Eike Rathke:
> As already pointed out on the comment list, for some occasions an
> IFERROR(expression;errorcase) does come handy. The function evaluates
> 'expression' as usual, and if that results in an error returns the
> result of an 'errorcase' expression, otherwise the result of
> 'expression'. Also ECMA/MOOXML defines this function.
>
> I propose to include IFERROR in our ODFF specification.
Great! That would be another way to handle the problem I noted earlier (namely, how to prevent writing a complicated expression twice just to override one value). I had originally proposed an "IFEQ" function, which is more general than IFERROR, but it special-cases Error values and isn't implemented anywhere currently.
One problem with IFERROR is that it masks WHICH error occurs - you may only want the substitution with one Error value (e.g., NA()). One advantage of IFERROR is that it's simple and straightforward, and that is a VERY GOOD thing. It being implemented somewhere is a good thing too.
Andreas J Guelzow:
>On first glance this looks equivalent to
>IF(ISERROR(expression);errorcase;expression)
>but of course this "equivalent" version evaluates 'expression' twice
>which can give a different result.
Right. Also, 'expression' may be complicated, so only having to say it once is a very good thing.
We could add IFERROR, IFEQ, both, or neither. I think we should add at _least_ IFERROR. Comments?
--- David A. Wheeler
