sprintf behaves strangely when given empty input cell

[Muthiah Annamalai]

> Date: Sat, 05 Jan 2008 21:20:58 +0100
> From: Thomas Weber <thomas.weber.mail at gmail.com>
> Subject: sprintf behaves strangely when given empty input cell
> To: bug at octave.org
> Message-ID: <1199564459.32293.27.camel at localhost>
> Content-Type: text/plain
>
> Bug report for Octave 3.0.0 configured for i486-pc-linux-gnu
>
> Description:
> -----------
>
> copyfile(f1,...) fails when f1 is an empty cell. This is because sprintf
> seems to have a problem with empty cells as well:
>
> ====================================
> octave:1> f1 = {}
> f1 = {}(0x0)
> octave:2> sprintf ("\"%s\" ", f1)
> ans = ""
> octave:3> sprintf ("\"%s\" ", f1{:})
> ans = "
> ====================================
>
> Note that in the second sprintf() call, the second quote is missing.
>   
 From whatever I can tell, it seems the problem seems to start with the 
parser.
Actually the call in sprintf() file-io.cc, os.printf() doesnt pass any 
'args' the arguments
but the format to the printf is OK. Consequently the do_printf routine 
breaks
as soon as a conversion ( '%s' ) doesnt have a matching argument, and 
forgets
the rest of the formatted-list. I dont know whats the expected behavior, 
and maybe
we can be forgiving and print the rest of the format-list. That way, 
what Thomas
encoutered would atleast be > ans = "".

The larger question is what do we want to treat {:} for empty cells, and 
I think
that is something JWE must address. Infact Octave behaves similarly even 
if you
sprintf("\'%s\'",{}{:})

so one easy way to solve this is make the 
octave_base_stream::do_printf() forgiving
and continue with other elements of the format-list hoping they are not 
conversions.

Also similar problems exist due because of the treatment / choices at 
parsing stage.
-Muthu