fix(go/adbc/driver/snowflake): try to suppress stray logs

[lidavidm]

It turns out gosnowflake also exposes a log level control; set this to "warn".

I also filed snowflakedb/gosnowflake#1332. We need the fix to be upstream since ultimately the log message is triggered by constructing a global, so we're powerless to suppress it. It was ultimately fixed by snowflakedb/gosnowflake#1327.

[lidavidm]

It appears gosnowflake uses logrus which defaults to "info" so that would explain why we're seeing the stray info log.

[lidavidm]

Ah...the log message comes from a constructor for a global object, so it runs before we can suppress the message. While this PR is still good to have, it can't fix the underlying issue; we need Snowflake to fix it.

[lidavidm]

The one thing I can do here is wire up the snowflake logger to the ADBC logger at least.

[paleolimbot]

Thanks for looking into this!

A while back I patched gosnowflake specifically for a CRAN submission because it was creating files when initializing a global (maybe the same one). If useful, the patch is here: #1315 (comment)

[lidavidm]

Ah, that's what I was vaguely remembering. That seems to be a different code path but I suppose the same line of thinking.

[lidavidm]

Hmm, seems Snowflake has a global logger + it's pretty tied to the logrus interface; might not be good to bridge the two by default. Going to leave this PR as-is then

[lidavidm]

FWIW Dewey, it seems the PR where Snowflake is fixing this is primarily adding some file-based cache of auth tokens...so we might just be trading one problem for another (if stray files are a concern for CRAN)

[paleolimbot]

Thanks for the heads up! We don't submit this one to CRAN and so it's fine, and even if we did, the problem is only if a file gets created when constructing a global (creating files during connection, while theoretically frowned upon by CRAN, is not something they would be able to detect because we very reasonably can't create a connection when running the tests on their servers).

[lidavidm]

@zeroshade any comments here?

[zeroshade]

Do we expose setting the log level anywhere in the API?

Should we?

[lidavidm]

Do we expose setting the log level anywhere in the API?

Should we?

I looked at it, but the struggle is that we have a per-database log level and Snowflake has a global logger. Maybe we can config a separate logger per Snowflake connection though. But for now I figured we should just set this (even though it doesn't help) and wait for Snowflake to fix the underlying issue upstream.

[lidavidm]

Looks like the snowflake PR merged, I will try and bump (temporarily) to see if it does fix the issue here