Repository navigation
Update to StatsFuns@2 - #2062
Conversation
The new `nbinomlogpdf` in StatsFuns@2 is not ForwardDiff-safe at the `p = 1, k = 0` boundary (the `0 <= p <= 1` guard misfires for `Dual(1.0, +ε)` under ForwardDiff's lex-ordered `<=`, and the all-Dual frule for `xlog1py(0, -1)` produces a NaN partial). The pre-existing hand-written `logpdf` block already handled this correctly, so keep it and only delegate `cdf`/`quantile`/etc. via explicit one-liners as before. Also bump the `StatsFuns` compat to `2.0.1`, which carries the upstream fix for `srdistcdf` (#211 in StatsFuns) that the `StudentizedRange` and `Truncated{StudentizedRange}` tests depend on.
| Statistics = "1" | ||
| StatsAPI = "1.6" | ||
| StatsBase = "0.32, 0.33, 0.34" | ||
| StatsFuns = "1.5.2" |
There was a problem hiding this comment.
Is it necessary to drop support for 1? It doesn't look like it required any code changes.
There was a problem hiding this comment.
Not anymore, it did in the initial version - but these additional changes would require even more fixes on the StatsFuns side, so I postponed them.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #2062 +/- ##
=======================================
Coverage 86.65% 86.65%
=======================================
Files 148 148
Lines 8919 8919
=======================================
Hits 7729 7729
Misses 1190 1190 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
The PR should not have been merged. It requires JuliaStats/StatsFuns.jl#211. Or, more precisely, without it the version bound is pretty useless. |
I don't think that was clear from the description. Wasn't it marked as ready? |
3 participants
No description provided.