Debian Codegraph @

benchmarks/api.py

raw
1 # ex: set sts=4 ts=4 sw=4 noet:
2 # ## ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ##
3 #
4 # See COPYING file distributed along with the datalad package for the
5 # copyright and license terms.
6 #
7 # ## ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ##
8 """Benchmarks of the datalad.api functionality"""
9
10 from os.path import join as opj
11
12 from datalad.api import (
13 create,
14 create_test_dataset,
15 drop,
16 install,
17 ls,
18 )
19
20 #
21 # Following ones could be absent in older versions
22 #
23 try:
24 from datalad.api import diff
25 except ImportError:
26 diff = None
27
28 try:
29 from datalad.api import status
30 except ImportError:
31 status = None
32
33
34 # Some tracking example -- may be we should track # of datasets.datalad.org
35 #import gc
36 #def track_num_objects():
37 # return len(gc.get_objects())
38 #track_num_objects.unit = "objects"
39
40
41 from .common import (
42 SampleSuperDatasetBenchmarks,
43 SuprocBenchmarks,
44 )
45
46
47 class testds(SuprocBenchmarks):
48 """
49 Benchmarks to test on create_test_dataset how fast we could generate datasets
50 """
51
52 def time_create_test_dataset1(self):
53 self.remove_paths.extend(
54 create_test_dataset(spec='1', seed=0)
55 )
56
57 def time_create_test_dataset2x2(self):
58 self.remove_paths.extend(
59 create_test_dataset(spec='2/2', seed=0)
60 )
61
62
63 class supers(SampleSuperDatasetBenchmarks):
64 """
65 Benchmarks on common operations on collections of datasets using datalad API
66 """
67
68 def time_installr(self):
69 # somewhat duplicating setup but lazy to do different one for now
70 assert install(self.ds.path + '_', source=self.ds.path, recursive=True)
71
72 def time_createadd(self):
73 assert self.ds.create('newsubds')
74
75 def time_createadd_to_dataset(self):
76 subds = create(opj(self.ds.path, 'newsubds'))
77 self.ds.save(subds.path)
78
79 def time_ls(self):
80 ls(self.ds.path)
81
82 def time_ls_recursive(self):
83 ls(self.ds.path, recursive=True)
84
85 def time_ls_recursive_long_all(self):
86 ls(self.ds.path, recursive=True, long_=True, all_=True)
87
88 def time_subdatasets(self):
89 self.ds.subdatasets()
90
91 def time_subdatasets_with_all_paths_recursive(self):
92 # to see if we do not get O(N^2) performance
93 subdatasets = self.ds.subdatasets(recursive=True, result_xfm='relpaths')
94 subdatasets2 = self.ds.subdatasets(path=subdatasets, recursive=True, result_xfm='relpaths')
95 assert subdatasets == subdatasets2
96
97 def time_subdatasets_recursive(self):
98 self.ds.subdatasets(recursive=True)
99
100 def time_subdatasets_recursive_first(self):
101 next(self.ds.subdatasets(recursive=True, return_type='generator'))
102
103 def time_uninstall(self):
104 for subm in self.ds.repo.get_submodules_():
105 self.ds.drop(subm["path"], recursive=True, what='all',
106 reckless='kill')
107
108 def time_remove(self):
109 self.ds.drop(what='all', reckless='kill', recursive=True)
110
111 def time_diff(self):
112 self.ds.diff(fr="HEAD^")
113
114 def time_diff_recursive(self):
115 self.ds.diff(fr="HEAD^", recursive=True)
116
117 # Status must be called with the dataset, unlike diff
118 def time_status(self):
119 self.ds.status()
120
121 def time_status_recursive(self):
122 self.ds.status(recursive=True)
123
124
125 supers.time_remove.warmup_time = 0