Using Borrow<Value> on infer_json_schema_from_iterator

[rguerreiromsft]

Which issue does this PR close?

None

Rationale for this change

In a project I'm working on, we need to save the json data and the schema into a different file. Today we need to clone the original json, where it'd be much easier to just use a read-only reference.

Are there any user-facing changes?

No, using Borrow makes any owned instance still work here.

[tustvold]

I've left a suggestion on how to avoid the Box.

FWIW the Decoder that accepts Value is intended to be deprecated soon, see #3718, I'd be interested to hear your thoughts on this

[tustvold]

Suggested change

	V: AsRef<Value>,
	V: Borrow<Value>,

I think this should remove the need for the boxing

[rguerreiromsft]

Thank you for both suggestions, I really appreciate them! I didn't remember about Borrow when coding it.

About the RawReader, awesome, I'll start using it right away!

However, I still need to use the infer_json_schema_from_iterator because I have to manually work with serde_json::Value before I attempt to generate the schema.

Then I save this schema into a file to be used later, but I'm using tokio::fs to work with files, it'd be nicer to have the RawReader also support async work. It's not a huge issue, though, I can use std fs for this particular operation.

On a separate note, I don't think exposing Values is a bad thing, sure a user doesn't need it, but when we're building a library on top of Arrow, it's nice to have access to lower level apis.

[tustvold]

it'd be nicer to have the RawReader also support async work

You might want to try out RawDecoder in that case, it is designed for use in async contexts. You could possibly use it in conjunction with AsyncBufRead.

FYI no operating systems currently support async file IO, tokio actually just calls tokio::spawn_blocking under the hood, unless you're using something like glomio or streaming data over a network, async is not necessarily helpful.

I don't think exposing Values is a bad thing,

The reason for moving away from Values is to avoid intermediate allocations, and allow for vectorised decoding, there wasn't a way to achieve these with the current API.

[tustvold]

Thank you

[ursabot]

Benchmark runs are scheduled for baseline = 221533f and contender = 59016e5. 59016e5 is a master commit associated with this PR. Results will be available as each benchmark for each run completes.
Conbench compare runs links:
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on ec2-t3-xlarge-us-east-2] ec2-t3-xlarge-us-east-2
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on test-mac-arm] test-mac-arm
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on ursa-i9-9960x] ursa-i9-9960x
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on ursa-thinkcentre-m75q] ursa-thinkcentre-m75q
Buildkite builds:
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

[rguerreiromsft]

Thank you so much! I'll take a look at the RawDecoder. Because I'm also streaming multiple json files before saving them locally into a single file, which I'm providing to the RawReader.

I couldn't test everything yet, I'm certain there are improvements I need to do before finalizing everything.