getting rid of append_samples

[Bob Carpenter]

I moved the discussion here from an RStan issue.

There's really no point in having append_samples,
at least until we have a way of restarting a chain.
I'd say we should just get rid of the feature, especially
if it hasn't already been documented.

- Bob

> On Jul 13, 2016, at 5:44 PM, Jonah Gabry <notifi...@github.com> wrote:
>
> Thanks for pointing this out. I'll add it to the documentation.
>
> append_samples is only relevant if sample_file is specified. If sample_file = posterior.csv and posterior.csv exists then setting append_samples = TRUE will append the samples to the existing posterior.csv instead of overwriting it.
>
> —
> You are receiving this because you are subscribed to this thread.
> Reply to this email directly, view it on GitHub, or mute the thread.
>

[Jonah Gabry]

Ok, I did end up documenting it on the develop branch, but should I undo that and have it throw an error if it's specified? I guess it's conceivable that some people _are_ using it even though it was undocumented (well, it was mentioned in the doc, just not explained fully), although I agree that it isn't particularly useful until we can restart chains.