OpenEuphoria

Re: routine_id()

• Posted by Robert Craig <rds at ATTCANADA.NET> Jul 04, 2000
• 605 views

Jiri Babor writes:
> To restrict the search for the routine's name to the point where the
> routine_id() is _located_ in the source code and not where it is
> actually _invoked_ is probably very convenient for the interpreter,
> but a bloody nuisance as far as I am concerned ;). Can anything be
> done about it, Robert? You would not need two passes for that,
> would you?

David Cuny has requested the same thing on several
occasions. I may reconsider it at some point, but I still tend
to believe what I said in a posting in March 1998:

------------------

... I could have defined routine_id() so it would search
forwards and backwards to find routines *anywhere*
in the program (assuming they have been defined at
the time routine_id is called.)

I chose the current behavior for routine_id for the following
reasons:

     * It matches the way routines are looked-up in the
       symbol table when a normal call is performed. This
       makes the language definition simpler.

     * I could imagine a lot of very weird bugs happening
       if a library .e file used routine_id *internally* and you
       included this .e file in your program, and you accidentally
       had a routine with the same name as routine_id() in the
       .e file was looking for. You might have no idea why
       your routine was getting called since you wouldn't
       necessarily understand the workings of the .e file. I felt it
       would be safer if it was your job to get the routine id of
       your own routine and pass it to any library that
       wants to call your routine. There would be no surprises.

     * I am still clinging to the somewhat controversial belief,
       that there is maintenance and readability value in having
       the language force everyone to order variables
       and routines such that they are defined before they are
       used. (It also opens some possibilities for optimization
       that I have already taken advantage of.)

       routine_id gives you a mechanism for breaking that
       order, but it's much less convenient than a normal call
       (especially to call a "later" routine), and the reader of
       your code can plainly see that you are doing something
       unusual when he sees the routine_id() calls and the
       explicit call_proc() and call_func() calls.

Regards,
   Rob Craig
   Rapid Deployment Software
   http://www.RapidEuphoria.com