Re: [PATCH] for strports.c: scm_c_eval_string_from_file_line

[Marius Vollmer]

Bruce Korb <address@hidden> writes:

> Marius Vollmer wrote:
> 
> >   void
> >   scm_c_primitive_load_from_string (const char *str,
> >                                     const char *filename, int line)
> >   {
> >     SCM port, exp;
> > 
> >     port = scm_open_input_string (scm_str2string (str));
> >     scm_set_port_file_name_x (port, scm_str2string (filename));
> >     scm_set_port_line_x (port, scm_int2num (line));
> > 
> >     while (!SCM_EOF_OBJECT_P (exp = scm_read (port)))
> >       scm_primitive_eval_x (exp);
> >   }
> > 
> > I think this quite straightforward, no?
> 
> No.

Err, I mean, once you see how it is done.  The manual can be improved,
I'm sure.

> [...]  I also try to avoid gratuitous creations of
> string SCM's since the file name is generally invariant.
> You omitted returning the final value, which is important :-).

As you see, there are quite some variants on this theme, and I think
the above code is clean enough to allow for easy modification.  There
is no black magic in it.

> It's easier to ignore if the caller doesn't want it.  And, no,
> I didn't find that the research required to do this was obvious.
> In fact, there is no ``scm_set_port_file_name_x'' in 1.6.3, so
> you must be talking 1.7 here.

Oops, it is "scm_set_port_filename_x".

> I'm still working with 1.4.  Therefore, I must directly assign to
> ``file_name''.  I may as well assign the line number, too.

One more reason to join the happy Guile 1.6 club...

> I'd rather it was ready made so I can just #define my
> ``ag_scm_c_eval_string_from_file_line'' into
> ``scm_c_primitive_load_from_string''.  You might tweak it a bit so a
> NULL file pointer bypasses the scm_set_port_file_name_x call.  I'd
> also use "file_line" in the name, but that is your call.

Given these minor variations, I suggest you just use the function from
above.  That way, you get exactly what you want.

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405