| precedence question [message #43003] |
Mon, 14 March 2005 10:35  |
Foldy Lajos
Messages: 268
Registered: October 2001
|
Senior Member |
|
|
Hi guys,
I am a little bit lost in operator precedence. I think that . and [] has
the same precedence, and associate from left to right. So I expect the
following three expressions to be equivalent:
a=replicate({l:lindgen(2,3)}, 4,5)
help, a[3:6] .l [*,0:1] ; a
help, (a[3:6]).l [*,0:1] ; b
help, ((a[3:6]).l)[*,0:1] ; c
a is the same as b, but c differs. Where am I wrong?
regards,
lajos
|
|
|
|
| Re: precedence question [message #43126 is a reply to message #43003] |
Tue, 15 March 2005 06:42  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Lajos writes:
> Sorry, David, I didn't want to attack you. Your tutorials (and the full
> web site) is excellent.
Oh, it would take a lot more than that to make me feel
attacked. :-)
> I just came over an expression, where the IDL sw and docs contradict. I
> wanted to know, whether this is a real contradiction, or there is a
> "hidden rule" somewhere I don't know of.
If it is a contradiction between the software and the
documentation, it is the first I have ever discovered
in IDL. :^)
> OK, let's close this topic.
OK. Good idea. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: precedence question [message #43128 is a reply to message #43003] |
Tue, 15 March 2005 06:35 |
Foldy Lajos
Messages: 268
Registered: October 2001
|
Senior Member |
|
|
On Tue, 15 Mar 2005, David Fanning wrote:
> =?ISO-8859-2?Q?F=F6ldy_Lajos?= writes:
>
>> David's Operator Precedence Tutorial has the same table, and refers to
>> '[]' and '.' as equal precedence operators, so it is wrong, too :-)
>
> All I know is that as soon as you start parsing "left", "right",
> and "equal" people's eyes glaze over. It's right enough for me.
> If it's not right *always*, well, I don't hold people to higher
> standards than I aspire to. :-)
>
> Cheers,
> David
>
Sorry, David, I didn't want to attack you. Your tutorials (and the full
web site) is excellent.
I just came over an expression, where the IDL sw and docs contradict. I
wanted to know, whether this is a real contradiction, or there is a
"hidden rule" somewhere I don't know of.
OK, let's close this topic.
regards,
lajos
|
|
|
|
| Re: precedence question [message #43129 is a reply to message #43003] |
Tue, 15 March 2005 06:13 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
=?ISO-8859-2?Q?F=F6ldy_Lajos?= writes:
> David's Operator Precedence Tutorial has the same table, and refers to
> '[]' and '.' as equal precedence operators, so it is wrong, too :-)
All I know is that as soon as you start parsing "left", "right",
and "equal" people's eyes glaze over. It's right enough for me.
If it's not right *always*, well, I don't hold people to higher
standards than I aspire to. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: precedence question [message #43130 is a reply to message #43003] |
Tue, 15 March 2005 05:55 |
Foldy Lajos
Messages: 268
Registered: October 2001
|
Senior Member |
|
|
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 :-)
regards,
lajos
|
|
|
|
|
|
| Re: precedence question [message #43133 is a reply to message #43003] |
Tue, 15 March 2005 01:42 |
Foldy Lajos
Messages: 268
Registered: October 2001
|
Senior Member |
|
|
Hi Marc,
On Tue, 14 Mar 2005, m_schellens@hotmail.com wrote:
> [] has higher (the highest after (expr) ) precedence.
>
> Cheers,
> marc
>
>
as concatenation, yes, but not as subscripting. The IDL reference guide
says:
Operator Precedence
The following table lists IDL's operator precedence. Operators with the
highest precedence are evaluated first. Operators with equal precedence
are evaluated from left to right.
Priority Operator
First (highest) ( ) (parentheses, to group expressions)
[ ] (brackets, to concatenate arrays)
Second . (structure field dereference)
[ ] (brackets, to subscript an array)
( ) (parentheses, used in a function call)
Third * (pointer dereference)
.
.
.
so my original question still holds.
regards,
lajos
|
|
|
|
|
|
| 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, F�ldy 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
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
