ARROW-17349: [C++] Allow casting map types

[wjones127]

• Implements casting functions from MapType -> [MapType, list<struct>, large_list<struct>], with support for mapping the field names to new ones.
• Adds a new Datum::View() method to map an array to a new type.
• Use Datum::View() in early return of cast dispatching (if types are "equal"). Right now this is useful for mapping field names for map arrays, but it could also be used for list types once we enable some way to check if list types are equal except for its field name. See also: https://issues.apache.org/jira/browse/ARROW-14999

[github-actions]

https://issues.apache.org/jira/browse/ARROW-17349

[wjones127]

Mapping MapType field names

import pyarrow as pa

# MapType
ty_named = pa.map_(pa.field("x", pa.int32(), nullable=False), pa.int32())
ty = pa.map_(pa.int32(), pa.int32())
arr_named = pa.array([[(1, 2), (2, 4)]], type=ty_named)
arr_named.type, arr_named.cast(ty).type

Before (wrong)

(MapType(map<int32 ('x'), int32>), MapType(map<int32 ('x'), int32>))

After (correct)

(MapType(map<int32 ('x'), int32>), MapType(map<int32, int32>))

Mapping Map<T, List> names

ty_named = pa.map_(pa.string(), pa.field("x", pa.list_(pa.field("x", pa.int32(), nullable=True)), nullable=False))
ty = pa.map_(pa.string(), pa.list_(pa.int32()))
arr_named = pa.array([[("string", [1, 2, 3])]], type=ty_named)
arr_named.type, arr_named.cast(ty).type

Before

pyarrow.lib.ArrowNotImplementedError: Unsupported cast from map<string, list<x: int32> ('x')> to map<string, list<item: int32>> (no available cast function for target type)

After

(MapType(map<string, list<x: int32> ('x')>), MapType(map<string, list<item: int32>>))

[wjones127]

Failures seem to be related to Gandiva, not changes in this PR.

[wjones127]

cc @pitrou @jorisvandenbossche

FYI one of the motivations for this is making sure it's easy to cast field names so the transition for Parquet to use compliant nested types is easier (ARROW-14196).

[pitrou]

Nice improvement @wjones127 . Here are some comments.

[pitrou]

Do you want to open a JIRA for this TODO?

[wjones127]

I'll be doing this in #13851

[wjones127]

These CI failures are unrelated.

[pitrou]

Thanks for the update. Just a couple more comments/questions about the tests.

[jorisvandenbossche]

Looks good, nice improvement!

One question: in general we allow renaming list fields while casting, but for structs the fields have to match (since for structs the names are much more important / meaningful than for lists).
Here, you also allow renaming the fields when casting to a struct. I think I am fine with that, but just wanted to mention it explicitly, since it's a grey area between both cases mentioned above (for structs the names are important, but for a map type they also don't matter much, so the first cast to struct can ignore them?)

[wjones127]

Here, you also allow renaming the fields when casting to a struct. I think I am fine with that, but just wanted to mention it explicitly, since it's a grey area between both cases mentioned above (for structs the names are important, but for a map type they also don't matter much, so the first cast to struct can ignore them?)

Yes, I was careful to make sure this only happens in the map -> list<struct> path and doesn't affect casts from structs. Lists and maps will support renaming fields through casting, structs will not (perhaps we want a different function for it?).

[wjones127]

@pitrou could you take another look?

[pitrou]

Sorry for the delay @wjones127 .
Also I forgot to mention that we should update the list of available casts here:
https://arrow.apache.org/docs/dev/cpp/compute.html#conversions

[pitrou]

Thanks a lot @wjones127 !

[ursabot]

Benchmark runs are scheduled for baseline = f25d88e and contender = 69aad53. 69aad53 is a master commit associated with this PR. Results will be available as each benchmark for each run completes.
Conbench compare runs links:
[Finished ⬇️0.0% ⬆️0.0%] ec2-t3-xlarge-us-east-2
[Failed ⬇️0.56% ⬆️0.0%] test-mac-arm
[Failed ⬇️0.27% ⬆️0.0%] ursa-i9-9960x
[Finished ⬇️0.39% ⬆️0.0%] ursa-thinkcentre-m75q
Buildkite builds:
[Finished] 69aad53e ec2-t3-xlarge-us-east-2
[Failed] 69aad53e test-mac-arm
[Failed] 69aad53e ursa-i9-9960x
[Finished] 69aad53e ursa-thinkcentre-m75q
[Finished] f25d88ed ec2-t3-xlarge-us-east-2
[Failed] f25d88ed test-mac-arm
[Failed] f25d88ed ursa-i9-9960x
[Finished] f25d88ed ursa-thinkcentre-m75q
Supported benchmarks:
ec2-t3-xlarge-us-east-2: Supported benchmark langs: Python, R. Runs only benchmarks with cloud = True
test-mac-arm: Supported benchmark langs: C++, Python, R
ursa-i9-9960x: Supported benchmark langs: Python, R, JavaScript
ursa-thinkcentre-m75q: Supported benchmark langs: C++, Java