Oh, I mean, if you take look at Haddocks over there, goal is to document it as much as possible
It's docs should read like a developer guide (just today I wrote two big notes about handling of names and unicode)
Currently I'm trying to decide on conflict behaviour of - - I wonder whether it wouldn't make sense to simply parse stream of "atomic" expressions and use traditional algo on it instead of building some weird tree with wrong fixities as GHC does
Hmm, Haskell2010 spec says that (https://www.haskell.org/onlinereport/haskell2010/haskellch10.html#x17-17700010.2):
But when I try e.g.
it responds with
:joy: : https://gitlab.haskell.org/ghc/ghc/-/blob/master/compiler/GHC/Parser/Lexer.x#L131
what caused you to try that?
just wondering
Hask
the toy compiler
Warning - writing Haskell compilers causes abuse of GHC :big_smile:
stalker!
See #Hask :smile:
there's an entire stream for it here!
what he said
I would guess that's the stream for discussing the category xP
because you don't pay attention!
did Mat announce it somewhere around here?
think so!
Hmm, can't find any announcement, but I've mentioned the original repo to few people: https://github.com/TheMatten/hask
I'm seriously considering turning this into bachelor thesis, because it's fun :big_smile:
"Writing Haskell compiler using modern Haskell"
maybe y'all can finish "write you a haskell"
Oh, I mean, if you take look at Haddocks over there, goal is to document it as much as possible
It's docs should read like a developer guide (just today I wrote two big notes about handling of names and unicode)
Alright, so should I file a bug? I mean, GHC ignores more valid characters than it accepts in this case :stuck_out_tongue:
So many missed opportunities to use segmented digits :big_smile: :
This time, spec does not follow
GHClogic - why is⋆
isascSymbol
when it's not ASCII? :big_smile:(https://www.haskell.org/onlinereport/haskell2010/haskellch10.html#x17-17700010.2)
the
asc
actually stands for ascendingHS2010 spec says that
, but wouldn't that mean that lambda can appear on left side of an operator, without parens?
Maybe they meant
aexp
?well, it's not like you can check if it can appear
because when it's parensless it scopes all the way to the right
or are you saying this would cause a conflict in the grammar?
Hmm, wait:
right :big_smile:
Hopefully it'll be enough to put
%shift
somewhere :sweat_smile::sweat_smile:
I think this is more of a fixity issue
precedence, that is
I used to struggle with this, but apparently it works fine with lambda in lexp
Currently I'm trying to decide on conflict behaviour of
-
- I wonder whether it wouldn't make sense to simply parse stream of "atomic" expressions and use traditional algo on it instead of building some weird tree with wrong fixities as GHC doesPrecedence of do block?
well the lambda binds less strongly than any operator, or something
not quite sure why my grammar does it right without any precedences if I think about it
Ah, so lambda as right associative prefix operator?
that's how I did it at first iirc, but now there's no associativity in that rule
I think tree-sitter might just choose the longer match in this case