Does it depend on Stratego/XT? As far as I know I have to live in the Eclipse environment to use anything depending on Stratego/XT. At 2014-04-08 04:53:06,"Tillmann Rendel" wrote:

Hi,

Ducis wrote:

...

Actually I wonder how to achieve that without rolling my own haskell parser or forking haskell-src-exts.

Adam vogt wrote:

...

Ideally there could be something like camlp4 for haskell, but as far as I know there isn't such a thing.

We built SugarHaskell to explore that design space. See our paper at the Haskell Symposium 2012:

http://sugarj.org/sugarhaskell.pdf

While there is not much going on about SugarHaskell specifically, the underlying language-independent framework is still actively developed, in the https://github.com/sugar-lang organization. The eclipse plugin for SugarHaskell from http://update.sugarj.org/ should work. I don't know what the status of the command-line version is, though.

Even if you don't want to use all of SugarHaskell, the underlying easily extensible grammar might still be a good starting point. It uses declarative layout constraints to specify Haskell's layout in a modular way. See our paper at the conference on software language engineering 2012:

http://sugarj.org/layout-parsing.pdf

Tillmann

So, it means that using Stratego outside of Eclipse is not that hard? Cool. I've heard of Stratego before. It's time to try again. Thanks for the information. At 2014-04-08 20:37:04,"Tillmann Rendel" wrote:

Hi,

Ducis asked about SugarHaskell:

...

Does it depend on Stratego/XT?

SugarHaskell uses Stratego and SDF, but all necessary tools are bundled, so you only need a Java virtual machine (and maybe a Java compiler, not sure). Most of the tools are bundled as Java class files. Two tools (sdf2table and implodePT) are bundled as a statically linked executable for common platforms. On a non-common platform, you would have to compile these from their C sources.

There is a release of SugarHaskell on hackage:

http://hackage.haskell.org/package/sugarhaskell

The tarball consists mostly of *.jar files and native executables. Looking at the timestamp, this is obviously outdated, but it should work and give you a first impression. I'm sure an updated release on Hackage could be provided.

(Eclipse is the main platform for SugarHaskell and the other Sugar languages, because we want to provide extensible syntax highlighting and other editor features. The process for making the Hackage release was to write a simple driver for the command-line interface that essentially calls java with the right parameters, and figure out which *.jar files from an Eclipse installation we need to bundle to make it work. The latter was done manually, and could be done manually again for a more recent SugarHaskell-in-Eclipse installation).

Tillmann

I just want to add that I find the syntax extremely confusing and counter intuitive. I thought it was just me, or that I was missed something. But it looks like I'm not the only one. On 7 April 2014 21:58, Kim-Ee Yeoh wrote:

On Mon, Apr 7, 2014 at 11:02 PM, ducis wrote:

...

It lets your write lambdas with 'slots' without inventing names for the parameters.

[s| ı + ı |] = \x y -> x+y

I have no background in this 'slot lambda' and a search reveals this package as the only hit.

Which may explain why I find the example given confusing. Why would [s| 1+1 |] not be equivalent to \x->x+x ?

-- Kim-Ee

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

-- *Alois Cochard* http://aloiscochard.blogspot.com http://twitter.com/aloiscochard http://github.com/aloiscochard

Thanks. [s|_0 + _1|] = \x y->x+y is supported. I simply like [s|_+_|] better personally. And [s|_+_ı|] is for \x->x+x. I have to admit that 'ı' character is not very recognizable though. Any suggestions of another 'Letter, Lowercase' unicode character that no one would use as a variable name and can be easily input in vim are welcome. At 2014-04-08 05:12:26,"Alois Cochard" wrote: In order to give a more constructive feedback, why not simply indexing arguments by position and use the `_N` syntax? The second example would be: _0 : _1 : _1 : _2 : _2 : _2 : _0 : [] One could just use `_` if there is only one arg: _ + _ (which mean what Kim thought it would mean) and use: _0 + _1 (for the case you wanted to support) Cheers On 7 April 2014 22:01, Alois Cochard wrote: I just want to add that I find the syntax extremely confusing and counter intuitive. I thought it was just me, or that I was missed something. But it looks like I'm not the only one. On 7 April 2014 21:58, Kim-Ee Yeoh wrote: On Mon, Apr 7, 2014 at 11:02 PM, ducis wrote: It lets your write lambdas with 'slots' without inventing names for the parameters. [s| ı + ı |] = \x y -> x+y I have no background in this 'slot lambda' and a search reveals this package as the only hit. Which may explain why I find the example given confusing. Why would [s| 1+1 |] not be equivalent to \x->x+x ? -- Kim-Ee _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe -- Alois Cochard http://aloiscochard.blogspot.com http://twitter.com/aloiscochard http://github.com/aloiscochard -- Alois Cochard http://aloiscochard.blogspot.com http://twitter.com/aloiscochard http://github.com/aloiscochard

On Tue, Apr 08, 2014 at 03:58:30AM +0700, Kim-Ee Yeoh wrote:

On Mon, Apr 7, 2014 at 11:02 PM, ducis wrote:

...

It lets your write lambdas with 'slots' without inventing names for the parameters.

[s| ı + ı |] = \x y -> x+y

I have no background in this 'slot lambda' and a search reveals this package as the only hit.

Which may explain why I find the example given confusing. Why would [s| 1+1 |] not be equivalent to \x->x+x ?

Because each slot stands for a new variable.

On Mon, Apr 7, 2014 at 5:14 PM, Alois Cochard wrote:

Understood and actually I forgot you can not use "_", so why not a symbol which don't look like a number?

I was also wondering how Turkish speakers/writers would react to that notation.... -- brandon s allbery kf8nh sine nomine associates allbery.b@gmail.com ballbery@sinenomine.net unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

:-) Ok, nevermind. That's just me which is has no understanding of the context then, like Kim. I'm booking some classes with Frege and Russell, looks like they are pretty busy atm. On 7 April 2014 22:57, Alexander Solla wrote:

On Mon, Apr 7, 2014 at 2:26 PM, Brandon Allbery wrote:

...

On Mon, Apr 7, 2014 at 5:14 PM, Alois Cochard wrote:

...

Understood and actually I forgot you can not use "_", so why not a symbol which don't look like a number?

I was also wondering how Turkish speakers/writers would react to that notation....

Native English speaker here. My first thought was, "Oh, a iota. And he's using it as a definite descriptor/anonymous variable. Nice."

But then again, I took too many classes on Frege and Russell.

-- *Alois Cochard* http://aloiscochard.blogspot.com http://twitter.com/aloiscochard http://github.com/aloiscochard

No, there is no theory behind it. I personally feel it convenient. 在 2014-04-08 04:58:30，"Kim-Ee Yeoh" 写道： On Mon, Apr 7, 2014 at 11:02 PM, ducis wrote: It lets your write lambdas with 'slots' without inventing names for the parameters. [s| ı + ı |] = \x y -> x+y I have no background in this 'slot lambda' and a search reveals this package as the only hit. Which may explain why I find the example given confusing. Why would [s| 1+1 |] not be equivalent to \x->x+x ? -- Kim-Ee