GHCi :edit command should jump to the the last error

[martijnislief]

GHCi has a :edit command for editing the current source file. Hugs's command of the same name automatically jumps the editor to the location of the last error; it would be nice if GHCi's did the same.

[simonmar]

I implemented :edit, but without the line-number-jumping facility for now.

[igloo]

Punting on this.

The only problem with fixing it is that we've forgotten where the first error was by the time :e is given; shouldn't be too hard to fix, though, by updating a mutable variable when appropriate.

[claus]

starting somewhere in 6.7, you can define something like that yourself. the trick is to have a :redirErr command that allows you to capture ghci command error output. then you do a :l or :r, capture the errors, and use :e to edit the first one.

a slight complication is that :l/:r invalidate all variable bindings, so we need to define a temporary :afterCmd, in order to restore stderr and to read the errors after the :l/:r.

please note that many editors support quick compile/edit cycles (in vim, try :help quickfix; emacs should also support something similar), which will do this kind of thing more easily in most cases.

add the attached le.ghci to your .ghci, or source it with :cmd readFile "le.ghci". then, :le <mod> will :load and :edit, and :le on its own will :reload and :edit. adapt the code to your needs.

there is a patch pending for ghc head that will allow us to avoid having to compress ghci definitions into single lines, which will make things like this a lot more readable and maintainable..

[claus]

example .ghci script implementing :le (load/reload and edit)

[guest]

actually, it turns out you can define all this even back in 6.4.1 (had i only known back then!-). you need to define your own :e command (:def is available), and ghci 6.4.1 is somewhat annoying about command names that are prefixes of each other. have a look at

 http://www.cs.kent.ac.uk/people/staff/cr3/toolbox/haskell/dot-squashed.ghci641

 http://www.cs.kent.ac.uk/people/staff/cr3/toolbox/haskell/dot-squashed.ghci

since .ghci files are not very readable with definitions squashed into single lines, you might want to check the explanations of the latter file in

 http://www.haskell.org/pipermail/haskell-cafe/2007-September/032260.html

(note: the names in the 6.4.1 version differ from those in the described version)

i'm not yet sure whether being able to define :e means that it shouldn't be part of ghci at all (and this ticket be closed), or whether there are still any advantages to having it built in? perhaps ghci releases should have a selection of .ghci files with useful definitions?

[simonmar]

You can define it yourself, but it's a bit hacky and it would clearly be better to have it built-in. This is also a small self-contained task (limited to the GHCi front end) that a new contributor could take on.