Bill Allombert on Wed, 17 Mar 2004 15:52:16 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: gp: anal.c more bugs


On Wed, Mar 17, 2004 at 02:31:49PM +0100, Karim Belabas wrote:
> * Bill Allombert [2004-03-17 11:54]:
> > If we are going this way, we need to be more thorough and add the
> > mentionned functions to the interpreter in libpari instead of adding
> > them as GP add-on. This is mainly a matter of removing the Class: gp
> > in the GP function database.
> 
> I have not understood the purpose of this Class: field then. All the routines
> in plotport and highlvl have 'Class: highlevel', just as many others 
> already included in libpari.

Or it is me who misunderstood the purpose of highlevel.h in the
first place. The Class flag determine in which  functions_xxx table the
function must go: 

functions_basic   ../language/init.h 
functions_gp      ../gp/gp_init.h 
functions_highlvl ../gp/highlvl.h

As far as I understand, only functions_basic is known by the interpreter
initially, functions_gp and  functions_highlvl ar GP specific add-on. 
Indeed, gp_init.c call

pari_addfunctions(&pari_modules, functions_highlevel,helpmessages_highlevel); 

So either we move that call to init.c, or we remove it and move the functions  
to the basic class (by removing the Class declaration since basic is the
default).

As for the 'many others already included in libpari', well, it happened
that some of the functions in highlevel were progressively moved to
libpari, but here we are doing things more consistently.

Cheers,
Bill