按列匹配python pyfits表

Question

I am trying to match two pyfits data objects by a OBJNO (object number) column. In IDL this is done with the match, A.objno, B.objno, ii, jj procedure. This returns two indicies ii and jj which index A and B such that A[ii].objno == B[ii].objno .

Is there a numpy/pythonic way of doing this? I would not like to make any stipulations on the ordering or size of either A or B since they may be large fits tables. This is what I am doing now:

ii = np.in1d(A.OBJNO, B.OBJNO).nonzero()[0]
jj = [np.where(B.OBJNO == objno)[0][0] for objno in A[ii].field('OBJNO')]

Is there a better numpy array matching algorithm?

Answer 1

Denoting the sizes of your arrays N and M ( N > M ), your solution is O(N*M).

Assuming your arrays are large enough, you'll be better off with a O(N*logN) solution. You can achieve that by first sorting the bigger array (call is A ), and then performing a binary search (eg using bisect ) of each element from B in sorted_A .