Closed nickrobinson251 closed 2 years ago
oh i think Jacob is away this week -- i wonder if someone else with permissions is around to review (and make a patch release). Maybe @nalimilan?
Thanks. Any particular reason to rush a release before Jacob is back? We can do it but it would be good to double check with him.
Thanks for reviewing!
Any particular reason to rush a release before Jacob is back? We can do it but it would be good to double check with him.
No more so than usual. Just for the benefit of @nilshg who hit this bug. Seemed an straight forward enough fix to release on its own.
Thanks
thanks for adding me to the repo, Jacob!
Before
inlinestrings([missing, "e"])
would fail, even thoughinlinestrings(["e", missing])
would succeed as wouldinlinestrings([missing, "ee"])
. This was because in first case the initial results vector eltypeeT
is set toMissing
and then when we see a stringy = "e"
we'd havey !== missing && sizeof(y) == 1
and hit the branch which callsnonmissingtype(eT)(y)
. We only hit this branch whensizeof(y) == 1
since the otherside of the check,sizeof(y) < sizeof(eT)
, is always false (sincesizeof(Missing) == 0
). (This bug was accidentally introduced by me in #26) The fix is to never hit this branch ifeT == Missing
.