Re: [Caml-list] [ANN] slacko-0.9.0

[Jacques Garrigue <garrigue@math.nagoya-u.ac.jp>]

On 2014/10/05 07:43, Marek Kubica wrote:
> 
> Hello Gabriel,
> 
> On Sat, 4 Oct 2014 17:23:02 +0200
> Gabriel Scherer <gabriel.scherer@gmail.com> wrote:
> 
>> `kick` now returns a [> error_cant_kick | `Success of 'a ], while
>> `invite` returns [> error_cant_invite | `Success of 'a ].
> 
> Thank you for this idea, I was actually looking for a way to define
> subsets of polymorphic variants, I didn't know that I can just put them
> into other types and was unaware of the # syntax. Now my functions
> return just the possible error variants, not all of them anymore.
> 
> I've reworked the code to get rid of the huge apierror type (which I
> suppose I can throw away completely, since a user of the API will never
> need it anyway), but it lead to quite huge blow-up of signatures:
> 
> <https://github.com/Leonidas-from-XIV/slacko/blob/37c3626bb9574bc99325267fb4c9f9c3c4f4730c/src/slacko.mli>
> 
> I am going to simplify these subsets a bit (since e.g. all history
> methods use the same polymorphic variants and some possible error types
> imply other error types) but it kinda looks very verbose now - any hints
> on what can be done?

I don’t think it’s particularly verbose.
You have documented all the possible errors for each function, which is
valuable information, and it would take the same space (or more) if you
were to write this in comments.
If you think this is hard to read, you can try some factorization, but I’m not
sure it would help. Also, from a stylistic point of view I would write variant
definitions on a single line, since there are never more than 3 cases.

This said, there is always a tension between safety an complexity.
If in practice static checking of possible errors doesn’t matter that much,
it would be easier to write this code using normal variants.

Jacques Garrigue