Thanks Bruno, for kicking off. I have had meetings all week, and worked on the forthcoming release of Generic Haskell. I'll leave for holidays for two weeks tomorrow, so I won't join the discussion until the week of October 23 again.
1) While discussing the different approaches, we should also develop a "template" for reporting our conclusions. I think that having a base template will make our life easier. At the moment I am using the following:
- Approach (Just the name of the approach) - Required features (What do we need) - Usage (how to use it, what different users do we have and what knowledge do they need) - Extensibility (is it extensible, if not can we do anything about it?) - Performance considerations (what are the performance impacts, is it possible to improve?) - Expressibility (what kind of generic functions can we write? producer and consumers? different arities? local redefinition?) - Boilerplate (what is the boilerplate involved, is it possible to remove it?) - Helpful extra features (what could Haskell have that would make the approach easier to use? Are those features realistically feasible?) - Discussion (General discussion, advantages/disadvantages, comparasion with other approaches)
but note that this is not, by any means, supposed to be a proposal for the final template. It is just my initial template. We probably should discuss the template as we go along and tune it (if you think it is a good idea to have it in the first place).
I definitely think it is a good idea to have such a template. I think you cover all relevant aspects. I recently thought about such a template again, and also included: - Subset of data types covered. Most proposals restrict the subset of data types on which generic functions can be applied. The restrictions vary for the different proposals. This is related to expressiveness, but different enough to warrant a special category, I think. Furthermore, I called Boilerplate `Amount of work per data type', at least, that is what I think you mean with boilerplate? The name boilerplate might be a bit confusing: what is the boilerplate of SYB (nothing! of course (well, except for deriving Typeable), bust still, I suppose you get my point). Something I'd also like to look at is: what are the kind of type- error messages you get when you make a mistake in your generic function? How `comprehensible' is the program (whatever that means, and it probably means different things to different people)? At least size plays a role here, but also other aspects. And is it easy to prove properties of generic programs? Finally, I called required features portability, but there is something to say for both.
2) Johan also mentioned a darcs/svn repository for the resulting paper/document. I think this is a great idea. Is anyone taking care of this? It would also be a good idea to use the repository to store some code for the different approaches. This way people who would like to experiment with code, would not have to repeat the task individually. What do you think?
Good idea. The darcs repository seems like a good idea. Manuel offered to creat one.
3) Related to 2), we could start by collecting code for the approaches from the authors of the papers. I guess most of them are in this mailing list anyway. For example, we could now ask Ralf Hinze or James Cheney (either one of you two reading this email ?) if they some code available for LIGD that they could submit. We could then add this code to darcs/svn and do some experiments with it: organize it like a library, define generic functions, etc. If no code is available, perhaps we should find a volunteer that implements the code (if that's the case for LIGD I can volunteer for it).
Ralf implemented some function in LIGD for our comparing paper, I attach it to this message (together with a file Xbits that is used somewhere; maybe there is a standard Haskell library for converting strings into bits and vv? These should appear in the Darcs repository when it is there. Note that the code is better viewed on paper then as code).
====================================================================== = Approach: A Lightweight Implementation of Generics and Dynamics
Required features: - Haskell 98 + existential types
Usage: - Library Writer: Responsible for the sums-of-products machinery, the representation(s) data type(s) to be defined in the library and defining the class and standard instances for "Representable". The library writer needs to be have a deep knowledge about generic programming. - Power User: Defines the boilerplate for new data types. Needs to be fairly familiar with the implementation. - User: defines generic functions by pattern matching on the structure of "Rep". This requires that he is comfortable with sums of products and isomorphisms. For convenience, the user should also provide a definition that uses the type class "Representable", which avoids the representations to be passed explicitly. - End User: If "Representable" is used, then the end can just use generic functions in the same way as any other ad-hoc (that is, type class overloaded) functions.
Extensibility: - Data types are non-extensible, however we can try to combine the approach with type classes to get some form of extensibility (RepLib proposes one solution like this).
What do you mean with extensibility here? Not data types I assume. You mean that LIGD functions are not extensible for new data types? This is a limitation of course. Maybe a more severe limitation is that generic functions can only be defined on types defined in Rep, we cannot add non-generic behaviour for particular data types. Open data types will solve this, but these do not exist in any Haskell implementation.
Performance considerations: - Sums of products based and therefore involving considerable performance penalties.
And types are passed at run-time, and interpreted. Sums of products need not incur considerable performance penalties if partial evaluators remove those intermediate layers. But we have to experiment with behaviour here. Passing and interpreting types at run- time is harder to avoid, I think.
Expressibility: - Can do both producer and consumer functions. - Multiple representations needed for generic functions of different arities; - No local redefinitions.
Boilerplate: - Isomorphisms between sums of products and data types - Instances of Representable - Smart constructors for representation of data types
Helpful extra features: - GADTs (already implemented in GHC) allow more direct definitions of generic functions, since there is no need to apply the type coercions explicitly; - A boilerplate generation mechanism. This would effectively mean that power users would not be necessary. - Open data types and functions in order for extensibility. - Kind polymorphism, for eliminating the need for multiple representations;
Discussion: An interesting approach which is quite lightweight and that is fairly easy to use (specially if we would not need to handle any of the boilerplate). It is a bit outdated because with GADTs, the use can be slightly simplified. However, this also takes us further away from Haskell 98 or even Haskell' (since GADTs will not be there). I think that LIGD is still a reference for a generic programming library that uses a data type for type representations and sums of products as a family of data types. A drawback may be performance, since we need to convert a data type into its isomorphic sum of products representation. =============================================================
I think LIGD is quite a nice library, that needs almost no Haskell extensions. But it has its limitations. -- Johan