sdk/cwl/arvados_cwl/http.py

   1 import requests
   2 import email.utils
   3 import time
   4 import datetime
   5 import re
   6 import arvados
   7 import arvados.collection
   8 import urlparse
   9 import logging
  10
  11 logger = logging.getLogger('arvados.cwl-runner')
  12
  13 def my_formatdate(dt):
  14     return email.utils.formatdate(timeval=time.mktime(now.timetuple()), localtime=False, usegmt=True)
  15
  16 def my_parsedate(text):
  17     return datetime.datetime(*email.utils.parsedate(text)[:6])
  18
  19 def fresh_cache(url, properties):
  20     pr = properties[url]
  21     expires = None
  22
  23     if "Cache-Control" in pr:
  24         if re.match(r"immutable", pr["Cache-Control"]):
  25             return True
  26
  27         g = re.match(r"(s-maxage|max-age)=(\d+)", pr["Cache-Control"])
  28         if g:
  29             expires = my_parsedate(pr["Date"]) + datetime.timedelta(seconds=int(g.group(2)))
  30
  31     if expires is None and "Expires" in pr:
  32         expires = my_parsedate(pr["Expires"])
  33
  34     if expires is None:
  35         # Use a default cache time of 24 hours if upstream didn't set
  36         # any cache headers, to reduce redundant downloads.
  37         expires = my_parsedate(pr["Date"]) + datetime.timedelta(hours=24)
  38
  39     if not expires:
  40         return False
  41
  42     return (datetime.datetime.utcnow() < expires)
  43
  44 def remember_headers(url, properties, headers):
  45     properties.setdefault(url, {})
  46     for h in ("Cache-Control", "ETag", "Expires", "Date", "Content-Length"):
  47         if h in headers:
  48             properties[url][h] = headers[h]
  49     if "Date" not in headers:
  50         properties[url]["Date"] = my_formatdate(datetime.datetime.utcnow())
  51
  52
  53 def changed(url, properties):
  54     req = requests.head(url)
  55     remember_headers(url, properties, req.headers)
  56
  57     if req.status_code != 200:
  58         raise Exception("Got status %s" % req.status_code)
  59
  60     pr = properties[url]
  61     if "ETag" in pr and "ETag" in req.headers:
  62         if pr["ETag"] == req.headers["ETag"]:
  63             return False
  64     return True
  65
  66 def http_to_keep(api, project_uuid, url):
  67     r = api.collections().list(filters=[["properties", "exists", url]]).execute()
  68     name = urlparse.urlparse(url).path.split("/")[-1]
  69
  70     for item in r["items"]:
  71         properties = item["properties"]
  72         if fresh_cache(url, properties):
  73             # Do nothing
  74             return "keep:%s/%s" % (item["portable_data_hash"], name)
  75
  76         if not changed(url, properties):
  77             # ETag didn't change, same content, just update headers
  78             api.collections().update(uuid=item["uuid"], body={"collection":{"properties": properties}}).execute()
  79             return "keep:%s/%s" % (item["portable_data_hash"], name)
  80
  81     properties = {}
  82     req = requests.get(url, stream=True)
  83
  84     if req.status_code != 200:
  85         raise Exception("Failed to download '%s' got status %s " % (req.status_code, url))
  86
  87     remember_headers(url, properties, req.headers)
  88
  89     logger.info("Downloading %s (%s bytes)", url, properties[url]["Content-Length"])
  90
  91     c = arvados.collection.Collection()
  92
  93     count = 0
  94     start = time.time()
  95     checkpoint = start
  96     with c.open(name, "w") as f:
  97         for chunk in req.iter_content(chunk_size=1024):
  98             count += len(chunk)
  99             f.write(chunk)
 100             now = time.time()
 101             if (now - checkpoint) > 20:
 102                 bps = (float(count)/float(now - start))
 103                 logger.info("%2.1f%% complete, %3.2f MiB/s, %1.0f seconds left",
 104                             float(count * 100) / float(properties[url]["Content-Length"]),
 105                             bps/(1024*1024),
 106                             (int(properties[url]["Content-Length"])-count)/bps)
 107                 checkpoint = now
 108
 109     c.save_new(name="Downloaded from %s" % url, owner_uuid=project_uuid, ensure_unique_name=True)
 110
 111     api.collections().update(uuid=c.manifest_locator(), body={"collection":{"properties": properties}}).execute()
 112
 113     return "keep:%s/%s" % (c.portable_data_hash(), name)