ndmitchell
Avoid isJust/fromJust
From https://code.google.com/p/ndmitchell/issues/detail?id=238 by @np:
In most situations code like:
if null lst then ... else ... head l ... tail l ...
could be replaced by a case
case lst of [] -> ... ; x : xs -> ... x ... xs ...
Same thing with isJust + fromJust, but maybe could be suggested instead of a case.
@ndmitchell replies:
if isNothing x then 0 else fromJust x + 3
Should suggest:
maybe 0 (+3) x
This relies on having a nice looking functional in the else. Alternatively could have patterns with context, i.e.
if isNothing x then y else _f x ==> maybe x _f x
where _f is the lambda that needs inserting to get it working.
I have the line
warn "fromJust" = fromJust ==> fromMaybe (error "location of call")
in my local hlint config. I prefer total functions and thus would always replace fromJust by a complete case analysis.
I prefer total programs, and would generally rather the compiler proved that for me, rather than trying to approximate it with complete case analysis - but I appreciate many people do go that route. The fromJust to fromMaybe pattern is only useful for implementations of Haskell which don't have stack handling, which I think the latest GHC will have (at least for fromJust).
I see the Foldable.sum version as more complex than the maybe, and also more special purpose, since it only works for numerical operations. Was that just one example, or a suggestion of a hint?
Just found a case of isJust $ a <$> b and I was wondering whether HLint had an option for that. I guess that isJust . fmap is a special case of what is being suggested here.
@ysangkok - I don't think that's a special case of this - there actually already is a hint for what you want - https://github.com/ndmitchell/hlint/blob/master/data/hlint.yaml#L623 - just no variant with <$>. I'd happily accept a patch to add that.