From help-octave-request at bevo dot che dot wisc dot edu  Tue Jan 20 12:50:13 1998
Subject: exist gotcha? (was Re: Re: How to tell a function from a string)
From: king at cogsci dot ucsd dot edu (Jonathan King)
To: help-octave at bevo dot che dot wisc dot edu
Cc: king at cogsci dot ucsd dot edu
Date: Tue, 20 Jan 1998 10:49:32 -0800

I *think* I've got this text attributed correctly...

John W. Eaton <jwe at bevo dot che dot wisc dot edu> writes:
>
>On 20-Jan-1998, Jim Van Zandt <jrv at mbunix dot mitre dot org> wrote:
>|
>| I also thought the documentation could have been more helpful.
>| Eventually I found this:
>| 
>|  - Built-in Function:  exist (NAME)
>|      Return 1 if the name exists as a variable, 2 if the name (after
>|      appending `.m') is a function file in the path, 3 if the name is a
>|      `.oct' file in the path, or 5 if the name is a built-in function.
>|      Otherwise, return 0.
>| 
>| and decided that "exist(NAME) > 1" was a reasonable test.  (Can it
>| ever return 4?)
>
>No, Octave's exist function doesn't currently return 4, but I'm not
>sure what Matlab uses that return code for, so if you want to write
>portable code, you might want to check that first.

Matlab4.2 allegedly returns 4 if A is a compiled SIMULINK(tm)
function.  I say "allegedly" only because I haven't tried it.

>| I'm now surprised to read JWE's reply:
>| 
>| > Unfortunately, there isn't a good way to do this from Octave, though
>| > you could probably wrap is_valid function from src/variables.cc in a
>| > DEFUN easily enough.
>| 
>| What's wrong with the above test?
>
>Uh, I just completely forgot about it.

I see one other "gotcha", as suggested by the following octave session:

mk26:~/matlab> octave
Octave, version 2.0.9 (i586-pc-linux-gnu).
Copyright (C) 1996, 1997 John W. Eaton.
This is free software with ABSOLUTELY NO WARRANTY.
For details, type `warranty'.

octave.bin:1> exist("who")  % this will return 5, like it should
ans = 5
octave.bin:2> % here's a stupid typo and its parse error
octave.bin:2> exist(who)
parse error:

>>> exist(who)
             ^

octave.bin:2> exist("who")  % now this will fail...
ans = 0
octave.bin:3> who           % until you use it correctly...
octave.bin:4> exist("who")  % and now it works again
ans = 5

I'm not sure what's going on here, but it looks funny.  As far as I
can tell, this only happens with built-ins; specifically, if you
precede the octave session with "touch who.m" in a directory on your
path, the parse error doesn't seem to block the correct operation of
exist (in the sense that it will correctly return ans=2, since it
notices the shadowing function).

I think I stumbled across this first when I accidentally defined a
variable called "save"...but I think that's all I want to say about
that particular incident.

jking