Hi everyone! I am using meltano to extract some data from a MySQL database, but for some reason loaders fail to validate table columns that are of JSON type: the column type in the catalog is

['null', 'object']

, but the json payload is sent to the loader as a string instead of a python dictionary and the validation fails:

jsonschema.exceptions.ValidationError: '{"content_types": ["formal"]}' is not of type 'null', 'object'

Is there a way to overcome this problem? I was trying to use the

_schema

extra to override the type of this field to string, but loaders seem to ignore this setting. Is there anything I'm doing wrong?

I'm guessing you are using target-bigquery as a loader here?

No, I've actually tried csv and jsonl (I'm still testing meltano locally). I have also opened an issue here about this problem (I've been using meltano since 24 hours, so I hope I am not doing any errors on my end). As a temporary workaround I've tried using the

meltano-map-transformer

to convert these columns to strings on their way to the loader, but this is not a great solution, especially because the mapper doesn't support wildcards in the stream id. If I have 3000 schemas - which I have - I can't create a rule like

*-table.column

, but I should create one rule for each schema, which is not sustainable. Finally, I have also tried changing the

_schema

extra, but it only changes the column type in the catalog, not in the actual records being sent.