Fix tuple patterns #102
Fix tuple patterns #102
Conversation
Instead of map (\(NewLine x ...) -> x) . filter isNewLine introduce prisms _TokenName and _NewLine and write mapMaybe _NewLine (also, reuse trimNewlines where relevant)
PR MarcWeber#99 attempted to fix the handling of multiple definitions by way of tuple unpackings -- eg (x,y,z) = ... and (x, y) = undefined Unfortunately, this was an edge case underexercised in the test suite, so the bugs in MarcWeber#99 weren't caught -- we were lucky it also broke the handling of operators, so 9.hs caught it. Moreover, the comma splitting code was impenetrably obscure, rewrote it as a simple unfold. Closes: MarcWeber#101 Fixes: MarcWeber#99
- Handle inexhaustive pattern matches:
- tail -> drop 1
- map head . groupBy -> map D.L.NonEmpty.head . D.L.NonEmpty.groupBy
- Replace unnecessarily partial conditions by function guards, list
comprehensions, etc
- Removed unnecessary imports
|
CI fails for 8.0 and 8.2 because of the Semigroup incompatibility. |
As recommended by @andreasabel, supporting those versions means needing to bridge the "Semigroup as superclass of Monoid" proposal. Unfortunately, pushing this commit stalled long enough that the CI logs were wiped, so it is hard to assess the cost of supporting those old versions.
|
In cleaning up my outstanding bugs list, I noticed I hadn't addressed this. Apologies. Unfortunately, enough time has elapsed that the CI logs have been wiped, so I can't assess how hard it'd be to bridge the Semigroup/Monoid change. Moreover, |
You need to install |
2 participants
#99 attempted to fix multiline tuple patterns, but in the process broke some test. In fact, it turns out tuple patterns were underexercised in tests, so we were lucky to catch this (it also broke the handling of prefix operator definitions). Moreover, the logic in that PR was hard to understand in places.
This PR fixes this by replacing the tuple-splitting logic by a simple
unfoldsplitting on delimiters.Also, to silence GHC warnings, it introduces two prisms
_TokenNameand_NewLineto use instead of egmap (\(NewLine x ...) -> x) . filter isNewLine.Closes: #101
Fixes: #99