Typechecker doesn't preserve HsPar in renamed source.

Hello all, The typechecker doesn't preserve parenthesis that occur at the head of applications. This results in some weird SrcSpans in the TypecheckedSource For example, given code foo a b c = (bar a) b c The typechecker will emit an HsApp with head spanning over `bar a) b` and tail spanning over `c`. Notice that the opening parenthesis is not included. On the other hand, the renamer will generate the expected SrcSpans that always include both parenthesis, or neither. This becomes an issue when you want to associate RenamedSource with its corresponding TypecheckedSource, as the SrcSpans no longer match and overlap partially. This occurs due to this line in TcExpr.hs tcApp m_herald (L _ (HsPar _ fun)) args res_ty = tcApp m_herald fun args res_ty I have a work in progress fix here: https://github.com/wz1000/ghc/commit/3b6db5a35dc8677a7187e349a85ffd51f452452... I have also created a ticket on trac: https://ghc.haskell.org/trac/ghc/ticket/15242#ticket

This looks pretty good to me. What's "in progress" about it? I would want to see a comment on the declaration for HsArgPar with an example, and a test case. Thanks! Richard
On Jun 7, 2018, at 6:38 AM, Zubin Duggal
wrote: Hello all,
The typechecker doesn't preserve parenthesis that occur at the head of applications.
This results in some weird SrcSpans in the TypecheckedSource
For example, given code
foo a b c = (bar a) b c
The typechecker will emit an HsApp with head spanning over `bar a) b` and tail spanning over `c`. Notice that the opening parenthesis is not included.
On the other hand, the renamer will generate the expected SrcSpans that always include both parenthesis, or neither. This becomes an issue when you want to associate RenamedSource with its corresponding TypecheckedSource, as the SrcSpans no longer match and overlap partially.
This occurs due to this line in TcExpr.hs
tcApp m_herald (L _ (HsPar _ fun)) args res_ty = tcApp m_herald fun args res_ty
I have a work in progress fix here: https://github.com/wz1000/ghc/commit/3b6db5a35dc8677a7187e349a85ffd51f452452... https://github.com/wz1000/ghc/commit/3b6db5a35dc8677a7187e349a85ffd51f452452...
I have also created a ticket on trac: https://ghc.haskell.org/trac/ghc/ticket/15242#ticket https://ghc.haskell.org/trac/ghc/ticket/15242#ticket _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

At the time I hadn't modified tcSeq and tcTagToEnum to take HsArgPars into
account. I have now done that, and also added a test case.
I've also submitted the fix to Phab, over here:
https://phabricator.haskell.org/D4822
On 12 June 2018 at 09:04, Richard Eisenberg
This looks pretty good to me. What's "in progress" about it?
I would want to see a comment on the declaration for HsArgPar with an example, and a test case.
Thanks! Richard
On Jun 7, 2018, at 6:38 AM, Zubin Duggal
wrote: Hello all,
The typechecker doesn't preserve parenthesis that occur at the head of applications.
This results in some weird SrcSpans in the TypecheckedSource
For example, given code
foo a b c = (bar a) b c
The typechecker will emit an HsApp with head spanning over `bar a) b` and tail spanning over `c`. Notice that the opening parenthesis is not included.
On the other hand, the renamer will generate the expected SrcSpans that always include both parenthesis, or neither. This becomes an issue when you want to associate RenamedSource with its corresponding TypecheckedSource, as the SrcSpans no longer match and overlap partially.
This occurs due to this line in TcExpr.hs
tcApp m_herald (L _ (HsPar _ fun)) args res_ty = tcApp m_herald fun args res_ty
I have a work in progress fix here: https://github.com/wz1000/ghc/commit/ 3b6db5a35dc8677a7187e349a85ffd51f452452a
I have also created a ticket on trac: https://ghc.haskell.org/trac/ ghc/ticket/15242#ticket _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

Great -- and glad to see this getting appropriate attention over there. Richard
On Jun 11, 2018, at 11:51 PM, Zubin Duggal
wrote: At the time I hadn't modified tcSeq and tcTagToEnum to take HsArgPars into account. I have now done that, and also added a test case.
I've also submitted the fix to Phab, over here: https://phabricator.haskell.org/D4822 https://phabricator.haskell.org/D4822
On 12 June 2018 at 09:04, Richard Eisenberg
mailto:rae@cs.brynmawr.edu> wrote: This looks pretty good to me. What's "in progress" about it? I would want to see a comment on the declaration for HsArgPar with an example, and a test case.
Thanks! Richard
On Jun 7, 2018, at 6:38 AM, Zubin Duggal
mailto:zubin.duggal@gmail.com> wrote: Hello all,
The typechecker doesn't preserve parenthesis that occur at the head of applications.
This results in some weird SrcSpans in the TypecheckedSource
For example, given code
foo a b c = (bar a) b c
The typechecker will emit an HsApp with head spanning over `bar a) b` and tail spanning over `c`. Notice that the opening parenthesis is not included.
On the other hand, the renamer will generate the expected SrcSpans that always include both parenthesis, or neither. This becomes an issue when you want to associate RenamedSource with its corresponding TypecheckedSource, as the SrcSpans no longer match and overlap partially.
This occurs due to this line in TcExpr.hs
tcApp m_herald (L _ (HsPar _ fun)) args res_ty = tcApp m_herald fun args res_ty
I have a work in progress fix here: https://github.com/wz1000/ghc/commit/3b6db5a35dc8677a7187e349a85ffd51f452452... https://github.com/wz1000/ghc/commit/3b6db5a35dc8677a7187e349a85ffd51f452452...
I have also created a ticket on trac: https://ghc.haskell.org/trac/ghc/ticket/15242#ticket https://ghc.haskell.org/trac/ghc/ticket/15242#ticket _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org mailto:ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
participants (2)
-
Richard Eisenberg
-
Zubin Duggal