Result ordering after using pmap to parallel

[xywang181]

Dear all,

I am trying to use pytraj.pmap to get pairwise distances efficiently. But I figured out that the array was different from using pytraj.distance. My question is how to wrap the result back to the order as if using pytraj.distance without pmap? To be more specific, for example, my trajectory contains 20 frames. For each frame, I am trying to compute 3 pairwise distances. And I am trying to use 5 cores to parallel the computation. The array is obtained by using "result = pt.pmap(pt.distance, traj, mask=mask_lists, n_cores=5 )".

Thanks in advanced for any help.

[hainm]

Dear @xywang181: Can you please try:

pytraj/tests/test_parallel_pmap/test_pmap.py

Lines 245 to 252 in 3d844a4

	angle_ = pt.angle(traj, ':3 :4 :5')
	distance_ = pt.distance(traj, '@10 @20')
	data = pt.pmap(['angle :3 :4 :5', 'distance @10 @20'], traj, n_cores=2)
	assert isinstance(data, OrderedDict), 'must be OrderDict'
	arr = pt.tools.dict_to_ndarray(data)
	aa_eq(angle_, arr[0])
	aa_eq(distance_, arr[1])

[hainm]

Basically try to use dict_to_ndarray

[xywang181]

Thanks for the answer!

I used the dict_to_ndarray. But it gave me a 3D array, which I guess is separated by the parallel computations. How can I merge the 3D array back to 2D array with the right frame order?

[hainm]

Does pt.tools.dict_to_ndarray(data) [0] return what you are expecting?

(I don't have my machine to test pytraj currently).