Re: Docstrings and Reference manual (Re: Discussing Guile and revising interfaces)

[Keisuke Nishida <kxn30 at po dot cwru dot edu>]

Mikael Djurfeldt <mdj@mdj.nada.kth.se> writes:

> I think your suggestions are very interesting.  The interface for such
> properties could be `procedure-property' et al.  The properties need
> not be loaded to begin with, but can be loaded on demand from a
> database file.

Right.  Any object which has a docstring may have properties in the same
way.  Probably a general interface to properties like `get-property' may
be convenient:

  (define (get-property obj key)
    (or (object-property obj key)
        (and (procedure? obj) (procedure-property obj key))
        (find-property-from-database-file (object-name obj) key)))

Object names are really essential for this.


"Greg J. Badros" <gjb@cs.washington.edu> writes:

> As I think we agree, the properties should be added to the function, not
> the docstring, but otherwise sounds reasonable.  What kinds of
> properties are you thinking about?

Any miscellaneous properties, such as file location (e.g., "alist.c:12"),
cross references, names of arguments (for procedures), and possible
value types (for customizable variables), can be included.  I haven't
thought very details, but I need those information to implement the new
describe procedure.