Coverage for local_installation/dynasor/post_processing/average_runs.py: 95%

1import numpy as np 

2from dynasor.sample import Sample 

3from typing import List 

4from copy import deepcopy 

5 

6 

7def get_sample_averaged_over_independent_runs( 

8 samples: List[Sample], live_dangerously=False) -> Sample: 

9 """ 

10 Compute an averaged sample from multiple samples obtained from identical independent runs. 

11 

12 Note, all the meta_data and dimensions in all samples must be the same, 

13 else ValueError is raised (unless ``live_dangerously`` is set to True). 

14 

15 Parameters 

16 ---------- 

17 samples 

18 list of all sample objects to be averaged over 

19 live_dangerously 

20 setting True allows for averaging over samples which meta-data information is not identical. 

21 """ 

22 

23 # get meta data and dimensions from first sample 

24 sample_ref = samples[0] 

25 data_dict = dict() 

26 meta_data = deepcopy(sample_ref.meta_data) 

27 

28 # test that all samples have identical dimensions 

29 for sample in samples: 

30 assert sorted(sample.dimensions) == sorted(sample_ref.dimensions) 

31 for dim in sample_ref.dimensions: 

32 assert np.allclose(sample[dim], sample_ref[dim]) 

33 

34 for dim in sample_ref.dimensions: 

35 data_dict[dim] = sample_ref[dim] 

36 

37 # test that all samples have identical meta_data 

38 if not live_dangerously: 38 ↛ 54line 38 didn't jump to line 54, because the condition on line 38 was never false

39 for sample in samples: 

40 assert len(sample.meta_data) == len(meta_data) 

41 

42 for key, val in meta_data.items(): 

43 if isinstance(val, dict): 

44 for k, v in val.items(): 

45 assert sample_ref.meta_data[key][k] == sample.meta_data[key][k] 

46 elif isinstance(val, np.ndarray): 

47 assert np.allclose(sample.meta_data[key], val) 

48 elif isinstance(val, float): 48 ↛ 49line 48 didn't jump to line 49, because the condition on line 48 was never true

49 assert np.isclose(sample.meta_data[key], val) 

50 else: 

51 assert sample.meta_data[key] == val 

52 

53 # average all correlation functions 

54 for key in sample.available_correlation_functions: 

55 data = [] 

56 for sample in samples: 

57 data.append(sample[key]) 

58 data_average = np.nanmean(data, axis=0) 

59 data_dict[key] = data_average 

60 

61 return sample.__class__(data_dict, **meta_data)