SymPy Bot

The goal of SymPy Bot is to do all the automated testing for a pull request and report back into the pull request with the results.

So far one has to run the bot manually, but eventually we would like to create a web service for it.

Usage

List all pull requests, sorted by date:

./sympy-bot list

Make an automatic review of a pull request:

./sympy-bot review 268

This will run all tests and then comment in the pull request (under your name) with the results.

To review all open pull requests, do:

./sympy-bot review all

to only review mergeable pull requests, do:

./sympy-bot review mergeable

Requirements

SymPy bot needs argparse to run. This is part of the standard library in Python 2.7, however it can be installed in earlier versions of Python.

Tips

By default, the sympy repository is fully downloaded from the web, so you don't need to have any local copy. However, if you do have a local copy already, you can skip most of the download (which might take a few minutes on slower connections) by passing a --reference option to sympy-bot:

./sympy-bot --reference ~/repo/git/sympy review 268

This gets passed to git, see git clone --help for more information. Then sympy-bot starts testing the branch immediately, even if you have a slower connections.

This has another advantage: if you run ./bin/use2to3 in the reference directory and use the --copy-py3k-sympy option to sympy-bot, the py3k-sympy directory will be copied over to the testing directory, saving time (only those files that are changed by the pull request will have to be converted by 2to3 again). If you want to always enable this, add copy_py3k_sympy = True to your sympy-bot.conf file (see the next section). To remain effective, you should run ./bin/use2to3 off of master in the reference SymPy git repository on a regular basis.

Note that this can occasionally produce false test failures. If this happens, delete and recreate the py3k-sympy directory from the reference SymPy git repository, or don't use the --copy-py3k-sympy option. If you have the option set in your config file, you can disable it with --copy-py3k-sympy=False.

Configuration

You can configure SymPy bot to remember your GitHub credentials, use an existing clone of sympy and run interpreters under different profiles. This is done in the ~/.sympy/sympy-bot.conf file. The configuration supports multiple profiles, but will always read in the [default] profile, so you should start your configuration with your GitHub credentials in the default profile:

[default]
user = username
password = password

If you have an existing clone of sympy, you can avoid having to clone the SymPy repository every time the bot is run:

reference = ~/path/to/sympy

You can specify the interpreters to use by giving a comma separated list of Python interpreters:

interpreter = /path/to/python, /path/to/other/python
interpreter3 = /path/to/python3, /path/to/other/python3

which sets the Python 2 and Python 3 interpreters, respectively. By default, the interpreters set by interpreter are run. If you pass the -3 flag or set python3 = True in the configuration file, then interpreter3 will be used instead. You can force both by passing -2 and -3 or setting both python2 = True and python3 = True in the configuration file. Setting interpreter = None will disable the Python tests, which can be useful in setting up a profile just for testing docs.

If you want to test the building of the HTML docs, you can use the -D flag or set build_docs = True in the configuration file. By default, this will disable running the tests. This can be overridden by setting python2 or python3 options, as above.

Any of the other options set by commandline parameters can be set in the configuration file. See sympy-bot review --help for more information (the configuration values are the long form of the option, with any dashes replaced with underscores, for example, --build-docs becomes build_docs).

The configuration also supports different profiles. To set these up, you put the name of the profile between square brackets. Then, when you pass --profile profile_name, the options in the specified section will override the default section. This is done in the config file:

[profile_name]
interpreter = /path/to/different/python
testcommand = bin/test --other-options

This can be useful for setting up various suites of tests, e.g. slow tests, 32-bit/64-bit tests, etc.

To see an example configuration file, see the sympy-bot.conf.example file. This file also explains how you can use variable interpolation to avoid duplication.

Foreign repositories

SymPy Bot can be also used with other remote repository than sympy/sympy. You can change the remote with -R flag to sympy-bot or by setting repository in configuration file. The new remote doesn't have to be SymPy's repository, but any repository on GitHub. Note that in this case you may need to setup customized testcommand.

Custom Master Commit

By default, sympy-bot merges with master before testing, failing if the merge fails. You can customize this behavior with the -m option to sympy-bot. Pass any valid git commit name to this option, and it will use it to merge the master branch. The default is origin/master, which is the current master. If you don't want to merge at all, pass HEAD, which will perform a noop merge against the branch you are testing.

If you use --reference, git will pull in all commits from the local repository. Thus, you can merge with commits that are not in the official sympy/sympy repository by using this and passing the SHA1 of the commit you want.

This is also useful for bisecting problems with SymPy Bot. Simply use git to bisect in your local SymPy repository and pass the SHA1's it picks to sympy-bot -n -m.

Web interface integration with github

This way is a bit complicated in set up than previous (poll github for new pulls), but that will update information about pulls in real time.

SymPy Bot web-interface (which located in under web/) supports integration with github via mechanism called hooks http://developer.github.com/v3/repos/hooks/

To use that feature you need to follow these steps:

1. Go to http://example.com/upload_pull, sign in as administrator and press generate button. After that, all admins will recieve notification with secret URL (you can see a log of all generations in table on that page)

2. You need to tell github to use this URL, so here steps (replace username and repo with you values):

   • Go to https://github.com/user/repo/admin/hooks

   • Click on WebHook URLs and add secret URL there.

   • Find the hook that you want to modify by:

     curl -u username https://api.github.com/repos/username/repo/hooks
     

     the id field gives the hook ID, copy and paste the path in the "url" field into the command:

     curl -u username -d '{ "events": [ "pull_request" ] }'
     https://api.github.com/repos/username/repo/hooks/ID
     

     You will see that the "events" part:

     "events": [
         "push"
     ],
     

     changed to:

     "events": [
         "pull_request"
     ],