Find a match in a structure array using arrayfun

Hi there,

I have some data stored in a structure array s, with the following fields: s.Month, s.Day, s.Hour and s.Data. The structure array has the size of 1x123, i.e.a time series of 123 time steps.

I try to find a certain set within the structure array, matching the date, say 08/21 00 UTC.

My idea was:

 m = 8; d = 21; h = 0;
 index = find( arrayfun(@(x) x.Month==m & x.Day==d & x.Hour==h, s) );
 data = s(index).Data;

But the result of arrayfun() is an empty matrix. (I know that the index must be 81.)

What actually works is:

 find( arrayfun(@(x) x.Month==m,s)   & ...
       arrayfun(@(x) x.Day==d,s)      & ...
       arrayfun(@(x) x.Hour==h,s) )
ans =

    81

My question is: why does the first option not work ?

Thank you for an answer.

[Image Analyst]

dpb, I'm trying to give you votes to get your privileges up but I can't vote for comments. That looks a lot like an answer to me.

[dpb]

If it walks like a duck, quacks like a duck, plays golf like the AFLAC duck, ... :)

OKAY, I always start out thinking I've just got a comment then something happens--like the realization of the [] to build the arrays above was after I was already to sign off having simply noted his solution worked for me. I don't follow what didn't in his case, yet, however...

I'd noticed you'd been pinging a bunch but I'm not sure I'll live long enough to get to the privilege level of any significance... :)

The only thing I miss as structured is that sometimes it would be good to re-edit a submission but it takes excessive amount of time already. What I really miss that doesn't seem to exist at all is for feedback from the privileged be considered in updating the forum. It wouldn't be necessary to have to re-edit so much code if the default weren't wordwrap and a few other details like that...there is so much wasted effort in cleaning up happening even as it is and not a tenth of what should gets done.

[Image Analyst]

You're not far from 2000 when you'll be able to edit posts of others. I'll help by giving you 2 points per vote so you'll be there shortly.

Answer 1

[dpb]

Not certain; it did for a small sample case here...but you can write it w/o arrayfun as

ix=find([s.Month]==m & [s.Day]==d & [s.Hour]==h);

making the arrays from the struct.

ADDENDUM

Or, add a datenum value as another field and make your search on the single variable might be faster searching and certainly less complicated to build the search pattern I think. It's certainly much simpler to do ranges that way.

[Alexander]

dpb, thanks for the answer.

I had a typing error, so the first mentioned code does it too, as well as yours:

 % variant 1
 find(arrayfun(@(x) x.Month==M & x.Day==D & x.Hour==H,s))

 % variant 2 without arrayfun
 find([s.Month]==M & [s.Day]==D & [s.Hour]==H)

But your code is 10 times faster, thus using arrayfun makes really no sense in this case.

Alex

[dpb]

Figured had to be something of the sort; glad you found it.

Interesting that there's that large a speed difference...I hadn't timed it as just used a very short handmade test vector to verify syntax.