. But my expectation is there would be some users out there.

[pitrou]

I'm ok to add a deprecation and wait for a release to see if we hear back. But my expectation is there would be some users out there.

Hopefully the deprecation will have them chime in? But my attempt at code-searching this on GH did not lead to many recent relevant results.

[rok]

I'm ok to add a deprecation and wait for a release to see if we hear back. But my expectation is there would be some users out there.

Hopefully the deprecation will have them chime in? But my attempt at code-searching this on GH did not lead to many recent relevant results.

Agreed. I'll take a look myself and report back.

[rok]

I've checked and found some projects using Tensor (1, 2, 3, ..), but I think they can all easily switch to tensor array. I agree with deprecating Tensor.

[pitrou]

@wgtmac @CurtHagenlocher @lidavidm Would you like to take a look?

[pitrou]

@github-actions crossbow submit preview-docs

[rok]

Should this also cover pyarrow.ipc.read_tensor, write_tensor, and get_tensor_size?

[pitrou]

We can tackle Python in a separate PR, IMHO.

[pitrou]

(I'm not sure what's with the "Release and Merge Script" CI jobs? @kou )

[github-actions]

Revision: a15a298

Submitted crossbow builds: ursacomputing/crossbow @ actions-19c98b222c

Task	Status
preview-docs

[pitrou]

Rendered docs at https://s3.amazonaws.com/arrow-data/pr_docs/51102/format/Other.html

[pitrou]

@github-actions crossbow submit preview-docs

[github-actions]

Revision: 9037076

Submitted crossbow builds: ursacomputing/crossbow @ actions-9ae2c160cd

Task	Status
preview-docs

[kou]

+1

The "Source Release and Merge Script" failure is a timing problem. It should be fixed by re-run. (I've re-run the job.)