Andrew Mascioli

I started working on fleshing this out more completely and found that the `pyarrow` solution wasn't scaling as I expected. This prompted doing a run of more careful performance testing,...

Polars has no other dependencies, so it seems it would be relatively lightweight. It's also easy enough to include it optionally like Brian suggested. Personally, I lean towards including it...

I've got a draft PR #799 that implements the main ideas here for the MCMC outputs. With it mostly focusing on abstracting out the CSV interaction, it does seem like...

I started working on this last night briefly - replacing existing CSV reading for draws should be quick. Pathfinder and laplace are like 3 line changes each (also gq). MLE...

Regarding dropping support for older versions: for what it's worth, I would lean towards being a bit more aggressive with only maintaining support for recent versions (or perhaps tying some...

We still want to add this, correct? It would be fairly straightforward to do so in line with the other output files we're tracking. With this enabled, we would then...

I have been taking a look at this today and have a question. When we do something like perform a `CmdStanModel.generate_quantities(..., previous_fit=[files])`, where `files` is a list of stan csv...

> Yes. I saw that one too... > > I just tried to do the same thing using `multiprocessing.Process` and saw the same result. New pseudocode: > > from multiprocessing...

Yeah if it were just this, I wouldn't be interested in adding the dependency, but with potentially quite a few output files that need to be parsed, I think it...

No worries! There was one test failure that was numerical. That was just noise. I'm scratching my head on these Windows failures, though. I tried to replicate in a VM...