aboutsummaryrefslogtreecommitdiff
path: root/test/requests/correlation_tests.py
blob: 0915dc68ae961f46f3201f1f4da7939a503040c4 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
import sys
import html
import requests
from lxml.html import parse
from link_checker import check_page

def corrs_base_data():
    return [
        {
            "dataset": "HC_M2_0606_P",
            "trait_id": "1435464_at",
            "corr_dataset": "HC_M2_0606_P",
        },
        {
            "dataset": "HC_M2_0606_P",
            "trait_id": "1457545_at",
            "corr_dataset": "HC_M2_0606_R",
        },
        {
            "dataset": "HC_M2_0606_P",
            "trait_id": "1442370_at",
            "corr_dataset": "BXDPublish",
        }
    ]
def sample_vals():
    return '{"C57BL/6J":"10.835","DBA/2J":"11.142","B6D2F1":"11.126","D2B6F1":"11.143","BXD1":"10.811","BXD2":"11.503","BXD5":"10.766","BXD6":"10.986","BXD8":"11.050","BXD9":"10.822","BXD11":"10.670","BXD12":"10.946","BXD13":"10.890","BXD14":"x","BXD15":"10.884","BXD16":"11.222","BXD18":"x","BXD19":"10.968","BXD20":"10.962","BXD21":"10.906","BXD22":"11.080","BXD23":"11.046","BXD24":"11.146","BXD24a":"x","BXD25":"x","BXD27":"11.078","BXD28":"11.034","BXD29":"10.808","BXD30":"x","BXD31":"11.087","BXD32":"11.029","BXD33":"10.662","BXD34":"11.482","BXD35":"x","BXD36":"x","BXD37":"x","BXD38":"10.836","BXD39":"10.926","BXD40":"10.638","BXD41":"x","BXD42":"10.974","BXD43":"10.828","BXD44":"10.900","BXD45":"11.358","BXD48":"11.042","BXD48a":"10.975","BXD49":"x","BXD50":"11.228","BXD51":"11.126","BXD52":"x","BXD53":"x","BXD54":"x","BXD55":"11.580","BXD56":"x","BXD59":"x","BXD60":"10.829","BXD61":"11.152","BXD62":"11.156","BXD63":"10.942","BXD64":"10.506","BXD65":"11.126","BXD65a":"11.272","BXD65b":"11.157","BXD66":"11.071","BXD67":"11.080","BXD68":"10.997","BXD69":"11.096","BXD70":"11.152","BXD71":"x","BXD72":"x","BXD73":"11.262","BXD73a":"11.444","BXD73b":"x","BXD74":"10.974","BXD75":"11.150","BXD76":"10.920","BXD77":"10.928","BXD78":"x","BXD79":"11.371","BXD81":"x","BXD83":"10.946","BXD84":"11.181","BXD85":"10.992","BXD86":"10.770","BXD87":"11.200","BXD88":"x","BXD89":"10.930","BXD90":"11.183","BXD91":"x","BXD93":"11.056","BXD94":"10.737","BXD95":"x","BXD98":"10.986","BXD99":"10.892","BXD100":"x","BXD101":"x","BXD102":"x","BXD104":"x","BXD105":"x","BXD106":"x","BXD107":"x","BXD108":"x","BXD109":"x","BXD110":"x","BXD111":"x","BXD112":"x","BXD113":"x","BXD114":"x","BXD115":"x","BXD116":"x","BXD117":"x","BXD119":"x","BXD120":"x","BXD121":"x","BXD122":"x","BXD123":"x","BXD124":"x","BXD125":"x","BXD126":"x","BXD127":"x","BXD128":"x","BXD128a":"x","BXD130":"x","BXD131":"x","BXD132":"x","BXD133":"x","BXD134":"x","BXD135":"x","BXD136":"x","BXD137":"x","BXD138":"x","BXD139":"x","BXD141":"x","BXD142":"x","BXD144":"x","BXD145":"x","BXD146":"x","BXD147":"x","BXD148":"x","BXD149":"x","BXD150":"x","BXD151":"x","BXD152":"x","BXD153":"x","BXD154":"x","BXD155":"x","BXD156":"x","BXD157":"x","BXD160":"x","BXD161":"x","BXD162":"x","BXD165":"x","BXD168":"x","BXD169":"x","BXD170":"x","BXD171":"x","BXD172":"x","BXD173":"x","BXD174":"x","BXD175":"x","BXD176":"x","BXD177":"x","BXD178":"x","BXD180":"x","BXD181":"x","BXD183":"x","BXD184":"x","BXD186":"x","BXD187":"x","BXD188":"x","BXD189":"x","BXD190":"x","BXD191":"x","BXD192":"x","BXD193":"x","BXD194":"x","BXD195":"x","BXD196":"x","BXD197":"x","BXD198":"x","BXD199":"x","BXD200":"x","BXD201":"x","BXD202":"x","BXD203":"x","BXD204":"x","BXD205":"x","BXD206":"x","BXD207":"x","BXD208":"x","BXD209":"x","BXD210":"x","BXD211":"x","BXD212":"x","BXD213":"x","BXD214":"x","BXD215":"x","BXD216":"x","BXD217":"x","BXD218":"x","BXD219":"x","BXD220":"x"}'

def do_request(url, data):
    return requests.post(
        url,
        data={
            "dataset": "HC_M2_0606_P",
            "trait_id": "1435464_at",
            "corr_dataset": "HC_M2_0606_P",
            "corr_sample_method": "pearson",
            "corr_return_results": "100",
            "corr_samples_group": "samples_primary",
            "sample_vals": sample_vals(),
            **data,
        })

def check_sample_correlations(baseurl, base_data):
    data = {
        **base_data,
        "corr_type": "sample",
        "corr_sample_method": "pearson",
        "location_type": "gene",
        "corr_return_results": "200"
    }
    top_n_message = "The top 200 correlations ranked by the Genetic Correlation"
    result = do_request(f"{baseurl}/corr_compute", data)
    assert result.status_code == 200
    assert (result.text.find(f"Values of record {base_data['trait_id']}") >= 0), result.text
    assert (result.text.find(top_n_message) >= 0), result.text

def check_tissue_correlations(baseurl, base_data):
    data = {
        **base_data,
        "corr_type": "tissue",
        "location_type": "gene",
    }
    result = do_request(f"{baseurl}/corr_compute", data)

    assert result.status_code == 200
    if (data["trait_id"] == "1442370_at"
        and data["corr_dataset"] in ("BXDPublish",)):
        top_n_message = (
            "It is not possible to compute the 'Tissue' correlations between "
            f"trait '{data['trait_id']}' and the data")
    else:
        top_n_message = "The top 100 correlations ranked by the Tissue Correlation"
        assert (result.text.find(f"Values of record {base_data['trait_id']}") >= 0), result.text

    assert (html.unescape(result.text).find(top_n_message) >= 0), (
        f"NOT FOUND: {top_n_message}")

def check_lit_correlations(baseurl, base_data):
    data = {
        **base_data,
        "corr_type": "lit",
        "corr_return_results": "200"
    }
    result = do_request(f"{baseurl}/corr_compute", data)

    assert result.status_code == 200
    if (data["trait_id"] == "1442370_at"
        and data["corr_dataset"] in ("BXDPublish",)):
        top_n_message = (
            "It is not possible to compute the 'Literature' correlations "
            f"between trait '{data['trait_id']}' and the data")
    else:
        top_n_message = "The top 200 correlations ranked by the Literature Correlation"
        assert (result.text.find(f"Values of record {base_data['trait_id']}") >= 0), result.text

    assert (html.unescape(result.text).find(top_n_message) >= 0), (
        f"NOT FOUND: {top_n_message}")

def check_correlations(args_obj, parser):
    print("")
    print("Checking the correlations...")
    corr_type_fns = {
        "sample": check_sample_correlations,
        "tissue": check_tissue_correlations,
        "lit": check_lit_correlations
    }
    host = args_obj.host
    failure = False
    for corr_type, corr_type_fn in corr_type_fns.items():
        for corr_base in corrs_base_data():
            try:
                print(f"\tChecking {corr_type} correlations...", end="")
                corr_type_fn(host, corr_base)
                print(" ok")
            except AssertionError as asserterr:
                print (f" fail: {asserterr.args[0]}")
                failure = True

    if failure:
        print("FAIL!")
        sys.exit(1)
    print("OK")