Update to handle change in sql_validate return type

[jclay]

This is to accommodate the changes made to the OmniSci client in heavyai/pymapd#325

[datapythonista]

Thanks @jclay. Do we need to require a higher version of pymapd after this change?

Also, would be good to have a test, and an entry in the release notes. Thanks!

[jclay]

@datapythonista I have updated the CI suite to use the latest pymapd released today. I will open a PR on the feedstock repo as well.

Note that this was only affecting OmnisciDB v5.3.0 users, so while Ibis had good test coverage for this already, we were testing against v5.2.2 so it was not showing up in the tests. I have added coverage of sql_validate functionality on the pymapd side.

I'm not sure if it's related to the pymapd version bump, but I am noticing a drastic increase in the time it takes for conda to resolve the dependency constraints. Once cudf 0.15 is released, we should be able to relax some of the dependency constraints on the pymapd side.

[xmnlab]

mmm for some reason it is raising this error:

ibis/omniscidb/tests/test_operations.py:87: AssertionError

not sure why it is changing the result size:

>       assert df.size == 80
E       assert 70 == 80

[datapythonista]

@jclay, pymapd=0.24 was giving us problems in the CI, conda never finished to solve the environment, and we had to pin the version to 0.23 in #2272. Does this PR still makes sense?

You can investigate the problems with pymapd and conda, but for now we'll keep pinning to a specific version, and upgrade only when we know that the solver is not causing trouble.

[datapythonista]

@jclay is this something we still want to get merged?

[jclay]

I think we will want to get this merged at some point given it fixes the sql_execute breakage we were seeing on OmniSci 5.3. Let me close this out for now until I have a chance to revisit and get the failing test resolved.

[tmostak]

Hi @jclay, any update on this? Spoke to Eric K and he mentioned it was causing major headaches for a prospect.