Brian Ward

What version of cmdstan are you using and what is the error you’re getting? A couple versions ago we made a change to try to decrease the frequency of errors...

> Also, why are you asking people to open a bug report given that this is expected behavior for which we have no better solution? If `rebuild_cmdstan()` does not resolve...

@carrascomj what action would you like to see, given that the latest cmdstan fixed this issue?

We test against the latest released version, but we currently support old versions on a best-effort basis, which can be seen by the [presence of version-based conditionals](https://github.com/search?q=repo%3Astan-dev%2Fcmdstanpy+if+cmdstan_version_before&type=code) in the code....

> Removing that early return and testing that it satisfies the expectations would make sense. Removing the early return seems fine at first glance, since the `validate` function already should...

I'm reasonable certain what is slow is not seeking around the file or skipping the `#` rows, but actually parsing the numbers into floats. Right now, sampling does this in...

I did some rough timings on a fake CSV file with 1000 rows and 45 columns ```python def read_manual(filename): # similar to current cmdstanpy code draws = np.empty((1000, 45), dtype=float)...

Nice find @amas0. I suspect you might be able to get event faster if we wrote our own little TextIO wrapper to avoid needing to `.join` the entire `no_comments` iterator,...

That's no problem, just wanted to check! I think a good starting place is a function like: ```python def split_csv_comments(path: os.PathLike) -> Tuple[string, BytesIO]: comments = b"" data = b""...

I think it would be a really nice benefit of this if all of the code ended up sharing more pieces in the end. That being said, if you wanted...