Eric Soroos

It's possible what you want is 5 and up -- that's getting to the point that the planner might choose to use the index, and that's something that an explain...

As a side note -- I'm not sure it's worth specifically creating unique indexes when doing the loads from datapusher, unless you're specifically trying to constrain additional entries that are...

This is a draft for comment -- * I'm not clear that I'm hooking into the best place to provide the HVD only catalog. * The resource refactoring is pretty...

I can definitely see wanting an alternate datastore system -- and there are potentially two approaches to do that: 1) A different datastore backend, e.g. a different database technology, where...

So one of the ideas I had was a web-accessible sql enabled data lake. 1) Get a bunch of parquet files, either on s3, other object storage, or in the...

Another use case -- Datastore should transparently use a sqlite backend when there's a sqlite database uploaded as a resource, at least in the simple case of a single table...

I've long been in favor of deprecating the advice to install extensions in editable mode, and ensuring that appropriate dependencies are picked up by the solver. I think in most...

@duttonw Install order shouldn't matter, and can't matter if the requirements are specified and obeyed. It only matters if you're installing one dependency, and then separately installing another extension with...

Order certainly matters in ckan.ini, and yeah, that's a pain. What I'd recommend is look at the string of pip installs that you're doing and see if anything is ever...

@smotornyuk I'd add -- Must/Should not set any versions on dependencies that are also dependencies of ckan core.