| Re: keyword_set([0]) [message #48073 is a reply to message #48072] |
Mon, 20 March 2006 12:24  |
JD Smith
Messages: 850
Registered: December 1999
|
Senior Member |
|
|
On Mon, 20 Mar 2006 10:04:49 -0800, vlk.astro wrote:
> Hello,
>
> We are finally being forced to migrate from 5.3 to 6.2, and
> somewhere along the way IDL has bowled a googly with
> keyword_set.
>
> With 5.3, an array [0] was recognized as having been set:
> IDL> print, !version
> { sparc sunos unix 5.3 Nov 11 1999}
> IDL> a=0 & b=[a] & print, keyword_set(a), keyword_set(b)
> 0 1
> IDL> a=1 & b=[a] & print, keyword_set(a), keyword_set(b)
> 1 1
>
> With 6.2, 0 and [0] are indistinguishable:
> IDL> print, !version
> { sparc sunos unix Solaris 6.2 Jun 20 2005 64 64}
> IDL> a=0 & b=[a] & print, keyword_set(a), keyword_set(b)
> 0 0
> IDL> a=1 & b=[a] & print, keyword_set(a), keyword_set(b)
> 1 1
>
> Now, I am not one to rail against progress, especially when
> the new behavior matches the documentation. But nevertheles,
> it is damned inconvenient, because I think I have around a
> hundred off procedures that depended on the 0/[0] dichotomy.
>
> My question : is there a simple way to replace the old calls to
> keyword_set() with one- or two-liners that will work in both 5.3
> and 6.2 and one that will know the difference between a scalar 0
> and a vector 0? I don't want to roll my own function because of
> issues of speed (some of the keyword_set's are deeply nested,
> and I'd rather not have the extra overhead of a new function call)
> and aesthetics (i.e., as much native functionality as possible).
> Right now all I have is an ugly concoction that involves size(),
> n_elements() _and_ keyword_set().
>
> Thanks,
> Vinay
Try,
IDL> print, size(b,/N_DIMENSIONS) gt 0 OR keyword_set(b)
The scalar vs. single element vector dichotomy is one which the RSI
engineers wish didn't exist, since it no doubt forces many many lines
of special purpose code into IDL which do nothing other than maintain
this artificial distinction. I think this change in KEYWORD_SET in
IDL 5.6 was a limited nod towards reducing the most egregious
annoyances which result from this distinction. Here's the blurb:
The KEYWORD_SET function returns true if its argument is defined
and is nonzero, and false (0) otherwise. The specific rules by
which the value is determined are given in the IDL Reference
Guide. With IDL 5.6, there has been a small change to these rules,
designed to make KEYWORD_SET useful in a larger number of cases.
Previously, KEYWORD_SET would return true if it's argument was an
array, regardless of the value. This behavior has been changed:
Arrays with more than 1 element are treated as before, but
1-element arrays are treated in the same way as scalar arguments,
and the value returned by KEYWORD_SET depends on the value of the
element.
Personally, I would change the logic so that a single element vector
isn't required to be true by default, or isn't ever passed in for
boolean arguments/keywords. Changing the logic in this way should be
compatible across IDL 5.3/6.2, without new and hard to maintain tests
like this.
JD
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
