rhunwicks commented on issue #3302: Create a PandasDatasource URL: https://github.com/apache/incubator-superset/issues/3302#issuecomment-329281110 @xrmx it's true that if someone developed a model that had a dependency on a core Superset model, and then that model changed significantly, then the extension might stop working. But that would be the case whether or not the dependent model was in the core `superset/migrations/versions` directory or not. The impression I have from @mistercrunch comments above was that if we can write a `PandasConnector` it would be non-core anyway. And so given that objective it makes sense to try and keep the extension as loosely coupled as possible. Alembic, and thus Flask-Migrate, is quite capable of dealing with a branched migration tree, so changes can be made to the main models in the main tree without affecting the extension migrations provided that either they are not dependent on any upstream models or the upstream models don't change in a breaking way. For example, our models have a foreign key to `ab_user` but as long as changes to that model don't alter the primary key I don't think it will make any difference. If someone wrote an extension that did stop working as a result of change to an upstream model then they would need to update their component to account for the changes in order to continue working with the most recent version of Superset. This is a normal workflow for an add on component for an open source project. On balance, I think that it is preferable to have an `ADDITIONAL_VERSION_LOCATIONS` config parameter that would allow me to (attempt to) develop a `PandasConnector` that is minimally invasive to core Superset. I think that smaller the impact it has on core Superset the better chance it has of being successful. I think that the same logic might apply to other potential extensions - as I described at the start of this Issue there have been many requests for additional connectors, and making it easier for me and others to develop them without needing changes to core Superset or maintain a fork would likely increase the availability of such connectors. However, my main objective is to write a connector that can work with core Superset on an ongoing basis. So if the core team decide that Connectors should live in a contrib directory and keep their migrations in the main `superset/migrations/versions` directory then I'll happily comply. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
With regards, Apache Git Services
