Jon Harrop schrieb:

On Friday 19 February 2010 18:46:30 Tom Tobin wrote:

...

On Fri, Feb 19, 2010 at 1:57 PM, Jon Harrop wrote:

...

GHCI seems to require type definitions to be described in a separate file and then loaded into the REPL, which is seriously tedious.

Are there any Haskell REPLs that don't have such limitations and, preferably, use JIT compilation instead of an interpreter? I run Haskell in an inferior shell in Emacs alongside haskell-mode; it's easy to get immediate feedback to my changes that way.

What's your procedure for getting ghci to interpret a type definition?

Typing ":r" to reload the currently loaded but later changed or extended file. C.

On Sat, Feb 20, 2010 at 9:31 AM, Jon Harrop wrote:

On Friday 19 February 2010 18:46:30 Tom Tobin wrote:

...

On Fri, Feb 19, 2010 at 1:57 PM, Jon Harrop wrote:

...

GHCI seems to require type definitions to be described in a separate file and then loaded into the REPL, which is seriously tedious.

Are there any Haskell REPLs that don't have such limitations and, preferably, use JIT compilation instead of an interpreter?

I run Haskell in an inferior shell in Emacs alongside haskell-mode; it's easy to get immediate feedback to my changes that way.

What's your procedure for getting ghci to interpret a type definition?

I just hit C-c C-r in my Haskell source file's buffer to get it reloaded in the inferior ghci shell. I think it's best to think of the source file as your scratchpad for editing, with the "main" function changed as you go along to show relevant output for testing, and the inferior ghci buffer as just an "output window" of sorts. It's not the same as writing code directly shell, but I actually prefer it this way for all languages now, not just Haskell -- particularly since it's much nicer editing code in a proper editor vs. in a shell.

On Sat, Feb 20, 2010 at 1:44 PM, Jon Harrop wrote:

On Saturday 20 February 2010 16:33:30 Tom Tobin wrote:

...

I just hit C-c C-r in my Haskell source file's buffer to get it reloaded in the inferior ghci shell.

I think it's best to think of the source file as your scratchpad for editing, with the "main" function changed as you go along to show relevant output for testing, and the inferior ghci buffer as just an "output window" of sorts.  It's not the same as writing code directly shell, but I actually prefer it this way for all languages now, not just Haskell -- particularly since it's much nicer editing code in a proper editor vs. in a shell.

I see. What happens if your program is producing non-trivial output that takes a long time to compute? How do you get around complete recomputation at the evaluation of every new expression?

Hm, I'm not sure. I do seem to lose any local "let" bindings I do in the ghci window when I reload, and the manual for ghci doesn't seem to offer any way to save values between reloads [1]. I know you can avoid that in other languages' shells (e.g., Python). One workaround I can think of is to use the "binary" package to save intermediate computations out to disk, but that seems somewhat cumbersome when you're trying to quickly prototype your code. [1] http://www.haskell.org/ghc/docs/latest/html/users_guide/interactive-evaluati...