Re: term_expansion/2 with DCG clauses

On 14/08/2013, at 10:24, Jan Wielemaker <[email protected]> wrote:

> On 08/14/2013 04:28 AM, Richard A. O'Keefe wrote:
>> 
>> On 14/08/2013, at 3:44 AM, Jan Wielemaker wrote:
>>> This seems a logical way to access DCG expansion. The only price is that
>>> you cannot created rules for -->/2 (nor for []/0 and ./2). Do these
>>> systems have an agreed way to creates DCGs from expansion rules? I guess
>>> Richard will know the rationale for not doing DCG after user expansion.
>> 
>> If I recall correctly, DCG expansion was done *before* user expansion.
>> With there being no goal_expansion in those days, this was important if
>> you wanted users to be able to plug their own goal_expansion code in and
>> have it apply to the result of DCG expansion.
>> Further, DCG expansion was available as library code, so if you wanted it,
>> you could very easily call it yourself.
> 
> So not doing DCG expansion if user expansion is involved seems out of
> date.

I think the interesting question here is where and how the user can intervene if a call to term_expansion/2 produces a grammar rule (or a list of grammar rules). 

> The idea to chain `translation packages' is also implemented in
> Ciao and ECLiPSe (and Logtalk?), although the way the packages are
> specified there is much more fine-grained.

In the case of Logtalk, the user can define his/her own chaining. The fundamental difference is that the user explicitly states (using a directive in a source file or a compiler option in the predicate that compiles and loads a source file) which object contains the expansion rules to be used. In this object the user can define if and how the chaining works.

> If you want DCG translation and do further term expansion on the result,
> you can call dcg_translate_rule/2. Would be nice if this had the same
> name in all systems supporting DCGs ...

That was one of my goals when I served as the editor of the ISO WG17 DCGs standardization proposal, backed by a reference implementation (and a set of unit tests). You know what happened.

> As I understand it, some people
> assume you can call expand_term/2 to achieve DCG translation. There are
> two issues with that: it might do a lot more and it is probably a bad
> idea to call expand_term/2 from term_expansion/2.  It surely easily gets
> you into a loop.
> 
> I see no reason to change anything right now.

I agree. It will be a long time until any sort of de facto standard expansion mechanisms emerges.

Cheers,

Paulo


-----------------------------------------------------------------
Paulo Moura
Logtalk developer

Email: <mailto:[email protected]>
Web:   <http://logtalk.org/>
-----------------------------------------------------------------