Hi, I’m trying to connect GoodData to a database that uses

__

to prefix certain columns. This is preventing me from adding adding these tables using the LDM. I get the following error:

Column name is not aligned with naming conventions name=__cubeJoinField separator=__ numOfParts=2 expectedNumOfParts=3

The answer to this previously asked question says that the LDM should give me an option to ignore columns https://community.gooddata.com/data-sources-59/why-are-column-names-with-invalid-514 However I can’t find this option anywhere. Can anyone confirm that this should be possible? Thanks!

Hi Eugene. Can you please confirm that you are working with GoodData CN in this case?

Hi Joseph, yes I believe it is CN as I signed up for a trial

The default configuration of the LDM is based on the Physical Data Model, so you will need to ensure that the PDM is set up accordingly.

If I understand your needs well, you are mixing two use cases into one and request something what is not feasible only with GoodData UI apps / resp. with GoodData raw APIs. GoodData provides two use cases: • scan -> result is so called PDM, it can be stored to backend • generateLogicalModel -> result is LDM generated from the stored PDM The scan provides options to filter views/tables, their names with prefixes, not no filtering of columns is possible. However, PDM columns are not directly used by end users. The generateLogicalModel can generate LDM from PDM, so you may not create it manually. LDM entities are exposed to end users. It accepts a lot of arguments, one of them is

wdfPrefix

, which causes no LDM entity (attribute/fact/..) is generated from matching columns. If you are looking for this feature (limit generation of LDM entities from columns), this prefix can be used, but unfortunately

__

is not valid pattern in this case, because we expect prefix (e.g.

wdf

) followed by seperator(e.g.

__

). Generally, if you even want to exclude columns generated into LDM or if you want to customize the generated LDM, you can always utilize our Python SDK. You can execute scan/generateLogicalModel with the SDK and then apply any logic on it in Python language. I did it in many demos and can give you an example if you are interested.

Thanks for your help with understanding this, Jan. I’ve been able to get around this in the LDM UI using the ‘Start From Scartch’ > Empty Dataset and then manually mapping the DB fields. I’ll have to look at the python sdk workflow when we need to introspect more tables in the future