Unusable software

While trying my best to be productive in VS 2012 I get this message:

Thanks allot for hogging the UI thread piece of shit.  I was just being productive with it!

In sunny South Africa, we have problems.  The least of which is our poor Internet development.  Entry level DSL speeds was just a few years ago 384 kbps. That is stupid slow. Just a tiny bit faster than typical EDGE speeds.  And to make matters worse, the amount of data you can stream over these small pipes is also limited.  Your typical fixed line caps came in 2, 5 and 10 Gigs per month. That's per MONTH not day!

So imagine my delight when WebAfrica, my ISP (that I will love forever) announced all capped customers get 50 TB!! for use during after hours in September for free. Not just that: if I download more than 50 Gigs I could win a pair of Google Glasses.  WebAfrica: Challenge accepted!

Using some Linux tools I will present here my scripts for downloading videos  from the Internet scheduled to run only in after hours time.  Hopefully this will help some others to have a good shot at entering the competition. (And if you win with my help, I demand the privilege of at least holding the Google glasses. ;)

First a script for checking if it is in fact after hours download time.

#!/usr/bin/python import datetime import sys now = datetime.datetime.now() nowTime = now.time() downloadTimeStart = datetime.time(0,0) downloadTimeEnd = datetime.time(6,00) if nowTime > downloadTimeStart and nowTime < downloadTimeEnd : # In download time, exit with 0 sys.exit(0) else: sys.exit(1)

(Note 0 means success and 1 means fail, as per normal bash return codes.)

Now the script that will be run in the cron job (more on that later.):

#!/bin/bash if /home/drbergie/scripts/isAfterHoursDownloadTime.py; then echo "starting download" DOWNLOAD_PROGRESS_FILE=/tmp/downloadInProgress LINKS_FILE=/home/drbergie/youtubelinks.txt if [ ! -f $DOWNLOAD_PROGRESS_FILE ] ; then echo "1" > $DOWNLOAD_PROGRESS_FILE cd /tmp/ if [ ! -d yt ] ; then mkdir yt fi cd yt /usr/local/bin/youtube-dl --all-formats -a $LINKS_FILE rm * cd .. rm $DOWNLOAD_PROGRESS_FILE else echo "Skipping download, already in progress" fi fi

This relies on youtube-dl a nice little python app that makes downloading videos easy.  I installed it by running sudo pip install --upgrade youtube_dl

Notice the bash script doesn't run if cron job is called again while it is still busy. It uses a file called youtubelinks.txt that contains links to your desired youtube video doanloads.  It downloads videos in all the available formats, and then deletes them when finished.

Now tie it all together by adding your these lines to your cron script (to edit cron file run crontab -e and add below lines to file):

#Download comp scripts */1 * * * * /home/drbergie/scripts/downloadComp.sh 0 6 * * * /usr/bin/killall youtube-dl

This runs the download script once a minute (that is what */1 means). And kills any in progress downloads at 6 am every day using killall.

Obviously paths in these scripts are my username so change/home/drbergie to your relevant paths.  (Cron doesn't like PATH environment variables)

After much more work than I though it would be, (gasp, a schedule that was overrun!) we are proud to announce Toddler Playbook, the Android app.  

The app is simply a list of activities with clear illustrations that a bored/tired mom or dad can use to keep a toddler or small child occupied. The activities are simple and typically use items found around the house.

The app is written in Java for Android, compatible all the way back to Android 2.2 (released May 2010! That is the same age as Iron man 2.)  Thirdparty libs include ActionBarSherlock and ORMLite.

The original source content for the activities came to me in the form of a word document (what else?).  This had to be translated to HTML so that the Android app can show it using a simple WebView. 

Googleing around for a nice way of converting doc files to HTML I saw one guy suggesting importing it to Google Drive, and exporting it as HTML. 

Great I thought, let the Google machine do the heavy lifting for me.

"Not so fast kid" says the Google machine.

As you might be aware: HTML does not equal XML.  So step one: convert HTML to XHTML.  Done yet? Nope.  The XHTML must now be split up one HTML document per activity.  Ok, split the files using xpath.  Ok, done that, now we have valid XML files, but not very pretty HTML files.  So add some XSLT to the mix, to wrap it in play nice please XHTML.  So far so good.  Every thing works.  Almost.

The conversion process from HTML to XHTML escaped the troubling characters like apostrophe (apos), left double quote (ldquo), right double quote (ldquo) as these are valid HTML characters.  After the XSLT step however, everything is back to XML, and this doesn't have a mapping for the left quote and right quote characters.  So it output blank.

A problem as you can imagine. Luckily the Python lxml lib  XSLT library can parse XML or HTML for XSLT transforms.  So it handled that better.

The development of all theses scripts probably took the same amount of time as the actual Android app. 

We had a version of the word document on Google Drive, to stay sane.  After some edits we decided to make a new build with some mistakes in the docs fixed.  Not a problem I thought, just run my uber-super-amazing-script.py and be done with it. Not so simple!!2111@!

When transforming the HTML file part in step 3, the script looks for specific CSS class names to identify section headers in the exported HTML. Unfortunately Google doc generates a random set of about 3 different names depending on the weather. The average global weather. On Jupiter. :(  So my script has to be edited each time for each generation.  (The right solution here would be to first look for known styles or strings, and try and read the CSS class from them.)

One last tale of suffering:  The HTML files are served locally as a resource with the Android app.  Android 2.3 does not like spaces in the URL for local access.  %20 doesn't work either.  In Android 4.x it works fine.  My suspicion is that the server reads the space, but doesn't escape it correctly when trying to access the actual unix filesystem.  So file://sdcard/htmlfiles/my hammer.html  should translate to  /mnt/sdcard/htmlfiles/my\ hammer.html but doesn't do it. :(

Our Android app we have made for Toddlers  (more about that soon) needed Facebook integration! How else would our product spread like a mind virus!?

I started on my merry way, implementing the standard way apps share info on Androd: sending an Intent. All goes well, Gmail, Twitter and Google + works fine with this method but alas Facebook (dun dun duuuuuuuuuuuuuh) doesn't support Intents with images and text. Another example of Facebook's non standard approach to Android.  (On an aside I suppose their argument for not allowing this is a valid one. The amount of machine generated crap on Facebook has gotten out of hand as of late.  I guess we just wanted to be part of the bot to bot conversation.  (Perhaps some actual humans would listen in.)  From what I hear though, sending an image with text is allowed on the iOS version of Facebook app. Not fair.)

Only alternative, the people of the Stack, tells me is Facebook SDK for Android.

"Fuck that" I thought.  Don't what to learn a crappy social network's SDK that can't use simple Intents on my platform. I will outsource it! 

So off to Freelancer I went.  It was a fast and easy process to describe what I wanted done, and I was pleasantly surprised when about 10 freelancers clamoured to bid for my project.  

About a day later I picked a programmer from India (where else?) to help me with the Facebook SDK part of the app.  I chatted with him before I awarded him the project.  "Posting images, with text, and no stupid web log in?"  "Yes yes yes, no problems."

$20 he said, and should be done in 2 days.  Cool I thought, and went away for the weekend like a globalisation loving capitalist.

Before I that, I wanted to give him a skeleton of our app. I wasn't about to hand him over all my source codez just like that (obviously our app is a unique, winning idea) so a subset of the app had to be built.  The app simply, takes photo, posts Intent, with comment saying /* Add Facebook share here. */

First version he sent had a stupid Web log in, but besides that, the basic functionality was there. Posting a photo to my Facebook timeline. 

It wasn't a public post though, only I could see it. I asked our friend from over the pond if that is a setting in the way the Android app builds the post, or the Facebook app's setting? "No" he said, that is a Facebook setting per app that needs to be set.

Hu? Our poor stressed out, and sleep deprived users definitely won't be able to go through the seven click process of changing default share options for a Facebook app.  I don't think they even know they have Facebook apps.

The relationship started breaking down.  I  made the classic manager mistake, of meddling in development.  "Just use the Graph API" I insisted, without completely understanding what that means.  He started loosing patience with me, and that meant his "US English Level 1" score of 85% dropped to about 30%.

At the end of the project, 3 weeks later he mailed me his code, and said "thanks for ur patience and understanding. the thing that u have asked not possible. i added the apk and the code also"    To be fair the project would probably never have taken more than a week, if I had given him quicker feedback, or lead him down the Open Graph API maze, that he never worked with before.  I didn't feel bad paying him the full(!) requested amount of $20.

Now I had to incorporate his (mostly copied and pasted) code into the real app.  Didn't take too long, but I was stuck with posting the image to my wall (timeline? What's the difference?). Facebook's API kept returning "permissions error". Googleing didn't really help (mostly because the Facebook API has changed so much, almost all of what is written online about it is broken). So I turned to our Indian friend again, and he was very helpful, even after monies have exchanged hands, and star ratings have been made. (I gave him 3.4 stars, he gave me 5. Am I a bad person?) Unfortunately, his advice didn't really help.  I was still stuck.

So there I was a few weeks back, reading through the getting started Android Facebook SDK guide exactly what I wanted to pay someone to avoid doing.

Luckily though I could skim over most of it and find the bit that was missing.  (I was never asking for publish_action permission when starting the Android post to Facebook workflow.)

I would say that time wise, I probably came out poorer than I would have just implementing the posting code my self.  This app is a side job though, and learning a new API I would classify as harder cognitive work, than just responding to emails.  After a hard day of real development, it is a bit easier on the coconut.

Also getting to know the process and pitfalls of outsourcing was also a valuable experience.  I would say that next time I dive into an outsourcing project like this, I would probably talk the specs through in more detail with potential freelancers.  Understanding the problem and how best to solve it must be clear for all parties.

My wife had a baby recently, so as you can imagine our lives have changed quite a bit.  Probably forever. 

The part that I am most involved with: sussing to sleep of the baby, has all the components of a good simulator game.  You have limited resources: parent stamina, baby drowsiness, parent sanity. Power ups include: dummy to get kid to sleep deeper (with downside of if it falls out later there is a 20% chance of a reawakening), drugs for baby (works well in short term, but affects baby IQ, and therefore, final game score). 

Giving daddy some beer while baby eats, increases sanity, but makes it more difficult to control the rhythmic walking needed to get baby to sleep. (Drunk meter!)

After the baby is asleep in the players arms, the final move (KO as it were) is to put the baby down in her crib.  The angle has to be right, and so too when you do it with regards to the rest of your rhythm. Get it wrong, and start all over again. (Parent moral -- ) In my experience, the baby is usually sleepier second time round, so should be easier.

I imagine the walking mechanics of the game to be like Guitar hero. Hitting the right spots at the right time increases baby drowsiness more effectively. (Dodging the odd pet or grumpy sibling for bonus points!) Different types of rhythms have different effects depending on the sleepy state of the baby.

Another mini game: Go to the shops, and try to avoid all the cooing old ladies, that want to touch your angel with their probably unwashed hands.

The Nightmare difficulty level (Disclaimer will read: "Only if you are a real parent, should you attempt this difficulty level. It is not even remotely fair for the player") none of the indicators will be visible.  New variables for baby unhappiness will be in play: temperature, random cramps, being held too long.

Thinking about educational games lately, I started to realise that most games that stick in my memory have had some educational element to them. Mostly as a by product of me just stubbornly trying to beat the game. I had no choice but to learn some new tricks, in the new world it presented me with.

Truism: Having fun makes people more susceptible to learning. 

Here are some of the early games I enjoyed, and the impressions they left on me.

(I still can't quite believe that you can play most of these games in your browser today! Blows. My. Mind.)

Rampage taught me eating bombs is bad, and eating people can increase HP. Yum.

Maniac mansion taught me breaking into a mansion was easier than it sounds.  Geeks sometimes wins even in the real world, and nuclear reactors need water to cool down. Conversations end up with very different outcomes depending on how I answer. Also tentacle mating calls ends up being fatal. Play NES version here.

Early King's quest games made my learn basic English grammar and spelling. Increased my medieval related vocabulary. Play here

Dune II resource gathering and spending it wisely. (Something which I am sure influenced my attitude towards my finances. "I want a PS4." "YOU HAVE INSUFFICIENT VESPENE GAS!") Play here, requires free sign on

Blockout probably first 3D game I played, or available on a 80286, made me think about minimising rotations of a 3D object to get it to my desired orientation. Flash port

Sopwith was the first game where I started to get an intuitive feel for physics, at higher than day to day speeds for a 10 year old.

Obviously more modern games had their effects as well (HL1: what is an enemy. Those poor charging up dog things, I tried to just avoid them. GTA 3: City navigation.). 

Windows 8 makes me wait, then shows me ants.

Am I really the only one that worries where the dots go after the first rotation of the new Windows 8 wait spinner? Will they ever come back? (It’s probably just some developer animation timing fuckup that some UI designer thought looked cool)

Samsung has done really well with their Android devices.  The Galaxy, and then the Galaxy S lines are top notch phones.  My only problem is the crapletts that they apply on top of normal, generally clean Android. 

Since the SII was released in early 2011 lots of people have agreed, that Samsung's bloatware can get too much.

The latest gimmick that Samsung is trying to pull on us is Smart stay.  I must admit, it sounds like a neat idea that could be useful.  Using the front facing camera...

the screen maintains a bright display as long as you're looking at it.

Sounds good right?  Screen dims only when you are not reading the screen. GENIUS! No more jiggling the scroll position, to keep the screen on.

Hold on, not so fast girls!  The fatal, and brain dead flaw in their plan is that there is this annoying eye, looking at you, in the notification bar all the time.   All the freaking time!  No clean notification free bar for you, ever again! OCD rage! 

The best part is that the moment you look away, a different, animated scroll up notification flashes on the notification bar. Since the human eye's peripheral vision has been trained by millions of years of not being eaten by pterodactyls to detect motion, I reflexively look at the tablet screen.  Ridiculously distracting, like the tablet is some attention seeking toddler. I'm sure the pixies controlling the tablet has a good laugh at me just around then. 

The basics of a UI event framework is quite simple to understand.  There is an event queue that reads user inputs, clicks, types mouse moves, (these days swipes) etc., and the code posts the events to a UI element, a button say, and tells it "Hey, button1, something happened!" The button then can decide if it wants to do something with the event or not.

As a programming idea this works well. I only have one small tweak, that will make a big difference to all UX on all platforms:  Before a UI element is allowed to receive an event, it must be of a certain age.  Hundred milliseconds, say.  This is simply because a UI typically, is used by a slow human brain, that can't read the dialog box, that has only finished drawing on the screen 5ms ago. So why accept input from the user if you are rather sure she doesn't know what she is doing.

Firefox has recognised that this is a potential security problem, and have built in a delay on all plug-in install confirmation buttons.  Before that safeguard, one unlucky double click on the wrong spot, could lead to an infected PC.

Most recently the keyboard version of this has happened to me in FTL.  Pressing the 1 and 2 keys targets weapon 1 or 2.  Only problem is, that 1 and 2 is also the game choice dialog popup's choice short cut keys.  So when I thought I would launch the final missile by pressing 2 to destroy the pirate spaceship, what I have actually done is graciously accept their offer of truce, and going our separate ways... No scrap for me.

A thirparty software company we work with regularly has stopped supporting email as a means of support communication.  When you send them an email it auto replies, telling you to please log the bug in their *cough* Jira *cough* issue tracking system.  Not that there is anything specifically wrong with Jira. It is designed with usability in mind. Unfortunately though, for most installations I have seen, it is usually customised (read butchered), to fit some IT drone's idea of being useful. This usually results in 15 drop downs for each issue, magic tick boxes, and an indecipherable colour scheme for each listed issue.

I don't think it is a good idea for a company, to ignore emails, the usual route for most support queries that has worked for years, and ask their users to learn a new interface that isn't email. It all feels to much like they are trying to crowd source their support mail sifting. 

There are good products out there that integrates with email and, gasp, logs things in an issue tracker automatically.  (Spoiler alert: one of them is Jira.)

Windows XP, bless her, never had Start menu search. Such simple times, we never thought to ask. We didn’t know what we were missing. That is, until Suse Linux’s incarnation of KDE 3.x added the search bar for all your items in your Start menu.

Such a nice idea: type to search for your installed apps. And it worked! No longer did we have to start up our entire brain to try and figure out what arbitrary category a application will be in. Just try and remember an approximate spelling and you will find what you need.

Keep it simple stupid.

Developers like complexity, which isn't a problem in itself.   We need complex models and code to solve complex problems.

The beef I have is the tendency to increase the complexity of code or a design for no real benefit.

What happens when that code is signed off, and all that complexity is swapped out of the dev's brain and 6 months later a sneaky bug crops up? If that much time has elapsed, chances are its gonna take a while to reconstruct the mental model needed to fix that bug.

I say code should be simpler, and more readable, as one of the top priorities when writing software.  Simpler code makes it easier for new people to understand your design, and bugs should hopefully be simpler to find. 

I was always fascinated with machine generated sentences.  Part of the allure is that it is a gamble if it comes out coherent at all or not. Its the infinite monkeys infinite type writer kinda thing.

A VCS commit comment is a very specific kind of sentence.  Usually it is a short description of work done.  It seems to be just as likely a nonsensical blurt about something that only makes sense in a developer's head at that exact moment.  So to read them out of context is sometimes quite amusing.

And so, commitconversations.com is born!

I decided to build a small webapp that presents the user with a collection of randomly chosen commit comments.  Some hilarity will probably ensue.

Using the GitHub API the app does a search on GitHub for repositories.  It then copies the latests few commit comments for a few of these repos into a local database.  The rest is simple java script.  I built the back-end with Google App engine in Python. (Python is a bit of a quirky language but if you keep the problem small enough, it stays simple to read.)