annotate validate_affy_metadata.py @ 11:96fbb0504ac9 draft

Uploaded
author greg
date Thu, 15 Nov 2018 10:22:56 -0500
parents d15ae203013e
children 977280cfef63
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
1 #!/usr/bin/env python
ae030e6e838c Uploaded
greg
parents:
diff changeset
2 """
ae030e6e838c Uploaded
greg
parents:
diff changeset
3 Validate the metadata file associated with Affymetrix 96 well plate data.
ae030e6e838c Uploaded
greg
parents:
diff changeset
4 """
ae030e6e838c Uploaded
greg
parents:
diff changeset
5 import argparse
ae030e6e838c Uploaded
greg
parents:
diff changeset
6 import datetime
ae030e6e838c Uploaded
greg
parents:
diff changeset
7 import decimal
ae030e6e838c Uploaded
greg
parents:
diff changeset
8 import re
ae030e6e838c Uploaded
greg
parents:
diff changeset
9 import shutil
ae030e6e838c Uploaded
greg
parents:
diff changeset
10 import sys
ae030e6e838c Uploaded
greg
parents:
diff changeset
11
ae030e6e838c Uploaded
greg
parents:
diff changeset
12 parser = argparse.ArgumentParser()
ae030e6e838c Uploaded
greg
parents:
diff changeset
13 parser.add_argument('--input', dest='input', help='Metadata file for Affymetrix 96 well plate data')
ae030e6e838c Uploaded
greg
parents:
diff changeset
14 parser.add_argument('--output', dest='output', help='Output dataset'),
ae030e6e838c Uploaded
greg
parents:
diff changeset
15 args = parser.parse_args()
ae030e6e838c Uploaded
greg
parents:
diff changeset
16
ae030e6e838c Uploaded
greg
parents:
diff changeset
17 EMAIL_MAX_LEN = 255
ae030e6e838c Uploaded
greg
parents:
diff changeset
18 VALID_EMAIL_RE = re.compile("[^@]+@[^@]+\.[^@]+")
ae030e6e838c Uploaded
greg
parents:
diff changeset
19
ae030e6e838c Uploaded
greg
parents:
diff changeset
20
ae030e6e838c Uploaded
greg
parents:
diff changeset
21 def add_error_msg(accumulated_msgs, msg):
ae030e6e838c Uploaded
greg
parents:
diff changeset
22 return "%s\n%s" % (accumulated_msgs, msg)
ae030e6e838c Uploaded
greg
parents:
diff changeset
23
ae030e6e838c Uploaded
greg
parents:
diff changeset
24
ae030e6e838c Uploaded
greg
parents:
diff changeset
25 def empty_value(line_no, label, accumulated_msgs):
ae030e6e838c Uploaded
greg
parents:
diff changeset
26 return add_error_msg(accumulated_msgs, "The required %s value is missing on line %d." % (label, line_no))
ae030e6e838c Uploaded
greg
parents:
diff changeset
27
ae030e6e838c Uploaded
greg
parents:
diff changeset
28
ae030e6e838c Uploaded
greg
parents:
diff changeset
29 def stop_error(msg):
ae030e6e838c Uploaded
greg
parents:
diff changeset
30 sys.exit(msg)
ae030e6e838c Uploaded
greg
parents:
diff changeset
31
ae030e6e838c Uploaded
greg
parents:
diff changeset
32
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
33 def string_as_boolean_string(string):
d15ae203013e Uploaded
greg
parents: 8
diff changeset
34 if str(string).lower() in ['true', 'yes', 'on', '1']:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
35 return 'True'
d15ae203013e Uploaded
greg
parents: 8
diff changeset
36 else:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
37 return 'False'
d15ae203013e Uploaded
greg
parents: 8
diff changeset
38
d15ae203013e Uploaded
greg
parents: 8
diff changeset
39
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
40 def validate_date_string(line_no, date_string, accumulated_msgs):
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
41 if len(date_string) == 0:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
42 return accumulated_msgs
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
43 try:
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
44 datetime.datetime.strptime(date_string, '%Y-%m-%d')
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
45 return accumulated_msgs
ae030e6e838c Uploaded
greg
parents:
diff changeset
46 except ValueError:
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
47 return add_error_msg(accumulated_msgs, "Line %d contains an incorrect date format (%s must be YYYY-MM-DD)." % (line_no, date_string))
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
48
ae030e6e838c Uploaded
greg
parents:
diff changeset
49
ae030e6e838c Uploaded
greg
parents:
diff changeset
50 def validate_decimal(line_no, decimal_string, accumulated_msgs):
ae030e6e838c Uploaded
greg
parents:
diff changeset
51 try:
ae030e6e838c Uploaded
greg
parents:
diff changeset
52 decimal.Decimal(decimal_string)
ae030e6e838c Uploaded
greg
parents:
diff changeset
53 return accumulated_msgs
ae030e6e838c Uploaded
greg
parents:
diff changeset
54 except Exception:
ae030e6e838c Uploaded
greg
parents:
diff changeset
55 return add_error_msg(accumulated_msgs, "Line %d contains an incorrect decimal value (%s)." % (line_no, decimal_string))
ae030e6e838c Uploaded
greg
parents:
diff changeset
56
ae030e6e838c Uploaded
greg
parents:
diff changeset
57
ae030e6e838c Uploaded
greg
parents:
diff changeset
58 def validate_email(line_no, email, accumulated_msgs):
ae030e6e838c Uploaded
greg
parents:
diff changeset
59 if not (VALID_EMAIL_RE.match(email)):
ae030e6e838c Uploaded
greg
parents:
diff changeset
60 return add_error_msg(accumulated_msgs, "Line %d contains an invalid email address (%s). " % (line_no, email))
ae030e6e838c Uploaded
greg
parents:
diff changeset
61 elif len(email) > EMAIL_MAX_LEN:
ae030e6e838c Uploaded
greg
parents:
diff changeset
62 return add_error_msg(accumulated_msgs, "Line %d contains an email address (%) that is longer than the maximum length, %d characters." % (line_no, email))
ae030e6e838c Uploaded
greg
parents:
diff changeset
63 return accumulated_msgs
ae030e6e838c Uploaded
greg
parents:
diff changeset
64
ae030e6e838c Uploaded
greg
parents:
diff changeset
65
ae030e6e838c Uploaded
greg
parents:
diff changeset
66 accumulated_msgs = ""
ae030e6e838c Uploaded
greg
parents:
diff changeset
67 # Parse the input file, skipping the header, and validating
ae030e6e838c Uploaded
greg
parents:
diff changeset
68 # that each data line consists of 31 comma-separated items.
ae030e6e838c Uploaded
greg
parents:
diff changeset
69 with open(args.input, "r") as ih:
1
613186952d42 Uploaded
greg
parents: 0
diff changeset
70 for i, line in enumerate(ih):
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
71 if i == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
72 # Skip the header.
ae030e6e838c Uploaded
greg
parents:
diff changeset
73 continue
ae030e6e838c Uploaded
greg
parents:
diff changeset
74 line = line.rstrip("\r\n")
ae030e6e838c Uploaded
greg
parents:
diff changeset
75 if i > 97:
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
76 accumulated_msgs = add_error_msg(accumulated_msgs, "The input file contains more than 97 lines (must be 1 header line and no more than 96 data lines).")
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
77 stop_error(accumulated_msgs)
3
017f975149b2 Uploaded
greg
parents: 2
diff changeset
78 items = line.split(",")
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
79 if len(items) != 29:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
80 accumulated_msgs = add_error_msg(accumulated_msgs, "Line %d contains %s columns, (must be 29)." % (i, len(items)))
2
f3ca1cd544fa Uploaded
greg
parents: 1
diff changeset
81 stop_error(accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
82 # Required and validated.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
83 date_entered_db = items[0]
4
f8488ab257ac Uploaded
greg
parents: 3
diff changeset
84 accumulated_msgs = validate_date_string(i, date_entered_db, accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
85 # Required.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
86 user_specimen_id = items[1]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
87 if len(user_specimen_id) == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
88 accumulated_msgs = empty_value(i, "user_specimen_id", accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
89 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
90 field_call = items[2]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
91 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
92 bcoral_genet_id = items[3]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
93 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
94 bsym_genet_id = items[4]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
95 # Required.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
96 reef = items[5]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
97 if len(reef) == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
98 accumulated_msgs = empty_value(i, "reef", accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
99 # Required.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
100 region = items[6]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
101 if len(region) == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
102 accumulated_msgs = empty_value(i, "region", accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
103 # Required and validated.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
104 latitude = items[7]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
105 accumulated_msgs = validate_decimal(i, latitude, accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
106 # Required and validated.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
107 longitude = items[8]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
108 accumulated_msgs = validate_decimal(i, longitude, accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
109 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
110 geographic_origin = items[9]
d15ae203013e Uploaded
greg
parents: 8
diff changeset
111 # Optional.
d15ae203013e Uploaded
greg
parents: 8
diff changeset
112 sample_location = items[10]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
113 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
114 latitude_outplant = items[11]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
115 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
116 longitude_outplant = items[12]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
117 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
118 depth = items[13]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
119 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
120 dist_shore = items[14]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
121 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
122 disease_resist = items[15]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
123 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
124 bleach_resist = items[16]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
125 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
126 mortality = items[17]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
127 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
128 tle = items[18]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
129 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
130 spawning = string_as_boolean_string(items[19])
d15ae203013e Uploaded
greg
parents: 8
diff changeset
131 # Required.
d15ae203013e Uploaded
greg
parents: 8
diff changeset
132 collector_last_name = items[21]
d15ae203013e Uploaded
greg
parents: 8
diff changeset
133 if len(collector_last_name) == 0:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
134 accumulated_msgs = empty_value(i, "collector_last_name", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
135 # Required.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
136 collector_first_name = items[22]
d15ae203013e Uploaded
greg
parents: 8
diff changeset
137 if len(collector_first_name) == 0:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
138 accumulated_msgs = empty_value(i, "collector_first_name", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
139 # Required.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
140 org = items[23]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
141 if len(org) == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
142 accumulated_msgs = empty_value(i, "org", accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
143 # Required and validated.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
144 collection_date = items[24]
4
f8488ab257ac Uploaded
greg
parents: 3
diff changeset
145 accumulated_msgs = validate_date_string(i, date_entered_db, accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
146 # Required and validated.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
147 contact_email = items[25]
5
17d62e93d029 Uploaded
greg
parents: 4
diff changeset
148 accumulated_msgs = validate_email(i, contact_email, accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
149 # Required.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
150 seq_facility = items[26]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
151 if len(seq_facility) == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
152 accumulated_msgs = empty_value(i, "seq_facility", accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
153 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
154 array_version = items[27]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
155 # Optional.
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
156 public = string_as_boolean_string(items[28])
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
157 # Optional.
11
96fbb0504ac9 Uploaded
greg
parents: 9
diff changeset
158 public_after_date = items[29]
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
159 accumulated_msga = validate_date_string(public_after_date)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
160
6
11d22a27c24a Uploaded
greg
parents: 5
diff changeset
161 if len(accumulated_msgs) > 0:
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
162 stop_error(accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
163
8
0735cfcc7490 Uploaded
greg
parents: 7
diff changeset
164 shutil.copyfile(args.input, args.output)