python - Numpy: Tuple-Indexing of rows in ndarray (for later selection of subsets) -

i'm new numpy, , not expierenced python programmer, please excuse me if seems trivial ;)

i writing script extract specific data out of several molecular-dynamics simulation. therefore read data out of files , modify , truncate them uniform length , add row-wise, form 2d-array each simulation run.

these arrays appended each other, 3d-array, each slice along z-axis represent dataset of specific simulation run. goal later on easy manipulation, e.g. averaging on simulation runs.

this give basic idea of done:

import numpy np  = np.zeros((2000), dtype = bool) = a.reshape((1, 2000))  # appending different rows form '2d-matrix',  # actual data per simulation run in xrange(1,103):     b = np.zeros((2000), dtype = bool)     b = b.reshape((1, 2000))     = np.concatenate((a, b), axis=0)  print a.shape # >>> (2000, 103)  c = np.expand_dims(a, axis=2) = np.expand_dims(a, axis=2)  print a.shape # >>> (2000, 103, 1)  # appending different '2d-matrices' form 3d array,  # each slice along z-axis representing 1 simulation run in xrange(1,50):     = np.concatenate((a, c), axis=2)   print a.shape # >>> (2000, 103, 50) 

so far good, actual question:

in 1 2d-array, each row represents different set of interacting atom-pairs. later on want create subsets of array, depending on different critera - e.g. 'show me pairs, distance x 10 < x <= 20'.

so when first add rows in for in xrange(1,103): ..., want include indexing of rows set of ints each row. data of atom pairs there anyway, @ moment i'm not including in ndarray.

i thinking of tuple, 2d-array like

[ [('int' a,'int' b), [false,true,false,...]],   [('int' a,'int' d), [true, false, true...]],  ... ] 

or that

[ [['int' a], ['int' b], [false,true,false,...]],   [['int' a], ['int' d], [true, false, true...]],  ... ] 

can think of or easier approach kind of filtering? i'm not quite sure if i'm on right track here , doesn't seem straight-forward have different datatypes in array that.

also notice, indexes ordered in same way in each 2d-array, because sort them (atm based on string) , add np.zeros() rows occur on other simulation runs. maybe lookup-table right approach?

thanks lot!

update/answer:

sorry, know question little bit specific , bloated code wasn't relevant question.

i answered question myself, , sake of documentation can find below. specific, maybe helps handle indexing in numpy.

short, general answer:

i created look-up-table python list , did simple numpy slicing operation selection mask, containing indices:

a = [[[1, 2],       [3, 4],       [5, 6]],       [[7, 8],       [9,10],       [11,12]]] = np.asarray(a)  # selects rows 1 , 2 each 2d array mask = [1,2] b = a[ : , mask, : ] 

which gives b:

[[[ 3  4]   [ 5  6]]   [[ 9 10]   [11 12]]] 

complete answer, specific question above:

this 2d array:

a =[[true, false, false, false, false],     [false, true, false, false, false],     [false, false, true, false, false]] = np.asarray(a) 

indexing of rows tuples, due specific problem e.g.:

lut = [(1,2),(3,4),(3,5)] 

append other 2d array form 3d array:

c = np.expand_dims(a, axis=0) = np.expand_dims(a, axis=0)  = np.concatenate((a, c), axis=0)  

this 3d array a:

 >[[[ true false false false false]     [false  true false false false]     [false false  true false false]]     [[ true false false false false]     [false  true false false false]     [false false  true false false]]] 

selecting rows, contain "3" in look-up-table

mask = [i i, v in enumerate(lut) if 3 in v]   > [1, 2] 

applying mask 3d-array:

b = a[ : , mask, : ] 

now b 3d array a after selection:

[[[false  true false false false]    [false false  true false false]]    [[false  true false false false]    [false false  true false false]]] 

to keep track of new indices of b: create new look-up-table further computation:

newlut = [v i, v in enumerate(lut) if in mask]   >[(3, 4), (3, 5)]