comp.lang.idl-pvwave archive: archive

Home » Public Forums » archive » precedence question

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: precedence question [message #43175 is a reply to message #43130]

Fri, 18 March 2005 15:45

JD Smith
Messages: 850
Registered: December 1999

Senior Member

On Tue, 15 Mar 2005 14:55:16 +0100, Fï¿½ldy Lajos wrote:

>
> On Tue, 15 Mar 2005, m_schellens@hotmail.com wrote:
>
>> What do you want to hear?
>> Obviously the docmentation is wrong here and brackets have a higher
>> priority.
>> marc
>>
>
> Well, I'd like to know whether it is a software bug, a documentation bug,
> or simply I am reading something wrong. It is not obvious for me :-)
>
> You voted for documentation bug, thanks.
>
> David's Operator Precedence Tutorial has the same table, and refers to
> '[]' and '.' as equal precedence operators, so it is wrong, too :-)

Well, it's actually my tutorial David hosts, and I admit I stole the
precedence table straight from the manual without extensive
verification. That said, I think you guys have this whole issue
wrong.

You're indexing a 3D array using only two of three dimensions. Now
you might complain that IDL has irregular behavior in this case, but I
don't see it as a precedence issue. Consider instead:

IDL> a=replicate({l:randomu(sd,10)},5)
IDL> print,a[2].l[6]
0.0162049
IDL> print,(a[2]).l[6]
0.0162049
IDL> print,((a[2]).l)[6]
0.0162049

Looks pretty good. In fact, if you think about it, there is no way
for `.' and `[]' to have anything but equal precedence. Let's say the
precedence of '[]' really was higher. How would IDL parse a[2].l[6]?
Let's see, a[2] is a scalar structure, but now we must first subscript
that with [6], since that has higher precedence.... hmmm.

What you've really run into is the apparently variable way IDL treats
indexing expressions which specify some, but not all dimensions:

IDL> a=findgen(10,10,10)
IDL> print,a[4,5:6,0]
54.0000
64.0000
IDL> print,a[4,5:6]
54.0000
64.0000

So here IDL just assumes you meant the first plane, as a convenience
to you. It could have thrown an error and said "Sorry try again
buddy", but it was being friendly instead. The only difference
between your case A and case C is that in case A IDL does it's
indexing/structure derefencing one by one and builds up the array from
the structure, whereas in case C you *first* construct the array of
size [2,3,4], and then index it with the incomplete index set [*,0:1],
at which point the "assume he meant the first plane" rule for arrays
kicks in, and you get the different result. By the way, if you had
used

IDL> help, a[3].l[*,0:1]

instead you would have found consistent behavior as well, since there
you're indexing only a 2x3 array.

JD

Report message to a moderator

[Message index]

		precedence question By: Foldy Lajos on Mon, 14 March 2005 10:35
		Re: precedence question By: David Fanning on Tue, 15 March 2005 06:42
		Re: precedence question By: Foldy Lajos on Tue, 15 March 2005 06:35
		Re: precedence question By: David Fanning on Tue, 15 March 2005 06:13
		Re: precedence question By: Foldy Lajos on Tue, 15 March 2005 05:55
		Re: precedence question By: JD Smith on Fri, 18 March 2005 15:45
		Re: precedence question By: marc schellens[1] on Tue, 15 March 2005 04:43
		Re: precedence question By: Foldy Lajos on Tue, 15 March 2005 01:42
		Re: precedence question By: marc schellens[1] on Mon, 14 March 2005 23:53

Previous Topic:	Forcing READ_ASCII output to be a set of strings...
Next Topic:	HIST_ND() used for resampling point data onto grid

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Fri Oct 10 01:38:33 PDT 2025

Total time taken to generate the page: 1.20351 seconds