Re: [Caml-list] [ANN] ppx_deriving 0.1

[Peter Zotov <whitequark@whitequark.org>]

On 2014-07-24 00:37, Yotam Barnoy wrote:
> OK, I was just making some suggestions based on experience I have with
> haskell's deriving, which I use all the time, and trying to bring as
> much of that ease of use as possible to ocaml.

Thank you for the suggestion. I see some cases where modules would be 
useful,
(including, eventually, implicits), but as most deriving plugins 
generate
a single function now, I think it would be pointless to generate such 
modules.

>  Any idea where I can find some info on ocaml's upcoming implicits? A
> google search was not particularly helpful.

There are two articles by Alain Frisch:
http://www.lexifi.com/blog/implicit-values
http://www.lexifi.com/blog/implicit-arguments

They get the general idea across, but do not represent the ongoing work.
The ongoing work is being done in this git repository:
https://github.com/def-lkb/ocaml/commits/implicit-module

As far as I know, it is not yet documented in any way and it is not 
clear
whether it will be merged in OCaml at all.
However, I have high hopes for this feature.

> 
> On Wed, Jul 23, 2014 at 4:19 PM, Peter Zotov
> <whitequark@whitequark.org> wrote:
> 
>> On 2014-07-23 18:36, Yotam Barnoy wrote:
>> 
>>> Very nice.
>>> 
>>> I've never used any of the deriving extensions before, but I have
>>> an
>>> aesthetic suggestions.
>> 
>> I humbly suggest trying to use it before proposing aesthetic
>> changes.
>> 
>>> Would it perhaps make sense to generate a
>>> module per derived type? For example a type t would generate a
>>> module
>>> T_ (the underscore or any other suffix would reduce mixups with
>>> pre-existing modules). You could then use code such as
>>> 
>>> 'if T_.(a = b && b = c) ...'
>>> 
>>> or 'T_.show x ...'
>>> 
>>> which allows you to keep the infix notation for = which is
>>> important
>>> IMO.
>>> 
>>> You could even generate T_ as having internal Eq, Ord, and Show
>>> modules (as requested by the user), which would be included in
>>> the T_
>>> module. This would allow you to easily pass them as first class
>>> modules to functions accepting Eq, Ord or Show signatures.
>> 
>> I believe this is best left to the upcoming implicits, which will
>> hopefully be merged soon.
>> 
>> I have based the current design on existing patterns across OCaml
>> ecosystem; I don't want to change the way people structure their
>> modules, I want to reduce the amount of boilerplate to write.
>> 
>> --
>> Peter Zotov
>> sip:whitequark@sipnet.ru

-- 
Peter Zotov