Tech Blog

I have been using git for a while now, and I believe I have the the basic workflow under control. Committing, reverting, using local branches for major refactoring work: been there, done that! However, I recently got some collaborators on my github-projects, I had to start working with other remote repositories and branches.

I found this blog post, which was really helpful. I am sharing some others things that helped me in the last couple of weeks. Hopefully, this saves other people some time Googling. If you know a better ways to accomplish these tasks, please let me know!

Things to remember about remote branches

Because I had some troubles discovering how to properly work with remote repositories, I am sharing what I found. The most important things to remember:

• Your local branches are not really connected to the remote branches. It is therefore possible to mix up branches, but this is usually not what you want
• Never forget to switch to the correct local branch (using git checkout local-branch.
• The names of the local and remote do not need to match.
• Before you can push changes from a local branch to a remote branch, all the commits of the remote branch have to be included in your local branch. This can be done using by executing git pull remote-name remote-branch in your local branch.

Checking out a newly created branch in a remote repository

Bart is implementing Merb log parsing for request-log-analyzer. He has put his progress in a separate branch of the github project. My local repository does not yet include this branch, but I want to check it out. Note that I am using a different name than the branch name on the github project.

$ git branch merb origin/generic_base
$ git checkout merb

When the new functionality is finished, the following commands will merge the changes in the merb branch to the master branch.

  # goto my local master branch and merge the merb-branch
$ git checkout master
$ git merge merb
 
  # push the changes to the master branch on github
$ git push origin master

Merging back a fork

Wes Hays is helping me out on the scoped_search plugin. He implemented OR in the query language in his own fork. I wanted to merge his changes back to master branch:

  # add a reference to the remote repository
$ git remote add gbdev git://github.com/gbdev/scoped_search.git
 
  # create a local branch for the fork to follow a remote branch
$ git branch gbdev-fork gbdev/master
$ git checkout gbdev-fork

Now, my local gbdev-fork branch contains Wes’s code. Because Wes’s repository was forked from my repository, git will know that most of the history of my master branch and gbdev-fork branch is the same.

After some testing, I was ready to include his changes by merging the gbdev-fork branch into my local master branch:

  # go back to my master branch, and merge the changes
$ git checkout master
$ git merge gbdev-fork
  # push the changes to the master branch at hithub
$ git push origin master

Pushing tags to a remote repository

You can create tags locally, but you probably want to send them to the remot repository as well:

  # create a local tag "tagname" with the given message.
$ git tag -a "tagname" -m "message" 
  # send your tags to the remote repository "origin"
$ git push origin --tags

Over the last year we’ve tried a lot of different applications to keep track of bugs and other issues, but none of those apps sticked with us. They were either not suited for the job (Basecamp) or to difficult and ugly (Trac, Bugzilla). But now I think that we finally have found a solution that just gets the job done: Lighthouse by activereload.

Lighthouse’s tagline is “beautiful simple issue tracking” and that really sums it up perfectly. It tracks issues, not only a bugs, so you can use it for all the issues that need to be dealt with. It’s simple and it’s easy to use, it just does what it needs to do without loads of unneeded features. A great advantage for us is that our designers can (and want) to use it too. It’s also very nicely priced and it has a sweet API for us geeks to build your own scoreboard (see picture).

There are some minor things that can be improved (for example, it’s somethings a little confusing to work with multiple projects) but it’s by far the best issue tracker I’ve ever worked with. To everybody at activereload, we really love your product and keep up the good work!

I personally am a fan of the git version control system. The best part of git is its speed, and the simplicity of using local branches.

Local branches are very helpful if you are working on different features at the same time but want to keep them apart. An example: it happens all the time that I am working on some feature and than I have to put my current work aside to work on a high priority issue. Once this issue is solved, I need to commit the changes and usually do a deploy of the web application so that the problem is solved as soon as possible. With Subversion, I sometimes commit files that were part of the unfinished feature I was working on before I started on the high priority issue. If I am not careful and deploy those files, unfinished work will be put into production and this can go horribly wrong, like every page request returning a 500-error of our high traffic site .

Using git, I can put my current work aside easily by using git stash. When I am finished with the high priority issue, I can revert to my previous work with git stash apply.

Another option: branching the project (using git branch feature) if the feature I am working on is invasive and than switch branches for high priority issues using git checkout master. I can go back to the feature branch with git checkout feature, followed by git merge master to merge back the changes I just made in the master branch. Branching and merging is very fast in git and merging is not the PITA like it is in Subversion.

However, our main code repository will probably remain in SVN for now. Luckily for me, I can use git-svn locally to profit from these advantages. I found an informative page on installing and getting started with git-svn on OSX. If you know Subversion, this page is helpful to translate Subversion commandos to their git alternatives.

I wanted to get some data from a web service using Javascript. I looked at several Javascript classes (like this), but because the web service was running on another server it got a little troublesome. As a solution I tried to call the web service through a proxy, but that didn’t make it any easier.

Jaap suggested to take a look at NuSOAP, a -kinda old- SOAP toolkit for PHP. With an AJAX request I could call a PHP page that uses NuSOAP to consume the web service. It was actually easier then I thought it would be.

To make the AJAX call from Javascript I used Prototype and this script:

  function doRequest() {
    var url = "ajax/consume_webservice.php";
    var param1 = "value1";
    var param2 = "value2";
    var params = "param1="+ param1 +"&param2="+ param2;
 
    new Ajax.Request ( url, { method: 'POST', parameters: params,
      onComplete: onResult } );
  }
 
  function onResult( result ) {
    alert( result.responseText );
  }

The PHP file consume.php looks something like this:

< ?php
 
  $param1 = isset( $POST_['param1'] ) ? $POST_['param1'] : false;
  $param2 = isset( $POST_['param2'] ) ? $POST_['param2'] : false;
 
  // this is the only file I used from the NuSOAP project
  require_once( "nusoap.php" );
 
  $url = webserviceurl;
  $params = array( "param1" => $param1, "param2" => $param2 );
 
  $soap = new nusoap_client( $url, true, false, false, false, false, 0, 60 );
  $proxy = $soap->getProxy();
  $proxy->functionname( $params );
 
  echo $proxy->response;
 
?>

That’s all. Do a AJAX request from Javascript to a PHP page. Then the PHP page uses NuSOAP to consume the web service and returns the result. Back in Javascript you can do whatever you want with the given data.

In college I followed a course in numerical analysis. The main point of the course was to be careful with floating point arithmetic, because it is vulnerable to rounding errors that can significantly influence the result of complex computations. Until yesterday I never had encountered such a problem. Now that I have lost my innocence in this matter, I would like to share my tale of nasty debugging and frustration.

After receiving some bug reports of Floorplanner designs that failed to save properly, we dove into the code to see what was going wrong. After some time, we found that the errors were caused by the script that loads the design after it has been saved with a unique name. This unique name is passed to the script to be able to find the design. We used the current timestamp as a unique name for the design. The current timestamp simply is the number of seconds passed since January 1, 1970 and looks something like this: 1206712028. As a design name, this number was passed to different scripts, both client-side and server-side. However, at some point in this chain of scripts, the number was changed slightly to 1206712030 and because of this the associated design could not be found, resulting in an error.

At first, we investigated the possibility that the stored timestamp was overwritten by a newer timestamp, as this could explain the slight increase in the number. However, we were not able to find this anywhere in the code and sometimes, the number was decreased a bit instead of being increased.

Finally, we monitored the data being sent between the different scripts, and we found that ActionScript automatically converted the numeric design name into a number in scientific notation. In our case, this would be 0.1206712028 x 10^10. Unfortunately, this number was rounded to 0.120671203 x 10^10 because computers use floating point arithmetic to store numbers in scientific notation. This number would eventually be converted back to normal notation, but it was now 1206712030 because of the rounding error.

We fixed it by putting an ‘a’ in front of the timestamp, preventing the automatic conversion to a number. Not very elegant, but it works!

Today we changed our branching style of the Floorplanner code base. I had it in my head for a long time, but I actually changed it until I read this post.

In the old setup we used the style that is called Branch per Release.

Branch per Release
Every release is a new branch; common changes are merged between the releases. Branches are killed off only when the releases are no longer supported.

In the beginning we had a lot of custom projects running on one Floorplanner code base. We wanted to improve the Floorplanner itself but changes in the source could cripple the custom projects. That’s why we created a branch for every custom project. This way we had a stable codebase for every custom project and for the Floorplanner itself. When we fixed a global bug we merged the fix to all the branches.

Nowadays we don’t have any custom projects running on the same code base anymore. We do have one branch (not the trunk) in which we were working on the latest release. So today I decided that this had to change. We merged the branch to the trunk and new features (tasks) will be developed in a branch of their own. When the feature is done it will be merged back to the trunk.

This is called the Branch per Task.

Branch per Task
Every development task is a new, independent branch. Tasks are merged into the permanent main branch as they are completed.

If you want to know more about branching, read Jeff’s post Software Branching and Parallel Universes.

Jeff Atwood recently wrote a very interesting post about this topic. He states that:

Check out the whole story here: Rethinking Design Patterns