| Re: fastest way to find the first non-zero value in an array [message #59671] |
Tue, 08 April 2008 07:02  |
Vince Hradil
Messages: 574
Registered: December 1999
|
Senior Member |
|
|
On Apr 8, 8:34 am, David Fanning <n...@dfanning.com> wrote:
> Vince Hradil writes:
>> Wow - I'd be interested in knowing how slow 'where' is. Are we
>> talking the difference between 0.01 seconds and 0.05 seconds? Or even
>> the difference between 1 and 5 seconds? Time is money, but at what
>> point does our 'need for speed' end?
>
> I've gotten to the point where anything that takes less time
> than it takes to go get a cup of coffee is fast enough. I
> used to think fast, elegant programs were required. But
> when you are writing one-offs day after day, why bother?
> With Starbucks just across the street, I can afford to be
> a little loose with a FOR loop.
>
> Cheers,
>
> David
>
> P.S. That said, I just spent the entire weekend re-working
> a program I inherited from someone else. It is generally a
> good idea to write a program in such a way that someone else
> can get it to work in less time than it takes to write the
> darn thing from scratch. :-)
>
> --
> David Fanning, Ph.D.
> Fanning Software Consulting, Inc.
> Coyote's Guide to IDL Programming:http://www.dfanning.com/
> Sepore ma de ni thui. ("Perhaps thou speakest truth.")
Exactly - the google-mentality is making everyone think that 10
seconds is too long to wait for anything. But this makes sense to
me: it take about 2-4 hours for the scientist to acquire the data for
a certain experiment, it takes me about the same time to create the
one-off and run the analysis for said experiment. Sure I could spend
about 8 hours to tweak the analysis to make it run in 5 minutes
instead of half-an-hour, but why bother. Unless, of course we need
that extra half-hour? Then I end up writing some obfuscated code
(http://en.wikipedia.org/wiki/Obfuscated_code) that uses histograms,
and the next developer that comes along just says, 'unh?' and re-
writes the whole thing.
|
|
|
|
|
|
|
|
|
|
| Re: fastest way to find the first non-zero value in an array [message #59675 is a reply to message #59674] |
Tue, 08 April 2008 06:03  |
Spon
Messages: 178
Registered: September 2007
|
Senior Member |
|
|
On Apr 8, 9:04 am, smas...@locean-ipsl.upmc.fr wrote:
> Hi,
>
> I want to find the first non-zero value of an array. Is there a faster
> way to do this than with the "where" command: (where(array ne 0))[0]
> "Where" will look for all non-zero values and I only need the first
> one. It would be great if I could stop "where" in its search process
> as soon as it found one element...
>
> sebastien
Hi Sebastien,
I can think of two separate ways of going about it:
Firstly, for certain arrays it may be perfectly sensible to use the
'inefficient', FORTRAN 101 route:
i = 0
WHILE Array[i] EQ 0 DO i++
RETURN, i
Secondly, I've attempted to do it using HISTOGRAM below.
-----
; Is the first element non-zero?
; If it is, we can save ourselves a lot of
; hassle...
IF Array[0] NE 0 THEN RETURN, 0L
; We now know element 0 contains data = 0.
; Generate a histogram of the array such
; that this element will always be put in the
; first bin:
H = HISTOGRAM(CEIL(ABS(Array)), REVERSE_INDICES = RI)
; Are there non-zero elements in the array?
IF N_ELEMENTS(H) EQ 1 THEN MESSAGE, $
'Array contains only zeroes!'
; Array[0] is always going to be in the first bin.
; Get the contents of that bin:
BinContents = RI[RI[0]:RI[1]-1]
; How many drops in that bin?
NBC = N_ELEMENTS(BinContents)
; Where do the drop indices stop increasing linearly?
; That's where the first non-zero element must be.
Index = WHERE((LINDGEN(NBC) - BinContents) NE 0, Count)
; If all the zeroes in the array come before the first
; non-zero value, then we won't get any indices returned,
; but the next index will be the next element after the
; end of our BinContents vector. This must be non-zero.
IF Count EQ 0 THEN RETURN, NBC
; Otherwise, use the WHERE results to return
; the index of first non-zero element of the array.
RETURN, Index[0]
-----
As far as I can tell, both methods work; and which is going to be
faster (between these two and just using WHERE) is going to depend on
your array.
Let us know how you get on,
Regards,
Chris
|
|
|
|
|
|
| Re: fastest way to find the first non-zero value in an array [message #59814 is a reply to message #59678] |
Wed, 09 April 2008 13:02 |
karo03de
Messages: 21
Registered: March 2007
|
Junior Member |
|
|
On 8 Apr., 10:42, "Clemens" <yy...@hotmail.com> wrote:
> it depends on the size of array. Small array is a simple loop.
> Huge one can be :
>
> index_arr = where( arr NE 0)
> print, index_arr[0]
>
> on very huge arrays you can devide arr into big peaces.
>
> Clemens
>
> <smas...@locean-ipsl.upmc.fr> schrieb im Newsbeitragnews:ff3bf651-8707-47e8-8cff-5e694dd5622f@m71g200 0hse.googlegroups.com...
>
>> Hi,
>
>> I want to find the first non-zero value of an array. Is there a faster
>> way to do this than with the "where" command: (where(array ne 0))[0]
>> "Where" will look for all non-zero values and I only need the first
>> one. It would be great if I could stop "where" in its search process
>> as soon as it found one element...
>
>> sebastien
What about the system function ARRAY_EQUAL ?
Karsten
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
