annotate alveo_get_primary_text.py @ 4:3a9f20428cff draft

planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
author stevecassidy
date Wed, 31 Aug 2016 22:07:27 -0400
parents
children 2f4907372748
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
4
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
1 from __future__ import print_function
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
2 import json
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
3 import argparse
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
4 import pyalveo
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
5 import sys
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
6 import os
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
7 from fnmatch import fnmatch
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
8
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
9 API_URL = 'https://app.alveo.edu.au' # TODO: export constants to a separate module
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
10
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
11 def parser():
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
12 parser = argparse.ArgumentParser(description="Downloads documents in an Alveo Item List")
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
13 parser.add_argument('--api_key', required=True, action="store", type=str, help="Alveo API key")
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
14 parser.add_argument('--item_list_url', required=True, action="store", type=str, help="Item List to download")
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
15 parser.add_argument('--output_path', required=True, action="store", type=str, help="Path to output file")
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
16 return parser.parse_args()
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
17
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
18 def get_item_list(api_key, item_list_url):
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
19 client = pyalveo.Client(api_key=api_key, api_url=API_URL, use_cache=False)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
20 return client.get_item_list(item_list_url)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
21
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
22 # this file name pattern allows galaxy to discover the dataset designation and type
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
23 FNPAT = "%(designation)s_%(ext)s"
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
24
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
25 def galaxy_name(fname, ext):
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
26 """construct a filename suitable for Galaxy dataset discovery"""
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
27
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
28 fname = FNPAT % {'designation': fname, 'ext': ext}
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
29
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
30 return fname
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
31 import pprint
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
32 def download_documents(item_list, output_path):
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
33 """
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
34 Downloads a list of documents to the directory specificed by output_path.
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
35
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
36 :type documents: list of pyalveo.Document
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
37 :param documents: Documents to download
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
38
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
39 :type output_path: String
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
40 :param output_path: directory to download to the documents to
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
41 """
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
42 if not os.path.exists(output_path):
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
43 os.makedirs(output_path)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
44
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
45 downloaded = []
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
46
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
47 items = item_list.get_all()
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
48 filtered_documents = []
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
49 for item in items:
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
50 md = item.metadata()
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
51 fname = os.path.join(output_path, galaxy_name(md['alveo:metadata']['dc:identifier'], 'txt'))
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
52 content = item.get_primary_text()
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
53 if not content == None:
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
54 with open(fname, 'w') as out:
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
55 out.write(content)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
56
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
57 return downloaded
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
58
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
59 def main():
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
60 args = parser()
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
61 try:
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
62 api_key = open(args.api_key, 'r').read().strip()
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
63 item_list = get_item_list(api_key, args.item_list_url)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
64 downloaded = download_documents(item_list, args.output_path)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
65 # write out a list of downloaded files as a result?
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
66 except pyalveo.APIError as e:
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
67 print("ERROR: " + str(e), file=sys.stderr)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
68 sys.exit(1)
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
69
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
70 if __name__ == '__main__':
3a9f20428cff planemo upload commit f36456464c692ed9d39a9cf654d09fe793113cce-dirty
stevecassidy
parents:
diff changeset
71 main()