feat(go/adbc/driver/flightsql): handle is_update in prepared stmts

[ennuite]

AI Disclaimer

This change was created with AI assistance (Augment Code). All lines were manually reviewed by a human. The output is not copyrightable subject matter.

Closes#4074

[ennuite]

I'm unsure of which component to use given that I touched 3 different ones: ‎go/adbc/driver/flightsql, ‎python/adbc_driver_flightsql and ‎python/adbc_driver_manager

It's my first PR in this codebase and I'm not very familiar with Go. I would appreciate a review not only on the content but also on style. I'm available for improving this as requested.

It's not yet in a state where I feel like it's ready to merge, although it is fine as a PoC.

Things I want to improve:

1. Add some tests for the Python driver.
2. Change the way I'm fetching the option. Details: it is being fetched in a code path that is shared by all drivers and not just by the Flight driver, and in those cases it will always throw an exception (which is caught with no side effects). This might hinder performance, and I'm not a proponent of using exceptions for control flow. I intend to move the logic to a code path specific to the Flight SQL driver and remove the exceptions.

Question: I added a logger in Python, but I don't see logging being used anywhere else. I would like to know your opinion on log statements in this codebase.

[ennuite]

This PR is not an individual piece of work. It requires changes in Arrow-Go and the main repo.

[lidavidm]

Don't make any Python changes. It should be handled entirely in the driver.

[ennuite]

Don't make any Python changes. It should be handled entirely in the driver.

@lidavidm I thought about that and it would definitely be much easier. However, do we want this in ADBC Go? My understanding is that, in ADBC Go, queries that generate result sets go via Statement.executeQuery() and queries that don't go via Statement.executeUpdate(). Do we want DML queries in Statement.executeQuery()?

[lidavidm]

Yes, that should be handled properly. It'll simply return an empty resultset.

[ennuite]

I see you already approved. I will still add tests tomorrow, but the feature work is finished. After I add the tests, if you agree, I think it's time to put the spec change to a vote.

[ennuite]

@lidavidm I added tests for this new behavior. I also manually tested it end-to-end, using the Go driver and the Python driver, against backend servers with and without the change to ensure backwards compatibility.

Let me know if we can proceed with voting on the spec change.

[ennuite]

Converted to draft until I carry over the latest changes from apache/arrow-go#732

[lidavidm]

@ennuite is this still in progress?

[ennuite]

@lidavidm It's ready for review now. Sorry for the delay.

[lidavidm]

Seems good, but we need the arrow-go changes

[zeroshade]

I added a bunch of comments on the corresponding arrow-go PR that are waiting for responses

[ennuite]

@lidavidm The Arrow-Go changes have been merged, see apache/arrow-go#732
Let me know if you need anything else here. Do we need to wait for an Arrow Go release before merging this?

[lidavidm]

Well, yes, we can't merge in this state.

[zeroshade]

I should be doing an arrow-go release in the next week or so