add download progress polling thread

3 files changed, 79 insertions, 12 deletions

diff --git a/Assistant.hs b/Assistant.hs
index 413e5e90e..42175dd82 100644
--- a/Assistant.hs
+++ b/Assistant.hs
@@ -39,30 +39,32 @@
  - 	and maintains the DaemonStatus currentTransfers map.
  - 	(This uses inotify on .git/annex/transfer/, so there are
  - 	additional inotify threads associated with it, too.)
- - Thread 10: Transferrer
+ - Thread 10: TransferPoller
+ -	Polls to determine how much of each ongoing transfer is complete.
+ - Thread 11: Transferrer
  - 	Waits for Transfers to be queued and does them.
- - Thread 11: StatusLogger
+ - Thread 12: StatusLogger
  - 	Wakes up periodically and records the daemon's status to disk.
- - Thread 12: SanityChecker
+ - Thread 13: SanityChecker
  - 	Wakes up periodically (rarely) and does sanity checks.
- - Thread 13: MountWatcher
+ - Thread 14: MountWatcher
  - 	Either uses dbus to watch for drive mount events, or, when
  - 	there's no dbus, polls to find newly mounted filesystems.
  - 	Once a filesystem that contains a remote is mounted, updates
  - 	state about that remote, pulls from it, and queues a push to it,
  - 	as well as an update, and queues it onto the
  - 	ConnectedRemoteChan
- - Thread 13: NetWatcher
+ - Thread 15: NetWatcher
  - 	Deals with network connection interruptions, which would cause
  - 	transfers to fail, and can be recovered from by waiting for a
  - 	network connection, and syncing with all network remotes.
  - 	Uses dbus to watch for network connections, or when dbus
  - 	cannot be used, assumes there's been one every 30 minutes.
- - Thread 15: TransferScanner
+ - Thread 16: TransferScanner
  - 	Does potentially expensive checks to find data that needs to be
  - 	transferred from or to remotes, and queues Transfers.
  - 	Uses the ScanRemotes map.
- - Thread 16: WebApp
+ - Thread 17: WebApp
  - 	Spawns more threads as necessary to handle clients.
  - 	Displays the DaemonStatus.
  -
@@ -118,6 +120,7 @@ import Assistant.Threads.SanityChecker
 import Assistant.Threads.MountWatcher
 import Assistant.Threads.NetWatcher
 import Assistant.Threads.TransferScanner
+import Assistant.Threads.TransferPoller
 #ifdef WITH_WEBAPP
 import Assistant.Threads.WebApp
 #else
@@ -168,6 +171,7 @@ startAssistant assistant daemonize webappwaiter = do
 				, assist $ pushRetryThread st dstatus pushmap
 				, assist $ mergeThread st
 				, assist $ transferWatcherThread st dstatus
+				, assist $ transferPollerThread st dstatus
 				, assist $ transfererThread st dstatus transferqueue transferslots
 				, assist $ daemonStatusThread st dstatus
 				, assist $ sanityCheckerThread st dstatus transferqueue changechan
diff --git a/Assistant/Threads/TransferPoller.hs b/Assistant/Threads/TransferPoller.hs
new file mode 100644
index 000000000..d720bcc45
--- /dev/null
+++ b/Assistant/Threads/TransferPoller.hs
@@ -0,0 +1,49 @@
+{- git-annex assistant transfer polling thread
+ -
+ - Copyright 2012 Joey Hess <joey@kitenet.net>
+ -
+ - Licensed under the GNU GPL version 3 or higher.
+ -}
+
+module Assistant.Threads.TransferPoller where
+
+import Assistant.Common
+import Assistant.ThreadedMonad
+import Assistant.DaemonStatus
+import Logs.Transfer
+import Utility.NotificationBroadcaster
+
+import Control.Concurrent
+import qualified Data.Map as M
+
+thisThread :: ThreadName
+thisThread = "TransferPoller"
+
+{- This thread polls the status of ongoing transfers, determining how much
+ - of each transfer is complete. -}
+transferPollerThread :: ThreadState -> DaemonStatusHandle -> IO ()
+transferPollerThread st dstatus = do
+	g <- runThreadState st $ fromRepo id
+	tn <- newNotificationHandle =<<
+		transferNotifier <$> getDaemonStatus dstatus
+	forever $ do
+		threadDelay 500000 -- 0.5 seconds
+		ts <- currentTransfers <$> getDaemonStatus dstatus
+		if M.null ts
+			then waitNotification tn -- block until transfers running
+			else mapM_ (poll g) $ M.toList ts
+	where
+		poll g (t, info)
+			{- Downloads are polled by checking the size of the
+			 - temp file being used for the transfer. -}
+			| transferDirection t == Download = do
+				let f = gitAnnexTmpLocation (transferKey t) g
+				sz <- catchMaybeIO $
+					fromIntegral . fileSize
+						<$> getFileStatus f
+				when (bytesComplete info /= sz && isJust sz) $ do
+					putStrLn $ "download size " ++ show sz
+					updateTransferInfo dstatus t info
+						{ bytesComplete = sz }
+			{- can't poll uploads -}
+			| otherwise = noop
diff --git a/doc/design/assistant/progressbars.mdwn b/doc/design/assistant/progressbars.mdwn
index ee7384274..33d736afd 100644
--- a/doc/design/assistant/progressbars.mdwn
+++ b/doc/design/assistant/progressbars.mdwn
@@ -5,10 +5,24 @@ for it, in the terminal.
 Something better is needed for the [[webapp]]. There needs to be a
 way for the web app to know what the current progress is of all transfers.
 
-To get this info for downloads, git-annex can watch the file as it arrives
-and use its size.
+This is one of those potentially hidden but time consuming problems.
 
-TODO: What about uploads? Will i have to parse rsync's progresss output?
-Feed it via a named pipe? Ugh. Check into librsync.
+## downloads
 
-This is one of those potentially hidden but time consuming problems.
+* Watch temp file as it's coming in and use its size.
+  This is the only option for some special remotes (ie, non-rsync).
+  Can either poll every .5 seconds or so to check file size, or
+  could use inotify. Implemented.
+* Feed rsync output into a parser and parse out a progress value. Ugly,
+  failure prone, but potentially the least CPU-expensive option.
+* Use librsync. Note: It's not wire-compatiable with the actual rsync
+  command.
+* Set up a FIFO, have rsync read from or write to that, and the FIFO
+  feeder/reader then can update the transfer info. Generic enough to
+  work for most (all?) special remotes, but also the most expensive option,
+  involving another copy through memory of the whole file contents.
+
+## uploads
+
+Cannot use temp file, as we're not receiving it. Rsync progress parser,
+librsync, and FIFO all work.