batterseapower / ansi-wl-pprint Goto Github PK

And remove the link from the home page.

e.g.:

in

"a" </> "+" </> + "b" </> " = " </> "c" </> "/" </> "d"

one of the soft breaks around equals sign should have higher priority, so that

a + b
= c / d

or

a + b =
c / d

is printed instead of

a + b = c
/ d

Citing from http://hydra.cryp.to/build/626491/nixlog/2/raw:

ext/PrettyPrint/ANSI/Leijen.hs:312:24:
    Ambiguous occurrence ‘<$>’
    It could refer to either ‘Text.PrettyPrint.ANSI.Leijen.<$>’,
                             defined at Text/PrettyPrint/ANSI/Leijen.hs:372:3
                          or ‘Prelude.<$>’,
                             imported from ‘Prelude’ at Text/PrettyPrint/ANSI/Leijen.hs:76:8-35
                             (and originally defined in ‘Data.Functor’)

Currently, the only way that Union is exposed is through group. However, group can only fiddle with white space. There are cases where this is not good enough. For instance, pretty printing a Haskell case statement:

case e of C1 x -> e1 ; C2 x y -> e2 ; ...

vs.

case e of
  C1 x -> e1
  C2 x y -> e2
  ...

Line breaks must turn into semicolons when trying to lay out on a single line.

A union combinator might allow people to break the invariant, but I'm not sure how to better handle arbitrary new functionality of this type, and it'd be nice to have.

Could there be support for 256 colors?

..., so that less (at least on OS X) and tail would render consecutive lines correctly

My most recent stackage builds just had a bunch of test suite failures with the following text:

         Test Cases     Total         test-async: Text/PrettyPrint/ANSI/Leijen.hs:(1039,1)-(1053,39): Non-exhaustive patterns in function plain

This seems to have been triggered by usage of test-framework. This is based on version 0.6.7 of the package.

I don't think many programs break if <> changes from Doc -> Doc -> Doc to Monoid m => m -> m -> m, since everything else in the library has type Doc. The fixity is the same between the two versions.

This change will save people from having to write hiding ((<>))