pushover

pushover

Serve up git repositories over http and accept git pushes.

This library makes it super easy to set up custom git push deploy logic.

example

var pushover = require('pushover');
var repos = pushover('/tmp/repos');

repos.on('push', function (push) {
    console.log('push ' + push.repo + '/' + push.commit
        + ' (' + push.branch + ')'
    );
    push.accept();
});

repos.on('fetch', function (fetch) {
    console.log('fetch ' + fetch.commit);
    fetch.accept();
});

var http = require('http');
var server = http.createServer(function (req, res) {
    repos.handle(req, res);
});
server.listen(7000);

then start up the pushover server...

$ node example/simple.js 

meanwhile...

$ git push http://localhost:7000/beep master
Counting objects: 356, done.
Delta compression using up to 2 threads.
Compressing objects: 100% (133/133), done.
Writing objects: 100% (356/356), 46.20 KiB, done.
Total 356 (delta 210), reused 355 (delta 210)
To http://localhost:7000/beep
 * [new branch]      master -> master

and then...

$ node example/simple.js 
push beep.git/d5013a53a0e139804e729a12107fc212f11e64c3 (master)

or...

$ git clone http://localhost:7000/beep.git

and then...

fetch beep.git/d5013a53a0e139804e729a12107fc212f11e64c3

methods

var pushover = require('pushover')

var repos = pushover(repoDir, opts={autoCreate:true})

Create a new repository collection from the directory repoDir. repoDir should be entirely empty except for git repo directories.

If repoDir is a function, repoDir(repo) will be used to dynamically resolve project directories. The return value of repoDir(repo) should be a string path specifying where to put the string repo. Make sure to return the same value for repo every time since repoDir(repo) will be called multiple times.

The return value, repos is an EventEmitter that emits the events listed below in the events section.

By default, repository targets will be created if they don't exist. You can disable that behavior with opts.autoCreate.

If opts.checkout is true, create and expected checked-out repos instead of bare repos.

repos.handle(req, res)

Handle incoming HTTP requests with a connect-style middleware.

Everything is admin-party by default. Check the credentials further up the stack using basic auth or whatevs.

repos.create(repoName, cb)

Create a new bare repository repoName in the instance repository directory.

Optionally get a callback cb(err) to be notified when the repository was created.

repos.mkdir(dir, cb)

Create a subdirectory dir in the repo dir with an errback cb(err).

repos.list(cb)

Get a list of all the repositories in the callback cb(err, repos).

repos.exists(repoName, cb)

Find out whether repoName exists in the callback cb(exists).

events

repos.on('push', function (push) { ... }

Emitted when somebody does a git push to the repo.

Exactly one listener must call push.accept() or push.reject(). If there are no listeners, push.accept() is called automatically.

push is an http duplex object (see below) with these extra properties:

• push.repo
• push.commit
• push.branch

repos.on('tag', function (tag) { ... }

Emitted when somebody does a git push --tags to the repo.

Exactly one listener must call tag.accept() or tag.reject(). If there are no listeners, tag.accept() is called automatically.

tag is an http duplex object (see below) with these extra properties:

• tag.repo
• tag.commit
• tag.version

repos.on('fetch', function (fetch) { ... }

Emitted when somebody does a git fetch to the repo (which happens whenever you do a git pull or a git clone).

Exactly one listener must call fetch.accept() or fetch.reject(). If there are no listeners, fetch.accept() is called automatically.

fetch is an http duplex objects (see below) with these extra properties:

• fetch.repo
• fetch.commit

repos.on('info', function (info) { ... }

Emitted when the repo is queried for info before doing other commands.

Exactly one listener must call info.accept() or info.reject(). If there are no listeners, info.accept() is called automatically.

info is an http duplex object (see below) with these extra properties:

• info.repo

repos.on('head', function (head) { ... }

Emitted when the repo is queried for HEAD before doing other commands.

Exactly one listener must call head.accept() or head.reject(). If there are no listeners, head.accept() is called automatically.

head is an http duplex object (see below) with these extra properties:

• head.repo

push.on('response', function(response, done) { ... })

Emitted when pushover creates a resposne stream that will be sent to the git client on the other end.

This should really only be used if you want to send verbose or error messages to the remote git client.

response is a writable stream that can accept buffers containing git packfile sidechannel transfer protocol encoded strings. done is a callback that must be called when you want to end the response.

If you create a response listener then you must either call the done function or execute the following end sequence when you want to end the response:

response.queue(new Buffer('0000'))
response.queue(null)

If you never use the response event then the above data will be sent by default. Binding a listener to the response event will prevent the end sequence those from being sent, so you must send them yourself after sending any other messages.

http duplex objects

The arguments to each of the events 'push', 'fetch', 'info', and 'head' are http duplex that act as both http server request and http server response objects so you can pipe to and from them.

For every event if there are no listeners dup.accept() will be called automatically.

dup.accept()

Accept the pending request.

dup.reject()

Reject the pending request.

install

With npm do:

npm install pushover

license

MIT

kudos

Reading through grack was super handy.