Re: PATCH: += parameter assignments

[Oliver Kiddle <okiddle@xxxxxxxxxxx>]

Sorry if you get this twice - I'm resending it after it didn't turn up
first time and sorry if Yahoo re-wraps the text horribly.

Bart Schaefer wrote:

> I'm not entirely happy with the addition of the a+=val syntax because
> it
> conflicts (conceptually, not mechanically) with the += operator that
is
> already present in the math syntax.  Consider that the following:
>
>     integer i=4
>     typeset s=4
>     i+=5
>     s+=5
>     ((i+=5))
>     ((s+=5))
>     print $i $s
>
> yields
>
>     14 50

What would you have sooner expected - `14 14'?

>
> However, what I consider to be worse is that:
>
>     s=four
>     ((s+=5))
>     s+=5
>     print $s
>
> yields
>
>     55
>
> Interpreting strings as integers in math context makes some sort of
> sense, but doing both that, and also overloading += depending on the
> parameter type, is going too far.  I think the suggested change for
> -= would be even more confusing.

I'm not sure that I've entirely understood your argument here as your
example doesn't seem particularly confusing to me. The ((s+=5)) results
in 5 with s remaining a scalar - that was the case before. It remained
a scalar so the s+=5 results in 55.

I don't think it is ideal that the new += can be used for numeric
addition. I find += to be useful with arrays, scalars and associations
and it is these (with the addition of compound variables) for which I
imagine this feature was intended in ksh93. My intention was to
continue to use (( ... )) when dealing with numerics  - we could add a
note to recommend this in the documentation.

Have you got any ideas on how to solve this? Short of ditching +=, the
only thing I can think of would be to make scalar+=val identical to
(( scalar+=val )) but I think it would be a pity to lose the appending
to scalars functionality. Does anyone else have any view or ideas?

> type, is going too far.  I think the suggested change for -= would be
> even more confusing.

The results might not be imediately obvious after a mix of both the
math and non-math += for scalars but considered on their own I don't
find them "confusing".

> And the following can't be anything but a bug:

A bug maybe, but it has bugger all to do with the += code.

> (Previously

meaning 4.0.4 but 4.1.0-dev-1 behaves as now so the problem will have
been introduced sometime before that. 15292 at a guess as that is the
one in that time frame dealing with math.c.

Oliver



__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com