Debian Codegraph @

benchmarks/usecases.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 for some use cases, typically at datalad.api level"""
9
10 import os.path as osp
11 import sys
12 import tempfile
13 from os.path import join as opj
14
15 from datalad.api import create
16 from datalad.utils import (
17 create_tree,
18 get_tempfile_kwargs,
19 rmtree,
20 )
21
22 from .common import SuprocBenchmarks
23
24
25 class study_forrest(SuprocBenchmarks):
26 """
27 Benchmarks for Study Forrest use cases
28 """
29
30 timeout = 180 # especially with profiling might take longer than default 60s
31
32 def setup(self):
33 self.path = tempfile.mkdtemp(**get_tempfile_kwargs({}, prefix='bm_forrest'))
34
35 def teardown(self):
36 if osp.exists(self.path):
37 rmtree(self.path)
38
39 def time_make_studyforrest_mockup(self):
40 path = self.path
41 # Carries a copy of the
42 # datalad.tests.utils_testdatasets.py:make_studyforrest_mockup
43 # as of 0.12.0rc2-76-g6ba6d53b
44 # A copy is made so we do not reflect in the benchmark results changes
45 # to that helper's code. This copy only tests on 2 not 3 analyses
46 # subds
47 public = create(opj(path, 'public'), description="umbrella dataset")
48 # the following tries to capture the evolution of the project
49 phase1 = public.create('phase1',
50 description='old-style, no connection to RAW')
51 structural = public.create('structural', description='anatomy')
52 tnt = public.create('tnt', description='image templates')
53 tnt.clone(source=phase1.path, path=opj('src', 'phase1'), reckless=True)
54 tnt.clone(source=structural.path, path=opj('src', 'structural'), reckless=True)
55 aligned = public.create('aligned', description='aligned image data')
56 aligned.clone(source=phase1.path, path=opj('src', 'phase1'), reckless=True)
57 aligned.clone(source=tnt.path, path=opj('src', 'tnt'), reckless=True)
58 # new acquisition
59 labet = create(opj(path, 'private', 'labet'), description="raw data ET")
60 phase2_dicoms = create(opj(path, 'private', 'p2dicoms'), description="raw data P2MRI")
61 phase2 = public.create('phase2',
62 description='new-style, RAW connection')
63 phase2.clone(source=labet.path, path=opj('src', 'labet'), reckless=True)
64 phase2.clone(source=phase2_dicoms.path, path=opj('src', 'dicoms'), reckless=True)
65 # add to derivatives
66 tnt.clone(source=phase2.path, path=opj('src', 'phase2'), reckless=True)
67 aligned.clone(source=phase2.path, path=opj('src', 'phase2'), reckless=True)
68 # never to be published media files
69 media = create(opj(path, 'private', 'media'), description="raw data ET")
70 # assuming all annotations are in one dataset (in reality this is also
71 # a superdatasets with about 10 subdatasets
72 annot = public.create('annotations', description='stimulus annotation')
73 annot.clone(source=media.path, path=opj('src', 'media'), reckless=True)
74 # a few typical analysis datasets
75 # (just doing 2, actual status quo is just shy of 10)
76 # and also the real goal -> meta analysis
77 metaanalysis = public.create('metaanalysis', description="analysis of analyses")
78 for i in range(1, 2):
79 ana = public.create('analysis{}'.format(i),
80 description='analysis{}'.format(i))
81 ana.clone(source=annot.path, path=opj('src', 'annot'), reckless=True)
82 ana.clone(source=aligned.path, path=opj('src', 'aligned'), reckless=True)
83 ana.clone(source=tnt.path, path=opj('src', 'tnt'), reckless=True)
84 # link to metaanalysis
85 metaanalysis.clone(source=ana.path, path=opj('src', 'ana{}'.format(i)),
86 reckless=True)
87 # simulate change in an input (but not raw) dataset
88 create_tree(
89 aligned.path,
90 {'modification{}.txt'.format(i): 'unique{}'.format(i)})
91 aligned.save('.')
92 # finally aggregate data
93 aggregate = public.create('aggregate', description='aggregate data')
94 aggregate.clone(source=aligned.path, path=opj('src', 'aligned'), reckless=True)
95 # the toplevel dataset is intentionally left dirty, to reflect the
96 # most likely condition for the joint dataset to be in at any given
97 # point in time