hi, as a result of certs issue over the weekend, i though i would also upgrade snowflake-target from

v0.12.0

to

v0.16.3

-> what unfortunately resulted is the target trying to change variants to varchars. any idea why?

Hi Michal, it's probably related to this change: https://github.com/MeltanoLabs/target-snowflake/commit/3536b752532d18c4b5a3292c7a2f6a582ca39246. Do you know what's the original JSON schema for the field that's being changed to VARCHAR? Also, is tap-postgres the extractor you're syncing data from into Snowflake?

v0.12.0

doesn't try to change the data-type, but

v0.16.3

does. the tap is

pipelinewise-tap-mysql

; the source schema for the field is indeed JSON

Are you applying any schema overrides to the extractor?

We also have the same issue, after changing snowflake-target to

v.0.16.3

we get the error

cannot change column XYZ from type VARIANT to VARCHAR

. Tap is

tap-postgres

, source schema is JSON. Any suggestions ?

The thing that would help me the most in figuring this out is knowing the data type of the offending column in the source table, and also ideally the JSON schema type, e.g.

"xyz": {"type": [....]}

.

data type in source table is

jsonb

and we don't specify the JSON schema type

Ok, I got a PR that I think should fix this: https://github.com/MeltanoLabs/target-snowflake/pull/352

shouldn't we add a test or something before issuing out a new tag? or is it already covered by other tests?

shouldn't we add a test or something before issuing out a new tag? or is it already covered by other tests?

I'm curious what tests or in what context you mean? Tests that you (the end user) run, or tests in the target repo?

tests in the snowflake-target? just don't want to fix one issue only for another to pop up inadvertently; don't know the protocol here

At least for type mapping I've just added tests to prevent regressions: https://github.com/MeltanoLabs/target-snowflake/commit/9d6d4b370a0ef2e92fdcd6b8185df2bff165e7e4

@Edgar Ramírez (Arch.dev)

0.17.0

looks like it fixed it for us, thanks for the quick response!