Planning for 2.23 release

Developers
1

it’s 3/18 - according to schedule, release 2.23 is due out 4/18
(cf Should we do a CmdStan 2.22.2 release?)

questions:

@rok_cesnovar @serban-nicusor @stevebronder @syclik - thoughts?

1 Like
2

Strong yes for me

3

They are not really bugs but before RStan can switch to stanc3, we need to

  • Generate C++ code for standalone functions
  • Allow for quotation marks around paths to #included Stan files
3 Likes
4

Given that 4/18 is a Saturday, do we move this to 4/20 or 4/17 (not sure a friday release would be the best tho)?

Strong yes from me also. The question is only 4/10 (Friday) or 4/13 (monday)

On Math its probably all the neg binomial stuff. I didnt follow that PRs closely, @martinmodrak would know more.
I dont think there are any pressing issue on Stan/Cmdstan.

For stanc3 I think these three are the most pressing and hopefully simple:



I have an updated release notes (14 days behind to be exact) for Math ready (for the Release notes PR comments stuff we were discussing but I havent gotten around to yet). Stan/Cmdstan/stanc3 release notes should be easy and short and ready together with the RC release.

1 Like
5

The neg. binomial stuff is currently on the back burner due to Covid-related activity which currently feels of higher priority. Quite likely will return to normal in a few days so reasonably likely that I’ll be able to move the neg. binomial stuff over the finish line in time. Everybody stay safe.

5 Likes
6

no Friday releases! 4/20 release ?

(queue up emojis - 420 (cannabis culture) - Wikipedia)

1 Like
7

I am in. Oh yes, there will be 420 jokes :). @serban-nicusor 4/20 good with you? And 4/13 a release candidate release?

8

@rok_cesnovar Sounds good to me!

1 Like
9

code freeze is 4/13 - do we want to try and co-ordinate on bug fixes that folks would like to see get in by then?

pinging (some of) the usual suspects: @rok_cesnovar @bbbales2 @stevebronder @wds15

2 Likes
10

also pinging stanc3 devs - @rybern and @Matthijs

11

For Math I made an issue a few days ago to track anything urgent: https://github.com/stan-dev/math/issues/1826
There arent any pressing Math issues (I think) that must be fixed for the release.

Stan repo isnt really active currently, we can make a similar issue tho I am not sure its necessary there.

We can make similar ones for stanc3 and cmdstan. There are some pressing issues in stanc3 I listed above.

Or do we want keep the discussion in a Discourse topic for all repos?

2 Likes
12

a list for stanc3 would be good -
there’s this issue: https://github.com/stan-dev/stanc3/issues/485
others?

13

This one https://github.com/stan-dev/stanc3/issues/478

Maybe this https://github.com/stan-dev/stanc3/issues/471 but this one was there even for stanc2 so not pressing I guess.

14

Just reporting that new version of neg_binomial_2_lpmf got merged, but I didn’t get to work through neg_binomial_2_log_lpmf and most likely won’t be able to make it in time, sorry :-(

15

^Is that a bugfix or a performance improvement?

16

neg_binomial_2_log_lpmf has some outstanding numerical issues, the worst parts were fixed in https://github.com/stan-dev/math/pull/1622 but not all, but I think the issues now occur only for quite large phi. My aim with that was to very closely follow the current neg_binomial_2_lpmf (non-log) implementation and tests, only with the log substituted…

17

Ok, so I found some time for this and the issue should be fixed, https://github.com/stan-dev/math/pull/1830 (waiting for tests)

2 Likes
18

Shall we do the 2.23 release tomorrow - 4/22?

@rok_cesnovar @serban-nicusor ?

1 Like
19

Hey @mitzimorris that’s fine with me.

1 Like
20

Sure, I will post Stan/Cmdstan release notes so Nic can add them.

1 Like