cmd/web: stream large files asynchronously.

We had a nice chunkyreader() loop for writing files, but unfortunately,
Tornado captured the full content of those files before writing them to the
client.  Oops.

Change things around so we don't end up buffering some multiple of the
ENTIRE FILE in memory.

Signed-off-by: Avery Pennarun <apenwarr@gmail.com>

diff --git a/cmd/web-cmd.py b/cmd/web-cmd.py
index 4b988140338b40758be97b7dd9857c0abb6274c7..568811c29a259b276e697a699ebb99877e897d06 100755 (executable)
--- a/cmd/web-cmd.py
+++ b/cmd/web-cmd.py
@@ -68,7 +68,8 @@ class BupRequestHandler(tornado.web.RequestHandler):
 
     def head(self, path):
         return self._process_request(path)
-
+    
+    @tornado.web.asynchronous
     def _process_request(self, path):
         path = urllib.unquote(path)
         print 'Handling request for %s' % path
@@ -118,12 +119,23 @@ class BupRequestHandler(tornado.web.RequestHandler):
         self.set_header("Content-Type", ctype)
         size = n.size()
         self.set_header("Content-Length", str(size))
+        assert(len(n.hash) == 20)
+        self.set_header("Etag", n.hash.encode('hex'))
 
         if self.request.method != 'HEAD':
+            self.flush()
             f = n.open()
-            for blob in chunkyreader(f):
-                self.write(blob)
-            f.close()
+            it = chunkyreader(f)
+            def write_more(me):
+                try:
+                    blob = it.next()
+                except StopIteration:
+                    f.close()
+                    self.finish()
+                    return
+                self.request.connection.stream.write(blob,
+                                                     callback=lambda: me(me))
+            write_more(write_more)
 
     def _guess_type(self, path):
         """Guess the type of a file.