This is no good because it doesn't deal with the cache

This is really fiddly because either:

• You want to execute the pipeline, add it to the cache and return the dataframe
• Retrieve from the cache if exists
• Return a pipeilne with the right sqls enqueued

So there's not an easy version of this that uses a fresh sql pipeline. you can pass one in, but then uncear what to return

Another option would be to allow the pipeline to queue input dataframes, tha seems promising ?

in that case, a fresh sql pipeline is passed in. If it's in the cache, then the pipeline is returned with it as an input dataframe queued up

if it's not in the cache either

• materialisation case: a pipeline is returned with it in the cache
• non-materialisation case: a pipeline is returned with sql enqueued