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     parsed = email.utils.parsedate(text)
  18     if parsed:
  19         return datetime.datetime(*parsed[:6])
  20     else:
  21         datetime.datetime(1970, 1, 1)
  22
  23 def fresh_cache(url, properties):
  24     pr = properties[url]
  25     expires = None
  26
  27     logger.debug("Checking cache freshness for %s using %s", url, pr)
  28
  29     if "Cache-Control" in pr:
  30         if re.match(r"immutable", pr["Cache-Control"]):
  31             return True
  32
  33         g = re.match(r"(s-maxage|max-age)=(\d+)", pr["Cache-Control"])
  34         if g:
  35             expires = my_parsedate(pr["Date"]) + datetime.timedelta(seconds=int(g.group(2)))
  36
  37     if expires is None and "Expires" in pr:
  38         expires = my_parsedate(pr["Expires"])
  39
  40     if expires is None:
  41         # Use a default cache time of 24 hours if upstream didn't set
  42         # any cache headers, to reduce redundant downloads.
  43         expires = my_parsedate(pr["Date"]) + datetime.timedelta(hours=24)
  44
  45     if not expires:
  46         return False
  47
  48     return (datetime.datetime.utcnow() < expires)
  49
  50 def remember_headers(url, properties, headers):
  51     properties.setdefault(url, {})
  52     for h in ("Cache-Control", "ETag", "Expires", "Date", "Content-Length"):
  53         if h in headers:
  54             properties[url][h] = headers[h]
  55     if "Date" not in headers:
  56         properties[url]["Date"] = my_formatdate(datetime.datetime.utcnow())
  57
  58
  59 def changed(url, properties):
  60     req = requests.head(url, allow_redirects=True)
  61     remember_headers(url, properties, req.headers)
  62
  63     if req.status_code != 200:
  64         raise Exception("Got status %s" % req.status_code)
  65
  66     pr = properties[url]
  67     if "ETag" in pr and "ETag" in req.headers:
  68         if pr["ETag"] == req.headers["ETag"]:
  69             return False
  70     return True
  71
  72 def http_to_keep(api, project_uuid, url):
  73     r = api.collections().list(filters=[["properties", "exists", url]]).execute()
  74
  75     for item in r["items"]:
  76         properties = item["properties"]
  77         if fresh_cache(url, properties):
  78             # Do nothing
  79             cr = arvados.collection.CollectionReader(item["portable_data_hash"], api_client=api)
  80             return "keep:%s/%s" % (item["portable_data_hash"], cr.keys()[0])
  81
  82         if not changed(url, properties):
  83             # ETag didn't change, same content, just update headers
  84             api.collections().update(uuid=item["uuid"], body={"collection":{"properties": properties}}).execute()
  85             cr = arvados.collection.CollectionReader(item["portable_data_hash"], api_client=api)
  86             return "keep:%s/%s" % (item["portable_data_hash"], cr.keys()[0])
  87
  88     properties = {}
  89     req = requests.get(url, stream=True, allow_redirects=True)
  90
  91     if req.status_code != 200:
  92         raise Exception("Failed to download '%s' got status %s " % (req.status_code, url))
  93
  94     remember_headers(url, properties, req.headers)
  95
  96     logger.info("Downloading %s (%s bytes)", url, properties[url]["Content-Length"])
  97
  98     c = arvados.collection.Collection()
  99
 100     if req.headers.get("Content-Disposition"):
 101         grp = re.search(r'filename=("((\"|[^"])+)"|([^][()<>@,;:\"/?={} ]+))', req.headers["Content-Disposition"])
 102         if grp.groups(2):
 103             name = grp.groups(2)
 104         else:
 105             name = grp.groups(3)
 106     else:
 107         name = urlparse.urlparse(url).path.split("/")[-1]
 108
 109     count = 0
 110     start = time.time()
 111     checkpoint = start
 112     with c.open(name, "w") as f:
 113         for chunk in req.iter_content(chunk_size=1024):
 114             count += len(chunk)
 115             f.write(chunk)
 116             now = time.time()
 117             if (now - checkpoint) > 20:
 118                 bps = (float(count)/float(now - start))
 119                 logger.info("%2.1f%% complete, %3.2f MiB/s, %1.0f seconds left",
 120                             float(count * 100) / float(properties[url]["Content-Length"]),
 121                             bps/(1024*1024),
 122                             (int(properties[url]["Content-Length"])-count)/bps)
 123                 checkpoint = now
 124
 125     c.save_new(name="Downloaded from %s" % url, owner_uuid=project_uuid, ensure_unique_name=True)
 126
 127     api.collections().update(uuid=c.manifest_locator(), body={"collection":{"properties": properties}}).execute()
 128
 129     return "keep:%s/%s" % (c.portable_data_hash(), name)