Add ability to linkify hashes on isolate server

appengine/isolate/handlers_frontend.py

 # Copyright 2012 The LUCI Authors. All rights reserved.
 # Use of this source code is governed by the Apache v2.0 license that can be
 # found in the LICENSE file.
 
 """This module defines Isolate Server frontend url handlers."""
 
 import datetime
 import json
+import re
 
 import webapp2
 
 from google.appengine.api import memcache
 from google.appengine.api import users
 
 import acl
 import config
 import gcs
 import mapreduce_jobs
@@ skipping 125 matching lines @@
 ### Non-restricted handlers
 
 
 class BrowseHandler(auth.AuthenticatingHandler):
   @auth.autologin
   @auth.require(acl.isolate_readable)
   def get(self):
     namespace = self.request.get('namespace', 'default-gzip')
     # Support 'hash' for compatibility with old links. To remove eventually.
     digest = self.request.get('digest', '') or self.request.get('hash', '')
+    params = {
+      u'digest': unicode(digest),
+      u'namespace': unicode(namespace),
+      u'onload': '',
+    }
+    # Check for existence of element, so we can 400/404
+    if digest and namespace:
+      try:
+        model.get_content(namespace, digest)
+      except ValueError:
+        self.abort(400, 'Invalid key')
+      except LookupError:
+          self.abort(404, 'Unable to retrieve the entry')

[M-A Ruel, 2016/04/14 14:05:06]

alignment

[kjlubick, 2016/04/14 14:57:37]

Done.

+    self.response.write(template.render('isolate/browse.html', params))
+
+
+class ContentHandler(auth.AuthenticatingHandler):
+  @auth.autologin
+  @auth.require(acl.isolate_readable)
+  def get(self):
+    namespace = self.request.get('namespace', 'default-gzip')
+    digest = self.request.get('digest', '')
+
     content = None

[M-A Ruel, 2016/04/14 14:05:06]

This can be removed.

[kjlubick, 2016/04/14 14:57:37]

Done.

     if digest and namespace:
-      # TODO(maruel): Refactor into a function.
-      memcache_entry = memcache.get(digest, namespace='table_%s' % namespace)
-      if memcache_entry is not None:
-        raw_data = memcache_entry
-      else:
-        try:
-          key = model.get_entry_key(namespace, digest)
-        except ValueError:
-          self.abort(400, 'Invalid key')
-        entity = key.get()
-        if entity is None:
-          self.abort(404, 'Unable to retrieve the entry')
-        raw_data = entity.content
+      raw_data = None

[M-A Ruel, 2016/04/14 14:05:06]

You need to move up above the condition, otherwise it'd fail line 188 with
variable uninitialized.

[kjlubick, 2016/04/14 14:57:37]

Good catch.  Done.

+      try:
+        raw_data = model.get_content(namespace, digest)
+      except ValueError:
+        self.abort(400, 'Invalid key')
+      except LookupError:
+        self.abort(404, 'Unable to retrieve the entry')
+
       if not raw_data:
+        key = model.get_entry_key(namespace, digest)

[M-A Ruel, 2016/04/14 14:05:06]

In this case, the entity is effectively loaded twice, this is suboptimal.

[kjlubick, 2016/04/14 14:57:37]

Done.  entity is returned as the second argument.

         stream = gcs.read_file(config.settings().gs_bucket, key.id())
       else:
         stream = [raw_data]
       content = ''.join(model.expand_content(namespace, stream))
+
+      self.response.headers['X-Frame-Options'] = 'SAMEORIGIN'
+      del self.response.headers['Content-Type']
+      # Apparently, setting the content type to text/plain encourages the
+      # browser (Chrome, at least) to sniff the mime type and display
+      # things like images.  Images are autowrapped in <img> and text is
+      # wrapped in <pre>.
+      self.response.headers['Content-Type'] = 'text/plain; charset=utf-8'
+      self.response.headers['Content-Disposition'] = str('filename=%s' % digest)
       if content.startswith('{'):
         # Try to format as JSON.
         try:
           content = json.dumps(
               json.loads(content), sort_keys=True, indent=2,
               separators=(',', ': '))
+          # If we don't wrap this in html, browsers will put content in a pre

[M-A Ruel, 2016/04/14 14:05:06]

Thanks for the explanation, I didn't know this.

+          # tag which is also styled with monospace/pre-wrap.  We can't use
+          # anchor tags in <pre>, so we force it to be a <div>, which happily
+          # accepts links.
+          content = (
+            '<div style="font-family:monospace;white-space:pre-wrap;">%s</div>'
+             % content)
+          # Linkify things that look like hashes
+          content = re.sub(r'([0-9a-f]{40})',
+            r'<a target="_blank" href="browse?namespace=default-gzip' +

[M-A Ruel, 2016/04/14 14:05:06]

use the namespcae instead of hardcoding 'default-gzip'

[kjlubick, 2016/04/14 14:57:37]

Done.

+              r'&digest=\1">\1</a>',
+            content)
+          self.response.headers['Content-Type'] = 'text/html; charset=utf-8'
         except ValueError:
           pass
-      content = content.decode('utf8', 'replace')
-    params = {
-      u'content': content,
-      u'digest': unicode(digest),
-      u'namespace': unicode(namespace),
-      # TODO(maruel): Add back once Web UI authentication is switched to OAuth2.
-      #'onload': 'update()' if digest else '',
-      u'onload': '',
-    }
-    self.response.write(template.render('isolate/browse.html', params))
+
+    self.response.write(content)
 
 
 class StatsHandler(webapp2.RequestHandler):
   """Returns the statistics web page."""
   def get(self):
     """Presents nice recent statistics.
 
     It fetches data from the 'JSON' API.
     """
     # Preloads the data to save a complete request.
@@ skipping 89 matching lines @@
       # Administrative urls.
       webapp2.Route(r'/restricted/config', RestrictedConfigHandler),
 
       # Mapreduce related urls.
       webapp2.Route(
           r'/restricted/launch_mapreduce',
           RestrictedLaunchMapReduceJob),
 
       # User web pages.
       webapp2.Route(r'/browse', BrowseHandler),
+      webapp2.Route(r'/content', ContentHandler),
       webapp2.Route(r'/stats', StatsHandler),
       webapp2.Route(r'/isolate/api/v1/stats/days', StatsGvizDaysHandler),
       webapp2.Route(r'/isolate/api/v1/stats/hours', StatsGvizHoursHandler),
       webapp2.Route(r'/isolate/api/v1/stats/minutes', StatsGvizMinutesHandler),
       webapp2.Route(r'/', RootHandler),
 
       # AppEngine-specific urls:
       webapp2.Route(r'/_ah/mail/<to:.+>', EmailHandler),
       webapp2.Route(r'/_ah/warmup', WarmupHandler),
   ]
 
 
 def create_application(debug):
   """Creates the url router.
 
   The basic layouts is as follow:
   - /restricted/.* requires being an instance administrator.
   - /stats/.* has statistics.
   """
   acl.bootstrap()
   template.bootstrap()
   return webapp2.WSGIApplication(get_routes(), debug=debug)