Using multiple git configurations

It is fairly common that you might be contributing to multiple git repositories which make different git configuration necessary.

Mine and probably the most common case is that I am using different contributor email addresses for my private and work git repositories. You might also want want to have separate global git ignores or other options that differ in multiple repositories.

The way I do it is having one ~/.gitconfig which looks like this:

[user] 
  name = Christof Damian 
  email = christof@damian.net 

[core] 
  excludesfile = ~/.gitignore_global 

[includeIf "gitdir:~/devex/"] 
  path = ~/.gitconfig_devex 

This defines the defaults for my name, email and a global gitignore. It also tell git to look at another git config for all repositories that are below the ~/devex/ folder. This file is very short for me, as it just changes the email address (~/.gitconfig_devex):

[user] 
  email = christof.damian@devex.com 

The ~/.gitignore_global is just a normal gitignore file, which excludes all the files the editors or tools I use create. It could for example look like this:

*~ 
/.project 
# whatever files your local editor/tools config generates

You could use for example https://www.gitignore.io/ to generate this

Friday Links

The Three Flaws in Software Design & How to Avoid Them
http://www.youtube.com/playlist?list=PLOU2XLYxmsIJ7HGm2bv20QrtwcWemSRCI

Github: Updated Services UI
https://github.com/blog/1827-updated-services-ui

Atlassian: Now in JIRA Agile Labs: an all new control chart
http://blogs.atlassian.com/2014/04/now-jira-agile-labs-new-control-chart/

Atlassian: 6 key ways to optimize development with a control chart
http://blogs.atlassian.com/2014/04/6-key-ways-optimize-development-control-chart/

How Disqus Went Realtime with 165K Messages Per Second and Less than .2 Seconds Latency
http://highscalability.com/blog/2014/4/28/how-disqus-went-realtime-with-165k-messages-per-second-and-l.html

3 ways to do eager loading (preloading) in Rails 3 & 4
http://blog.arkency.com/2013/12/rails4-preloading/

Drowning in problems:
http://game.notch.net/drowning/

There were also some rants about TDD, which I am not going to bother to list 

WeLovePHP Talk: Methodologies and tools used by the Splendia development team

Today I gave a talk at WeLovePHP, which is a quarterly talk and workshop series organized by Softonic in Barcelona. They also do one about JavaScript.
I can highly recommend them if you are interested in PHP, JavaScript or related topics.

I talked about the processes and tools we are using at Splendia.
There are no pictures in the slides ... sorry.

A Code Review and Continuous Integration Workflow

As hopefully most people working on software projects nowadays we are writing unit tests and do code reviews.

Work-flow at Splendia

As our project is a large PHP site we are using PHPUnit for unit testing and various static code checking tools (checkstyle, phpmd, pdepend, phpcpd) to verify the quality of our code.

All of these are run on our integration branch, whenever a new feature is integrated.

Before this can happen the code will be reviewed by other people in the team and only if there is a consensus it will be merged into the integration branch.

For the code reviews we are using the pull request system of github. It works very similar to other code review tools, you see a diff view of the changes and are able to add comments to discuss the code.

In these comments we are using a convention of "+1", "-1" and "[B]" to give the pull request a thumbs up, down or mark it as blocked because a critical bug was found. Anyone in the team has a vote and is allowed to discuss any request.

In addition to this the unit tests and some smoke tests will be run on the code of the pull request to avoid merging broken builds.

Only if the tests are successful, there are three positive votes in total and no blocker the request will be merged.

In the beginning all of this was done manually. Developers had to run the tests before the created the pull request and only senior developers had the right to merge a pull request. Every couple of hours they would check the list of pull requests and verify that they had enough votes. It was a distraction and also prone to mistakes.

What gave us a big push in productivity was the introduction of two tools to the work-flow.

ghprb

First the GitHub pull request builder plugin, which enables Jenkins to automatically start a job for each pull request. It has additional features, like starting another build if more commits are added or recognising comments that instruct it to retest or white-list people who are allowed to create builds. It is similar to the pull request feature of Travis-CI only with the full possibilities of Jenkins to your disposal.
With this we are testing every pull request before it gets merged. We also run a limited set of static code analysis, because we want to keep this build fast to give quick feedback. Currently it takes around seven minutes.

PullRequester / plus-pull

The second tool is a simple script which was written by one of our developers and it was called PullRequester. It runs as a cron job and checks whether the pull requests satisfy the +1 and successful test rules. If all is OK it automatically merges it.
I have reimplemented this script to make it possible to publish it. You can find it under the name plus-pull on github. I have added some more features to make it also useful for open source projects.

Final Words

Even though these tools are not magic the increase of our productivity was visible. I account this mostly to way it enabled asynchronous working. No developer has to wait for a senior developer or poke someone to merge his code. As long as he can find three people who agree it will be merged automatically. And as everyone is interested in reviews themselves they tend to happen quickly.

In the screen-shot you can see how this works in practice.