Open MaximilianAlgehed opened 4 months ago
I tried to get a grip on the number of people we will annoy with this change by searching for the following query on github withMaxSuccess language:haskell NOT is:archived NOT is:fork
and it gave back ~600 files (with at most 50 files being instances of withMaxSuccess
used in a comment).
Now, the thing I wasn't able to ascertain was how many distinct repositories would be impacted, but would I suspect it's somewhere between 60 and 300.
The reason for making this change is the change in #383 that means that maxSuccess
is never really max success (it wasn't before either - but now it's even more so not the case). Weighing that against pissing up to 600 people off however doesn't immediately worth while to me.
My plan is to wait with this change until we have enough other changes that will annoy people that we can't ignore and just bundle all the breaking changes in one place.
Another option would be to introduce withNumTests
but not yet mark withMaxSuccess
as {-# DEPRECATED #-}
. We could still update the Haddocks to tell people to use withNumTests
instead.
@MaximilianAlgehed https://hackage-search.serokell.io/ is better corpus to search through
@MaximilianAlgehed https://hackage-search.serokell.io/ is better corpus to search through
Thank you! I'll do a more thorough search there.
There are about 50 odd packages on hackage that use withMaxSuccess
that we have to care about (i.e. aren't mentioning it only in a comment).
After a discussion about changing the interaction between
checkCoverage
andwithMaxSuccess
wherebycheckCoverage
will ignore the number given inwithMaxSuccess
(becausecheckCoverage
may require more tests than what is specified) it was decided that a more intuitive name forwithMaxSuccess
iswithNumTests
. The new name reflects the fact thatwithNumTests
does not set a hard limit (unlike what is indicated by theMax
inwithMaxSuccess
) on the number of tests that are run.