<https://github.com/linkedin/datahub/pull/2898> <@...
# ingestion
s
https://github.com/linkedin/datahub/pull/2898 @gray-shoe-75895 review please. I tested it on my local and it works. This is for adding this feature https://datahubspace.slack.com/archives/CUMUWQU66/p1625566182481100
g
I think it makes sense to add this for now, but adding this option per source feels unsustainable so there’s a chance we’ll change this in the future - ideally via a more generalized transformer
s
I was initially writing a transformer for allowing replacement of any property of the Schema. Got half way through but it felt hacky. Basically because
db_name
is not a property but part of uri. So a string replacement might have worked but it could break other things. Might be a different way to write the transformer that did not occur to me. From what I could see in the code the sql source could easily use this by default and in other cases (like bigquery) where something else (like gcp project) makes more sense can override the method. That way this option won't be per source and we don't need a separate transformer. But I did not have all sources to test out so did it for postgres only. Should be medium-ish effort to make it in the super class. Just need to be able to run the integration test
1
g
Yep so I agree that parsing the URN strings feels pretty hacky - luckily, we’re trying to slowly migrate to using “key aspects” which have all the components of the urn split out into separate fields - once we do that, writing the transformer won’t be quite as bad
1