Sunday, October 30, 2011

WifiAlive Android App released

WifiAlive is a very simple and low resource application aimed at maintenance of a Wifi connection. A connection is periodically (every 5 seconds) checked by trying to contact the gateway, as well as IP addresses or domain names which you can enter via the UI. WifiAlive tests each of these addresses by icmp and http head. If one of these tests succeeds, Wifi is considered up. If not, Wifi is reset (that is deactivated and activated). After Wifi (re-)activation Android on its own tries to establish a connection to an access point in reach. On a failed connection check, WifiAlive will only wait 20 seconds for a reset, scan and successful connection until the connection is reset again. Considering the time needed for a Wifi reset and a scan to complete, it can be said, that, as soon as a known network is available, there is a worst-case of roughly 30 seconds for the connection to be established. WifiAlive is thus much more simple then many of its competitors. The popular WifiFixer for example conveys about 6000 lines of code whereas WifiAlive has only around 500.

The code of WifiFixer is publicly available on github. I have checked those parts of the code which seemed important in respect of a problem I believed to have with WifiFixer 0.8.0.6. I was rather stunned by the complexity of the program, since I had in mind the solution I have just proposed. WifiFixer almost replaces the Wifi Settings Tab in Android and will continue this approach in upcoming versions. In some cases, settings from the Android Wifi Settings Tab can be overridden, it does its own management of known networks and behind the scenes does scans for networks in reach, matches them to known networks and harbours a small collection of hacks for devices that have to be dealt with in a special way - for example the Google Nexus One. WifiFixer is a great application. It helps many people with their Wifi troubles. But after I had seen its complexity and, in addition, found a bug in its matching of known to found networks on a scan, I decided that this was not the solution to my problem, as, additionally, a lot of code seemed to do things that Android normally does on its own. No offence David, I've learned a lot from looking at your code.

There are other applications, like Advanced Wifi Lock. But I couldn't get access to their code and very often the free (limited) versions did not seem to cover all aspects of Wifi maintenance that I wanted to ensure. Moreover I am already beyond trusting a random application from the market without knowing its code. The Advanced Wifi Lock application (a bare Wifi lock, if it does what the name suggests) does not solve my problems, as it sometimes takes Android a bit too much of time to re-establish a connection when re-entering the area of an access point or on walking from one access point to another. In my experience this could take up to two minutes, which was just too long for my purpose. These experiences may be better on other devices, still it feels very good to know, that as soon as a known network is in reach, there is a rough 30 second worst-case until you have a connection.

On some devices there also seem to be problems with the supplied Wifi drivers. These drivers may enter an undetermined state or cause wpa supplicant to run in endless loops. I have not delved into these problems. But it seemed to me, that, if at all, they can be resolved by doing a Wifi reset.

To sum up, I think that some applications aimed at taking care of your Wifi connection may be too complicated and question is, why such an application should do things like scanning for access points, as Android usually does a pretty good at this. Others may be doing the things I had in mind, yet there was no way for me to be sure.

That's, why I wrote WifiAlive. I hope you like it.

Feel Free to browse the code at http://darcs.factisresearch.com/pub/WifiAlive or just install the app from the android market.

Sunday, October 9, 2011

stm-stats: Retry statistics for STM transaction

Two days ago, Stefan Wehr talked at the Haskell in Leipzig workshop about the experiences that factis research had with Haskell when creating real-world applications. When he mentioned that we have code that keeps count of retries of STM transaction, someone from the audience asked us to publish the code. So, after some refactoring to make the interface generally usable, here it is. The stm-stats package provides a few functions (in Control.Concurrent.STM.Stats) that can replace atomically: trackSTM and variants that allow you to name the transaction or have more control about the warnings that will be emitted when the number of retries exceeds a certain value. The following code demonstrates how the module works:
import Control.Concurrent
import Control.Concurrent.STM
import Control.Monad

import Control.Concurrent.STM.Stats

main = do
var <- trackSTM $ newTVar 0
forkIO $ forM_ [1..23] $ \i -> do
    threadDelay (100*1000)
    trackNamedSTM "writer" $ writeTVar var i
putStrLn "Starting reader..."
trackNamedSTM "reader" $ do
    i <- readTVar var
    when (i < 23) retry
putStrLn "Reader finished."
dumpSTMStats
When run, you will see this output:
Starting reader...
STM transaction reader finished after 23 retries
Reader finished.
STM transaction statistics (2011-10-09 16:26:27.226675 UTC):
Transaction     Commits    Retries      Ratio
_anonymous_           1          0       0.00
reader                1         23      23.00
writer               23          0       0.00
PS: As I usually post on my own blog, I should explain why I from now on will also post on the factis research company blog. Factis research relies heavily on Free Software and wants to contribute back to the community where possible. But in the course of every-day work, this sometimes falls by the wayside. Therefore, I was hired to help out as the community interface: My task is identifying, packaging and uploading components of internal code that are of wider interest, following up on user requests and bug reports, and talk about it. This module is the first brought to you by this new strategy, but expect more to come.

Saturday, October 8, 2011

Talk about developing commercial software in Haskell

Yesterday, I gave a talk at the Haskell in Leizpig workshop about our experience in developing commercial software in Haskell. You can find the slides (in german) here. I was really surprised by the large number of attendees (about 50), given that the workshop's language was german (except the very interesting talk by Kevin Hammond).

Saturday, October 1, 2011

New Version of HTF with Diffs, Colors, and Pretty-printing

I've just uploaded version 0.8.1.0 of HTF to hackage. HTF (Haskell Test Framework) allows you to define unit tests, QuickCheck properties, and black box tests in an easy and convenient way. We use HTF at work all the time, where it has proven to be quite valuable for organizing our test suite. An earlier blog post describes HTF in more detail.

The new version comes with some cool new features:

  • Support for diffs and pretty-printing. If an equality assertions fails, you now get a proper diff of the two values involved. If possible, the values are pretty-printed (thanks to Edward Yang and his groom package for inspiration). Here's an example:
    [TEST] Main:diff (TestHTF.hs:68)
    assertEqual failed at TestHTF.hs:68
    * expected:
    PlusExpr
      (PlusExpr
         (MultExpr
            (PlusExpr (Variable "foo")
               (MultExpr (Literal 42) (Variable "bar")))
            (PlusExpr (Literal 1) (Literal 2)))
         (Literal 581))
      (Variable "egg")
    * but got:
    PlusExpr
      (PlusExpr
         (MultExpr
            (PlusExpr (Variable "foo")
               (MultExpr (Literal 42) (Variable "bar")))
            (PlusExpr (Literal 2) (Literal 2)))
         (Literal 581))
      (Variable "egg")
    * diff:
    6c6
    <         (PlusExpr (Literal 1) (Literal 2)))
    ---
    >         (PlusExpr (Literal 2) (Literal 2)))
    *** Failed!
        
  • As you can see from the example above, HTF now supports colored output.
  • There's a new commandline option --quiet which causes HTF to produce output only if absolutely necessary (e.g. for failed test cases).

Just get HTF from hackage, it now also works with GHC 7.0.* and 7.2.1!