Add Defer.fix#3208
Conversation
| } | ||
|
|
||
| forAll { n0: Int => | ||
| val n = n0 & 0x3FF // don't let it get too big |
There was a problem hiding this comment.
This is nitpicky but could you just write this out for clarity? I can't remember off the top of my head what this does for negative n0 for example.
There was a problem hiding this comment.
It's taking the 18 least significant bits. Perhaps a slightly clearer way to write it for those who aren't doing bit manipulation on a regular basis would be math.abs(n0 % 262144).
There was a problem hiding this comment.
I've changed it to be more direct Gen.choose(0, 1000) which I think is nicer anyway. Note 0x3ff = 1024, it is actually 8 + 2 bits.
Codecov Report
@@ Coverage Diff @@
## master #3208 +/- ##
==========================================
+ Coverage 93.05% 93.06% +0.01%
==========================================
Files 376 376
Lines 7412 7428 +16
Branches 192 192
==========================================
+ Hits 6897 6913 +16
Misses 515 515
Continue to review full report at Codecov.
|
| def fix[A](fn: F[A] => F[A]): F[A] = { | ||
| lazy val res: F[A] = defer(fn(res)) | ||
| res | ||
| } |
There was a problem hiding this comment.
This is rather nice! I never knew I wanted this.
johnynek
dismissed stale reviews from kailuowang, LukaJCB, and djspiewak
via
3398700
|
One confusing aspect of cats contributions is that it is not clear when and why things are merged. e.g. two people have approved this PR without comment, but not merged. Is it waiting on a third or is there some other criteria that determines when a PR is merged? |
|
@johnynek sorry about that, I must have forgotten to press the button. |
|
@johnynek Sorry, probably partly my fault—I should have responded with a follow-up 👍. (My understanding is if there are two maintainer sign-offs and no outstanding requested changes you should feel free to merge yourself or ask for a merge.) |
|
Okay, thanks for the merge. I wasn't sure if there was some reason for delay due to staging for releases. |
In my experience, some of the reviewers just drive-by review things and don't loop back to check the merging status. This has made cats reviews a bit of a two phase process, where the second pass is just going through and pressing merge on a bunch of already-reviewed PRs. Honestly I don't think this is entirely problematic so long as the maintainers are doing these phases frequently enough, but it is certainly confusing for contributors. |
6 participants
A pattern I often use in working with scalacheck.Gen or with Parsers is:
It occurred to me all the types I use this pattern with are generally Defer instances, and more over, we can express
fixnicely onDeferinstances since we know we can always return a value (there is no risk of infinite loop at fix application, although it is up to the user if evaluatingF[A]is safe.I think this is a useful function to have on Defer.