| Arg_Present [message #9641] |
Fri, 01 August 1997 00:00  |
J.D. Smith
Messages: 214
Registered: August 1996
|
Senior Member |
|
|
I'm just full of complaints this week. This particular one is about
Arg_Present. It is a lovely addition, and one for which I clamored
before the release of IDL v5, but it has something left out... it does
not correctly identify *inherited* keyword passed variables as return
variables. This may seem like a small detail, but consider this
scenario: A subclass' GetProperty method calls its superclass' method
with any extra keywords it receives. The superclass' GetProperty method
checks if keywords are present as return variables before computing a
property for return. But this doesn't work... _EXTRA keywords will not
trigger arg_present, and the whole system is broken.
Just another link in a long chain of gripes.
JD
|
|
|
|
| Re: Arg_Present [message #9723 is a reply to message #9641] |
Wed, 06 August 1997 00:00  |
J.D. Smith
Messages: 214
Registered: August 1996
|
Senior Member |
|
|
David Fanning wrote:
>
> J.D. Smith writes:
>
>> I'm just full of complaints this week. This particular one is about
>> Arg_Present. It is a lovely addition, and one for which I clamored
>> before the release of IDL v5, but it has something left out... it does
>> not correctly identify *inherited* keyword passed variables as return
>> variables. This may seem like a small detail, but consider this
>> scenario: A subclass' GetProperty method calls its superclass' method
>> with any extra keywords it receives. The superclass' GetProperty method
>> checks if keywords are present as return variables before computing a
>> property for return. But this doesn't work... _EXTRA keywords will not
>> trigger arg_present, and the whole system is broken.
>
> Well, as long as we are gripping about Arg_Present, let me
> get my two cents in. :-)
>
> I was under the impression that Arg_Present was introduced
> to solve the sometimes subtle problem of telling whether
> a keyword was *used* or not. And in limited circumstances
> it does this well enough. The danger is in using this
> seemingly useful function in place of the much more
> useful, but completely misnamed (at least for the purpose
> it is being used for), N_Elements.
>
> In particular, Arg_Present returns a 0 (meaning no argument
> present) when a keyword argument is present, but the argument
> is passed by value. It only returns a 1 if the argument is
> present and is passed by reference.
>
> Now, to give RSI credit, this is all documented correctly.
> It is just that most people using Arg_Present will *assume*
> it does what its name suggests it does and end up using it
> incorrectly. It's the Keyword_Set problem all over again.
> It seems to me this whole idea of knowing if and when a
> keyword parameter was used needs more thought and a much
> more consistent (and meaningfully named) interface.
>
You are indeed correct that a better interface would be useful.
Remember, however, that arg_present works on both keywords *and*
arguments, and serves to tell you if any variable passed is passed by
reference. As you point out, using n_elements or keyword_set on an
*undefined* passed argument (e.g. mypro, outvariable=undefinedvar) will
be to no avail... you'll not be able to discriminate between this case,
and the case of it never having been passed at all, unless of course you
give undefinedvar a value before passing it, which is not ideal.
But nonetheless, arg_present works as advertised. However, the present
issue is inherited keywords and the fact that they are always passed by
value, which introduces some problems with object oriented methodology.
Cf. the my latest article in this thread.
JD
|
|
|
|
| Re: Arg_Present [message #9731 is a reply to message #9641] |
Mon, 04 August 1997 00:00 |
J.D. Smith
Messages: 214
Registered: August 1996
|
Senior Member |
|
|
J.D. Smith wrote:
>
> I'm just full of complaints this week. This particular one is about
> Arg_Present. It is a lovely addition, and one for which I clamored
> before the release of IDL v5, but it has something left out... it does
> not correctly identify *inherited* keyword passed variables as return
> variables. This may seem like a small detail, but consider this
> scenario: A subclass' GetProperty method calls its superclass' method
> with any extra keywords it receives. The superclass' GetProperty method
> checks if keywords are present as return variables before computing a
> property for return. But this doesn't work... _EXTRA keywords will not
> trigger arg_present, and the whole system is broken.
>
> Just another link in a long chain of gripes.
>
> JD
Well, this is one of those difficulties which goes even deeper... What
is really happening is that, since _EXTRA keywords get passed in a
structure, they are no longer call by reference, and cannot get passed
out as return items at all. As a fix, I dynamically create a structure
for return of variables, and return it along the method chain, adding to
it as necessary as I go.
JD
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
