Matthew Kay

In order to implement this, it might make sense to create some systematic infrastructure for resolving subtype conflicts amongst MC / MCMC / weighted MC / weighted MCMC. This would...

With the (relatively) new `map_vec()` and `pmap_vec()` in purrr, I don't think `map_rvar()` or `pmap_rvar()` are needed anymore; if someone wants those they can use the `purrr` functions.

I have a feeling most of their data manipulation code will break... One option is to cast to a `distributional::dist_sample()`, which is a list type and (I think?) should work...

Have you tried anything besides putting a column in / getting it out? Last time I investigated this I got as far as this comment and concluded rvar is fundamentally...

Cool! The core is [this commit](https://github.com/stan-dev/posterior/commit/930e7a830fa0ce1962fb269c50c660204258ce23)? Looks pretty straightforward. I wonder about what we want the long term plan to be: do we want folks to migrate to the version...

What about three values to `.use_rvars`: - `FALSE`: doesn't use them - `TRUE`: uses them, and if errors happen doesn't try to catch them (as the user has opted in)...

Cool thanks! > We should be mindful of how this could scale with many variables and several summary functions, particularly catching errors and potential warnings/messages. Yeah good point. Ideally people...

Hey all --- I wonder if there's been any further thought on this idea @lionel- proposed? > I wonder if we could add a customisation point for specifying which index...

Sure, would love a fix! I'd probably double check with @paul-buerkner to see if there's a canonical way to get chain and iteration info out of `posterior_predict()` and the like

Some updates on this on the dev branch: ### Re: applying quantiles but respecting integer bins This should be possible now if you let ggplot know that the input data...