Hubzilla Documentation: Developers

Contents

Who is a Hubzilla developer? Should I read this?


Anyone who contributes to making Hubzilla better is a developer. There are many different and important ways you can contribute to this amazing technology, even if you do not know how to write code. The software itself is only a part of the Hubzilla project. You can contribute by
  • translating text to your language so that people around the world have the opportunity to use Hubzilla
  • promoting Hubzilla and spreading awareness of the platform through blog posts, articles, and word-of-mouth
  • creating artwork and graphics for project assets such as icons and marketing material
  • supporting project infrastructure like the project website and demo servers
Software developers are of course welcomed; there are so many great ideas to implement and not enough people to make them all a reality! The Hubzilla code base is an advanced and mature system, but the platform is still very flexible and responsive to new ideas.

We're pretty relaxed when it comes to developers. We don't have a lot of rules. Some of us are over-worked and if you want to help we're happy to let you help. That said, attention to a few guidelines will make the process smoother and make it easier to work together. All developers are expected to abide by our code of conduct. We have developers from across the globe with different abilities and different cultural backgrounds and different levels of patience. Our primary rule is to respect others. Sometimes this is hard and sometimes we have very different opinions of how things should work, but if everybody makes an effort, we'll get along just fine.

This document will help you get started learning and contributing to Hubzilla.

Versions and Releases


Hubzilla currently uses a standard version numbering sequence of $x.$y(.$z), for instance '1.12' or '1.12.1'. The first digit is the major version number. Major versions are released "roughly" once per year; often in December.

The second digit is the minor release number. If this number is odd, it is a development version. If the number is even, it is a released version. Minor versions are released (moved from dev to master) typically once per month when development is 'stable', but this is likely to increase. Going forward minor releases will be made somewhere between one and three months; corresponding to a stable code point and when there is general community consensus that the current code base is stable enough to consider a release.

The final digit is an interface or patch designator.

The release process involves changing the version number (by definition the minor version number will be odd, and the minor number will be incremented). Once a year for a major release the major version will be incremented, and the minor number reset to 0.

The release candidate is moved to a new branch; and testing will commence/continue for a period of 1-2 weeks afterward or until any significant issues have been resolved. This branch is usually labelled with RC (release candidate); for instance 1.8RC represents the pending release of version 1.8. At this time, the minor version number on the dev branch is incremented to the next odd number. (For instance 1.9). New development can then take place in the dev branch.

Bug fixes should always be applied to 'dev' and from there merged forward (typically with git cherry-pick) to the RC branch and if necessary applied to the master or official release branch.

At the time a release candidate is produced, the language strings file is frozen until a release is made. Translation work may continue, but all translations should be submitted to 'dev' and merged forward to RC.

Once RC testing is completed, RC is merged to 'master' and the RC version designator removed; resulting in one final checkin to change the version number. The CHANGELOG file should also be updated at or just prior to this time. If there are merge conflicts during this final merge, the merge will be abandoned; and 'git merge -s ours' applied. This results in a replacement of master with the contents of the RC branch. Conflicts often arise with string updates which were made to master after the last release and cannot easily be resolved without hand editing. Since this is a release of tested code, hand editing is discouraged, and the replacement merge strategy should be used instead. It is assumed that RC now contains the most recent well-tested code.

Once the release is live and merged to master, the RC branch may be removed.

Fixes may be made to master after release. Where possible these should be made to dev and 'git cherry-pick' used to merge forward; which preserves the commit info and prevents merge conflicts in the next cycle. Only rarely does a patch only apply to the master branch. If necessary this can be made. If the change is severe, the interface version number should be incremented. This is at the discretion of the community. In any event, a 'git pull' of the master branch should always result in the latest release with any post-release patches applied.

The interface number (the $z in $x.$y.$z) should be incremented in dev whenever a change is made which changes the interfaces or API in incompatible ways so that any external packages (especially addons and API clients) relying on a the current behaviour can discover and change their own interfaces accordingly at the point that it changed.  

Git repository branches


There are two official branches of the Hubzilla git repo.
  • The stable version is maintained on the master branch. The latest commit in this branch is considered to be suitable for production hubs.
  • Experimental development occurs on the dev branch, which is merged into master when it is deemed tested and stable enough.

Developer tools and workflows


Hub Snapshots


The hub snapshots page provides instructions and scripts for taking complete snapshots of a hub to support switching between consistent and completely known states. This is useful to prevent situations where the content or database schema might be incompatible with the code.

Translations


Our translations are managed through Transifex. If you wish to help out translating Hubzilla to another language, sign up on transifex.com, visit Transifex and request to join one of the existing language teams or create a new one. Notify one of the core developers when you have a translation update which requires merging, or ask about merging it yourself if you're comfortable with git and PHP. We have a string file called 'messages.po' which is gettext compliant and a handful of email templates, and from there we automatically generate the application's language files.  

Translation Process


The strings used in the UI of Hubzilla is translated at Transifex and then
included in the git repository at github. If you want to help with translation
for any language, be it correcting terms or translating Hubzilla to a
currently not supported language, please register an account at transifex.com
and contact the translation team there.

Translating Hubzilla is simple. Just use the online tool at transifex. If you
don't want to deal with git & co. that is fine, we check the status of the translations
regularly and import them into the source tree at github so that others can use them.

We do not include every translation from transifex in the source tree to avoid
a scattered and disturbed overall experience. As an uneducated guess we have a
lower limit of 50% translated strings before we include the language. This
limit is judging only by the amount of translated strings under the assumption
that the most prominent strings for the UI will be translated first by a translation
team. If you feel your translation useable before this limit,
please contact us and we will probably include your teams work in the source tree.

If you want to get your work into the source tree yourself, feel free to do so
and contact us with and question that arises. The process is simple and
Hubzilla ships with all the tools necessary.

The location of the translated files in the source tree is
    /view/LNG-CODE/
where LNG-CODE is the language code used, e.g. de for German or fr for French.
For the email templates (the *.tpl files) just place them into the directory
and you are done. The translated strings come as a "hmessages.po" file from
Transifex which needs to be translated into the PHP file Hubzilla uses.  To do
so, place the file in the directory mentioned above and use the "po2php"
utility from the util directory of your Hubzilla installation.

Assuming you want to convert the German localization which is placed in
view/de/hmessages.po you would do the following.

1. Navigate at the command prompt to the base directory of your
   Hubzilla installation

2. Execute the po2php script, which will place the translation
   in the hstrings.php file that is used by Hubzilla.

       $> php util/po2php.php view/de/hmessages.po

   The output of the script will be placed at view/de/hstrings.php where
   froemdoca os expecting it, so you can test your translation mmediately.
                                  
3. Visit your Hubzilla page to check if it still works in the language you
   just translated. If not try to find the error, most likely PHP will give
   you a hint in the log/warnings.about the error.
                                        
   For debugging you can also try to "run" the file with PHP. This should
   not give any output if the file is ok but might give a hint for
   searching the bug in the file.

       $> php view/de/hstrings.php

4. commit the two files with a meaningful commit message to your git
   repository, push it to your fork of the Hubzilla repository at github and
   issue a pull request for that commit.

Utilities


Additional to the po2php script there are some more utilities for translation
in the "util" directory of the Hubzilla source tree.  If you only want to
translate Hubzilla into another language you wont need any of these tools most
likely but it gives you an idea how the translation process of Hubzilla
works.

For further information see the utils/README file.

Known Problems


* Hubzilla uses the language setting of the visitors browser to determain the
  language for the UI. Most of the time this works, but there are some known
  quirks.
* the early translations are based on the friendica translations, if you
  some rough translations please let us know or fix them at Transifex.

Licensing


All code contributed to the project falls under the MIT license, unless otherwise specified. We will accept third-party code which falls under MIT, BSD and LGPL, but copyleft licensing (GPL, and AGPL) is only permitted in addons. It must be possible to completely remove the GPL (copyleft) code from the main project without breaking anything.

Coding Style


In the interests of consistency we adopt the following code styling. We may accept patches using other styles, but where possible please try to provide a consistent code style. We aren't going to argue or debate the merits of this style, and it is irrelevant what project 'xyz' uses. This is not project 'xyz'. This is a baseline to try and keep the code readable now and in the future.
  • All comments should be in English.
  • We use doxygen to generate documentation. This hasn't been consistently applied, but learning it and using it are highly encouraged.
  • Indentation is accomplished primarily with tabs using a tab-width of 4.
  • String concatenation and operators should be separated by whitespace. e.g. "$foo = $bar . 'abc';" instead of "$foo=$bar.'abc';"
  • Generally speaking, we use single quotes for string variables and double quotes for SQL statements. "Here documents" should be avoided. Sometimes using double quoted strings with variable replacement is the most efficient means of creating the string. In most cases, you should be using single quotes.
  • Use whitespace liberally to enhance readability. When creating arrays with many elements, we will often set one key/value pair per line, indented from the parent line appropriately. Lining up the assignment operators takes a bit more work, but also increases readability.
  • Generally speaking, opening braces go on the same line as the thing which opens the brace. They are the last character on the line. Closing braces are on a line by themselves.
  • Some functions take arguments in argc/argv style like main() in C or function args in bash or Perl. Urls are broken up within a module. e.g, given "http://example.com/module/arg1/arg2", then $this->argc will be 3 (integer) and $this->argv will contain:   [0] => 'module', [1] => 'arg1',  [2] => 'arg2'. There will always be one argument. If provided a naked domain  URL, $this->argv[0] is set to "home".

File system layout

[/tr]
DirectoryDescription
addonoptional addons/plugins
boot.phpEvery process uses this to bootstrap the application structure
docHelp Files
imagescore required images
includeThe "model" in MVC - (back-end functions), also contains PHP "executables" for background processing
index.phpThe front-end controller for web access
installInstallation and upgrade files and DB schema
libraryThird party modules (must be license compatible)
modController modules based on URL pathname (e.g. http://sitename/foo loads mod/foo.php)
mod/site/site-specific mod overrides, excluded from git
utiltranslation tools, main English string database and other miscellaneous utilities
version.inccontains current version (auto-updated via cron for the master repository and distributed via git)
viewtheming and language files
view/(css,js,img,php,tpl)default theme files
view/(en,it,es ...)language strings and resources
view/theme/individual named themes containing (css,js,img,php,tpl) over-rides

More information needing re-organization and updating...