Add helpful error message to isolate UI when downloading big file

appengine/isolate/handlers_frontend.py

 # Copyright 2012 The LUCI Authors. All rights reserved.
 # Use of this source code is governed under the Apache License, Version 2.0
 # that can be found in the LICENSE file.
 
 """This module defines Isolate Server frontend url handlers."""
 
 import datetime
 import json
+import logging
 import re
 
 import webapp2
 
 from google.appengine.api import memcache
+from google.appengine.api import modules
 from google.appengine.api import users
 
 import acl
 import config
 import gcs
 import handlers_endpoints_v1
 import mapreduce_jobs
 import model
 import stats
 import template
@@ skipping 198 matching lines @@
       if not raw_data:
         stream = gcs.read_file(config.settings().gs_bucket, entity.key.id())
       else:
         stream = [raw_data]
       content = ''.join(model.expand_content(namespace, stream))
 
       self.response.headers['X-Frame-Options'] = 'SAMEORIGIN'
       # We delete Content-Type before storing to it to avoid having two (yes,
       # two) Content-Type headers.
       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
-          # 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=%s' % namespace +
-              r'&digest=\1">\1</a>',
-            content)
-          self.response.headers['Content-Type'] = 'text/html; charset=utf-8'
-        except ValueError:
-          pass
+
+      # App Engine puts a limit of 33554432 bytes on a request, which includes
+      # headers. Headers are ~150 bytes.  If the content + headers might
+      # exceed that limit, we give the user an option to workround getting
+      # their file.
+      if len(content) > 33554000:
+        host = modules.get_hostname(module='default', version='default')
+        # host is something like default.default.myisolateserver.appspot.com
+        host = host.replace('default.default.','')
+        sizeInMib = len(content) / (1024.0 * 1024.0)
+        content = ('Sorry, your file is %1.1f MiB big, which exceeds the 32 MiB'
+        ' App Engine limit.\nTo work around this, run the following command:\n'
+        '    python isolateserver.py download -I %s --namespace %s -f %s %s'
+        % (sizeInMib, host, namespace, digest, digest))
+      else:
+        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
+            # 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=%s' % namespace +
+                r'&digest=\1">\1</a>',
+              content)
+            self.response.headers['Content-Type'] = 'text/html; charset=utf-8'
+          except ValueError:
+            pass
 
     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.
@@ skipping 117 matching lines @@
 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)