Personal tools

Diagrams/Contributing

From HaskellWiki

< Diagrams(Difference between revisions)
Jump to: navigation, search
(Submitting a pull request: info on submitting pull requests)
(advanced info on pull requests etc.)
Line 73: Line 73:
 
If any errors are pointed out, changes requested, etc., simply make some new commits and push them to your forked repo. There is no need to create another pull request; any newly pushed commits will be automatically added to the existing pull request.
 
If any errors are pointed out, changes requested, etc., simply make some new commits and push them to your forked repo. There is no need to create another pull request; any newly pushed commits will be automatically added to the existing pull request.
   
=== Old text ===
+
=== Pull requests and topic branches ===
   
XXX walk through example of checking out repo, making and committing changes, and submitting a pull request. Show how to do it manually and also using 'hub' utility.
+
The important thing to realize about pull requests is that they do not correspond to a particular set of commits, but to a ''branch''. In particular, you add new commits to a pull request simply by pushing to the branch which the pull request is from. This has some interesting implications:
   
* use topic branches
+
* If you are going to be working on multiple features/bug fixes at once---or even if you just want to be able to get started on a new feature while your previous one is still undergoing review---it is best to create a "topic branch" for each feature, rather than making all your changes on your "master" branch. You can create a new branch called <code>foo</code> with <code>git branch foo</code>; see all branches with <code>git branch</code>; and switch between branches using <code>git checkout</code>.
* encourage submitting update to documentation
+
* Once your pull request is merged, you can delete the branch, using <code>git branch -d foo</code> for the branch in your local repo, and something like <code>git push origin --delete foo</code> for a remote repo named <code>origin</code>.
* link to coding style document
+
* Another interesting point is that there's not necessarily any reason to wait until you are "done" to open a pull request. Just make sure you state that the feature is "in progess", and then you can get useful early feedback as you continue to work on the feature and push more commits. Of course, branches/pull requests also make for a nice way to work on a new feature collaboratively.
 
TODO: expand/clean up the below
 
 
But the short version is that a pull request is not for a set of
 
commits, it is for a *branch*. In particular, you add new commits to
 
a pull request simply by pushing to the branch which the pull request
 
is from. So best practice when working on a new feature is
 
 
* make a new branch (a "feature" or "topic" branch). If you have push access you can make a branch directly in the main repo; otherwise, fork it and make a branch in your fork.
 
* Make some commits in your branch.
 
* Open a pull request from your branch to 'master'.
 
* If necessary, make revisions etc. by pushing additional commits to the branch. There's no need to "revise" the original commits; it's useful to have the entire history of development with comments, etc.
 
* Once your commits have been merged, you can delete the branch.
 
 
The important point is that it's usually a bad idea to open a pull
 
request from the 'master' branch of your fork because then you can't
 
do anything else while waiting for your changes to be merged.
 
 
Another interesting point is that there's not necessarily any reason
 
to wait until you are "done" to open a pull request. Just make sure
 
you state that the feature is "in progess", and then you can get
 
useful early feedback as you continue to work on the feature and push
 
more commits. Of course, branches/pull requests also make for a nice
 
way to work on a new feature collaboratively.
 

Revision as of 16:05, 21 August 2012

This page is under construction -- please help!

Contents

1 Getting involved

There are two major ways to get involved in the diagrams community and find out what is going on:

  • The IRC channel (#diagrams on freenode.org) is fairly active and a good place to interact with other diagrams users and developers. (Be patient: sometimes no one is watching the channel, but if you say something or ask a question, you can be sure that someone will eventually see it and respond.)
  • The mailing list is the place to stay up-to-date with announcements, and also a good place to ask questions, especially longer or more involved ones.

2 Getting the sources

All the core diagrams code can be found in the diagrams organization on github.

If you don't already know how to work with repositories hosted on github, take a look at a tutorial. The hub utility is an optional but recommended tool for simplifying common tasks involving github.

3 Choosing a project

If you would like to begin contributing to the diagrams project but are not sure where to start, here are a few resources that may be helpful:

4 Best practices

As a running example, let's suppose you want to add a function to the Diagrams.TwoD.Shapes module (from the diagrams-lib package) to draw a diamond shape.

4.1 Getting the code

To begin, you will need a github account, and to fork and clone some of the diagrams repositories. See here for help on how to fork a repo. At a minimum, you will need the following repositories:

as well as one of the following two backends:

If you do not plan to make any modifications to the code in a certain repo, you may clone it directly instead of first forking on github and then cloning from your fork, for example

git clone https://github.com/diagrams/diagrams-core

or, using hub, simply

hub clone diagrams/diagrams-core

(In fact, hub makes it easy to later convert a direct clone into your own fork if you wish, via the hub fork command; see the hub documentation for more information.)

4.2 Building

It's recommended to use some sort of sandboxing tool while working on diagrams. Because diagrams consists of several separate packages, using cabal-dev can be something of a pain. Instead, we recommend using hsenv---though you will have to build it from source, and it is only known to work on Linux systems. Good suggestions of similar tools for Windows or Mac are welcome.

In any case, there is an important trick for building multiple local packages at once that you should know, which works with cabal as well as cabal-dev. Instead of installing each package one at a time, simply go up to the parent directory and issue a command such as

cabal install monoid-extras/ dual-tree/ diagrams-core/ diagrams-lib/ diagrams-cairo/ diagrams-contrib/

The trailing slashes tell cabal to install packages from local directories, rather than trying to download packages from Hackage. In addition, it doesn't even matter in what order you list the directories; cabal will figure out the correct order based on dependencies.

4.3 Making changes

Now that you have your cloned/forked repositories and know how to build them, go ahead and make some edits. You can see what changes you've made using the git diff command, stage certain changes with git add (try the -p flag!), and create a commit from staged changes with git commit.

4.4 Submitting a pull request

Once you have a set of commits you are happy with, push them to your forked repository on github and open a pull request. At this point your code will be reviewed by someone with push access to the repository. They may very well leave some comments; feel free to respond with comments of your own.

If any errors are pointed out, changes requested, etc., simply make some new commits and push them to your forked repo. There is no need to create another pull request; any newly pushed commits will be automatically added to the existing pull request.

4.5 Pull requests and topic branches

The important thing to realize about pull requests is that they do not correspond to a particular set of commits, but to a branch. In particular, you add new commits to a pull request simply by pushing to the branch which the pull request is from. This has some interesting implications:

  • If you are going to be working on multiple features/bug fixes at once---or even if you just want to be able to get started on a new feature while your previous one is still undergoing review---it is best to create a "topic branch" for each feature, rather than making all your changes on your "master" branch. You can create a new branch called foo with git branch foo; see all branches with git branch; and switch between branches using git checkout.
  • Once your pull request is merged, you can delete the branch, using git branch -d foo for the branch in your local repo, and something like git push origin --delete foo for a remote repo named origin.
  • Another interesting point is that there's not necessarily any reason to wait until you are "done" to open a pull request. Just make sure you state that the feature is "in progess", and then you can get useful early feedback as you continue to work on the feature and push more commits. Of course, branches/pull requests also make for a nice way to work on a new feature collaboratively.