Comments on A Neighborhood of Infinity: Programming with impossible functions, or how to g...

Your coroutines can be extended with choice (a la ...

2009-12-23T10:43:20.111-08:00

Your coroutines can be extended with choice (a la session types) with a few more lines of code:

readOpt f g (Left (c,cs)) = second Left $ f c cs
readOpt f g (Right (c,cs)) = second Right $ g c cs
writeRight o r is = let (v, os) = r (fromRight is) in (v, Right (o, os))
writeLeft o r is = let (v, os) = r (fromLeft is) in (v, Left (o, os))

Great stuff! Interestingly, I just also found out...

2009-12-22T10:19:36.398-08:00

Great stuff!

Interestingly, I just also found out about Bart Jacob's co-algebraic view on object-oriented programming. See for example his papers ...

Objects and classes, co-algebraically

Introduction to Coalgebra. Towards Mathematics of States and Observations

Just came back as I encountered this talk: http:/...

2009-12-09T22:18:40.326-08:00

Just came back as I encountered this talk:

http://blog.well-typed.com/2009/12/talk-at-the-functional-programming-exchange/

which also does something very reminiscent of the above.

Must be something in the air these days.

Dear Author blog.sigfpe.com ! It is possible and ...

2009-12-03T19:26:02.715-08:00

Dear Author blog.sigfpe.com !
It is possible and necessary :) to discuss infinitely

Looking forward to it!

2009-11-29T23:51:11.762-08:00

Looking forward to it!

I've read that exceptions and continuations ar...

2009-11-29T23:24:54.358-08:00

I've read that exceptions and continuations are not algebraic effects. It would explain why Haskell need Monad and Lawvere Theories are not enough.

But to be honest, I don't yet really understand it. The paper http://homepages.inf.ed.ac.uk/gdp/publications/Effect_Handlers.pdf is the cause of my confusion and I'll have to read it.

In Plotkin papers, effects are related to constructors and handlers (not algebraic) to deconstructors.

I wonder if there is a relation with coalgebra. I am just beginning to study effects so I have more questions than answers.

Janis, Correction: Swiestra handles the free mona...

2009-11-29T13:35:16.252-08:00

Janis,

Correction: Swiestra handles the free monad case. I intend to handle the non-free case.

Janis, Thanks! Those papers look awesome. My next...

2009-11-29T08:38:20.474-08:00

Janis,

Thanks! Those papers look awesome. My next blog post is going to be quite similar to Swiestra's "Monads for Free" section, except I plan to answer the bit where he says "It is not immediately obvious what the type of our algebra should be" slightly differently.

Regarding the idea of modelling effects by first c...

2009-11-29T03:23:42.863-08:00

Regarding the idea of modelling effects by first constructing ASTs of an embedded language and then writing a small interpreter, you may want to consider this paper:
http://doi.acm.org/10.1145/1291201.1291206

I think some of your code appears there almost literally.

And (to plug my own work :-)) regarding the idea of afterwards eliminating the ASTs and interpreter overhead, you may want to look at this:

http://dx.doi.org/10.1007/978-3-540-70594-9_20

(It's a generic treatment, but Section 4 applies it to the "monad specification setting".)

Finally, regarding how to nicely combine sublanguages for different effects, you may want to look at:

http://dx.doi.org/10.1017/S0956796808006758

Happy reading. :-)

Chreighton, yes, and "so called" can als...

2009-11-28T22:33:52.568-08:00

Chreighton, yes, and "so called" can also come across as "I'm an out of touch fuddy-duddy" - for example, when a judge writes something like "So-called computer "codes" direct the operation of the computer".

Colimits are the general categorical way of sticki...

2009-11-28T22:01:28.544-08:00

Colimits are the general categorical way of sticking things together.

Ooh, awesome. I actually hadn't read that pap...

2009-11-28T19:17:20.200-08:00

Ooh, awesome. I actually hadn't read that paper, but when Mr. Sigfpe said "quotient" my first thought was colimit. It looks like the coproduct isn't particularly useful in & of itself, but it's a nice proof of concept that colimits are the notion you want for combining monads.

Not sure how to "pullback" colimits on free monads of signatures to more primitive operations.

Also, this paper reminded me that one just shouldn't use the phrase "so called" in an academic work. Despite intentions, it sounds dismissive.

> an alternative universe in which languages li...

2009-11-28T18:19:37.665-08:00

> an alternative universe in which languages like Haskell develop along a path in which monads don't play a starring role.

http://en.wikipedia.org/wiki/Standard_ML

This paper (which also has Ghani as a coauthor) ma...

2009-11-28T17:21:38.165-08:00

This paper (which also has Ghani as a coauthor) may be relevant, they give a construction of a monad coproduct which is a quotient:

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.8.3581

Creighton, That sounds a bit like where I got to....

2009-11-28T14:02:46.856-08:00

Creighton,

That sounds a bit like where I got to. I can easily write code that takes pairs of sets of signatures, combines them, and makes you a monad. It works perfectly well with do-notation. But it's "too big" and you need an "interpreter" to cut it back down to size, ie. what you really want is a quotient.

There's a big caveat, imo, with that paper: y...

2009-11-28T13:50:58.893-08:00

There's a big caveat, imo, with that paper: you can combine the theories, but relating the monad made from them to the monad you morally _want_ is a little trickier.

It wasn't clear to me how that gets resolved.

I'll check out that paper, at first glance it ...

2009-11-28T13:43:16.442-08:00

I'll check out that paper, at first glance it looks like exactly what I want. I searched hard for papers with concrete examples but everything I found was pretty abstract.

http://www.cs.nott.ac.uk/~gmh/modular.pdf You prob...

2009-11-28T13:26:39.715-08:00

http://www.cs.nott.ac.uk/~gmh/modular.pdf
You probably would enjoy reading the above paper. It's a bit about how to implement the combination of signatures together.

It's a bit...ungainly, but seems relevant.

Now the thing that's really cute, and what ties in to Conal Elliott's Type Class Morphism paper, is that an actual implementation of a signature, a model, corresponds to a functor & morphisms between models correspond to natural transformations. Conal was looking at it more from the functorial point of view, you're starting a bit more directly from models.

Either way, there's an equivalence between the perspectives (in the categoric sense).