| Re: How to get matching elements of array efficiently [message #69889] |
Mon, 01 March 2010 11:18 |
JJ
Messages: 36
Registered: January 2007
|
Member |
|
|
>
> Are your values integers as in the example? If they are, and if there
> are no big gaps in B (the number of elements of B is not much smaller
> than max(B)-min(B)), histogram comes to mind:
>
> h=histogram(A,binsize=1,min=min(b),max=max(b),reverse_indice s=ri)
> if (max(h) gt 0L) then res=ri[n_elements(h)+1:*] else (deal with the
> case of none found in b)
>
> Which would give the result you want in res, but ordered by bin (their
> order in B). In this example, res would be [0,3,2,4].
What I actually used was something akin to this:
h = histogram(a,min=0,max=(max(a) > max(b)),reverse_indices=ri)
w = where(h[b] gt 0)
res = ri[[ri[b[w]],ri[b[w]+1]-1]]
-JJ
|
|
|
|
| Re: How to get matching elements of array efficiently [message #69890 is a reply to message #69889] |
Mon, 01 March 2010 10:58  |
JJ
Messages: 36
Registered: January 2007
|
Member |
|
|
Thanks for the solutions everyone. I didn't know that Craig Markwardt
had written a solution. I'll probably fetch that. In the mean time,
I was able to use the histogram with reverse_indices method.
JD, I'm not sure that the code you pointed me at would work for my
case, as I need the indices of all matches, even if they are
duplicates.
-JJ
|
|
|
|
| Re: How to get matching elements of array efficiently [message #69921 is a reply to message #69890] |
Wed, 24 February 2010 14:26 |
JDS
Messages: 94
Registered: March 2009
|
Member |
|
|
On Feb 23, 3:11 pm, JJ <j...@cornell.edu> wrote:
> Hi All,
>
> I'm looking for an efficient (non-loop) solution to this problem. I
> have two arrays A, B. The elements of A are not necessarily unique,
> nor is A necessarily sorted. I'd like to find all the indices in A
> that match values that occur in B.
>
> For example, if A = [7,1,8,7,8], B = [7,8], the result should be
> [0,2,3,4].
>
> A and B can have many (order millions) of unique values, so I'd rather
> avoid a loop if I can.
>
> Is there any way to do this efficiently?
http://www.dfanning.com/tips/set_operations.html
|
|
|
|
| Re: How to get matching elements of array efficiently [message #69926 is a reply to message #69921] |
Wed, 24 February 2010 06:46 |
Jeremy Bailin
Messages: 618
Registered: April 2008
|
Senior Member |
|
|
On Feb 23, 3:36 pm, pp <pp.pente...@gmail.com> wrote:
> On Feb 23, 5:11 pm, JJ <j...@cornell.edu> wrote:
>
>> I'm looking for an efficient (non-loop) solution to this problem. I
>> have two arrays A, B. The elements of A are not necessarily unique,
>> nor is A necessarily sorted. I'd like to find all the indices in A
>> that match values that occur in B.
>
>> For example, if A = [7,1,8,7,8], B = [7,8], the result should be
>> [0,2,3,4].
>
>> A and B can have many (order millions) of unique values, so I'd rather
>> avoid a loop if I can.
>
>> Is there any way to do this efficiently?
>
> Are your values integers as in the example? If they are, and if there
> are no big gaps in B (the number of elements of B is not much smaller
> than max(B)-min(B)), histogram comes to mind:
>
> h=histogram(A,binsize=1,min=min(b),max=max(b),reverse_indice s=ri)
> if (max(h) gt 0L) then res=ri[n_elements(h)+1:*] else (deal with the
> case of none found in b)
>
> Which would give the result you want in res, but ordered by bin (their
> order in B). In this example, res would be [0,3,2,4].
Of course, I'll have to pipe in here and say that if your values
aren't integers, or if there are large gaps in them, you can use UNIQ
and VALUE_LOCATE to map them into a set of consecutive integers, and
then use HIstOGRAM. :-)=
-Jeremy.
|
|
|
|
| Re: How to get matching elements of array efficiently [message #69928 is a reply to message #69926] |
Wed, 24 February 2010 00:54 |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
On 23 Feb., 21:11, JJ <j...@cornell.edu> wrote:
> Hi All,
>
> I'm looking for an efficient (non-loop) solution to this problem. I
> have two arrays A, B. The elements of A are not necessarily unique,
> nor is A necessarily sorted. I'd like to find all the indices in A
> that match values that occur in B.
>
> For example, if A = [7,1,8,7,8], B = [7,8], the result should be
> [0,2,3,4].
>
> A and B can have many (order millions) of unique values, so I'd rather
> avoid a loop if I can.
>
> Is there any way to do this efficiently?
>
> Thanks.
>
> -JJ
Hi,
I don't know the other solutions, but maybe this one will also work
for you :)
wh = where(total(((c=abs(rebin(transpose(a[*]),$
((nb=n_elements(b))),((na=n_elements(a))),/sample) - $
rebin(b[*],nb,na,/sample)))) eq 0,1))
Cheers
CR
|
|
|
|
| Re: How to get matching elements of array efficiently [message #69932 is a reply to message #69928] |
Tue, 23 February 2010 13:06 |
wlandsman
Messages: 743
Registered: June 2000
|
Senior Member |
|
|
You might look at Craig Markwardt's program match2 in
http://idlastro.gsfc.nasa.gov/ftp/pro/misc/match2.pro
For your example, one would type
IDL> match2, a, b, suba, subb
IDL> print,where(suba GE 0)
0 2 3 4
You might be able to make match2.pro more efficient, because it also
gives the index of a (not necessarily unique) value in B that matches
a value in A, which is information you don't need. --Wayne
On Feb 23, 3:11 pm, JJ <j...@cornell.edu> wrote:
> Hi All,
>
> I'm looking for an efficient (non-loop) solution to this problem. I
> have two arrays A, B. The elements of A are not necessarily unique,
> nor is A necessarily sorted. I'd like to find all the indices in A
> that match values that occur in B.
>
> For example, if A = [7,1,8,7,8], B = [7,8], the result should be
> [0,2,3,4].
>
> A and B can have many (order millions) of unique values, so I'd rather
> avoid a loop if I can.
>
> Is there any way to do this efficiently?
>
> Thanks.
>
> -JJ
|
|
|
|
| Re: How to get matching elements of array efficiently [message #69933 is a reply to message #69932] |
Tue, 23 February 2010 12:36 |
penteado
Messages: 866
Registered: February 2018
|
Senior Member
Administrator |
|
|
On Feb 23, 5:11 pm, JJ <j...@cornell.edu> wrote:
> I'm looking for an efficient (non-loop) solution to this problem. I
> have two arrays A, B. The elements of A are not necessarily unique,
> nor is A necessarily sorted. I'd like to find all the indices in A
> that match values that occur in B.
>
> For example, if A = [7,1,8,7,8], B = [7,8], the result should be
> [0,2,3,4].
>
> A and B can have many (order millions) of unique values, so I'd rather
> avoid a loop if I can.
>
> Is there any way to do this efficiently?
Are your values integers as in the example? If they are, and if there
are no big gaps in B (the number of elements of B is not much smaller
than max(B)-min(B)), histogram comes to mind:
h=histogram(A,binsize=1,min=min(b),max=max(b),reverse_indice s=ri)
if (max(h) gt 0L) then res=ri[n_elements(h)+1:*] else (deal with the
case of none found in b)
Which would give the result you want in res, but ordered by bin (their
order in B). In this example, res would be [0,3,2,4].
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
