The Agnostic GitHub API
It doesn't know, and you don't care!
agithub
is a REST API client with transparent syntax which facilitates
rapid prototyping — on any REST API!
Originally tailored to the GitHub REST API, AGitHub has grown up to support many other REST APIs:
- DigitalOcean
- GitHub
- OpenWeatherMap
- SalesForce
Additionally, you can add full support for another REST API with very little new code! To see how, check out the Facebook client, which has about 30 lines of code.
This works because AGithub knows everything it needs to about protocol (REST, HTTP, TCP), but assumes nothing about your upstream API.
Use
The most striking quality of AGitHub is how closely its syntax emulates
HTTP. In fact, you might find it even more convenient than HTTP, and
almost as general (as far as REST APIs go, anyway). The examples below
tend to use the GitHub API as a reference point, but it is no less easy to
use agithub
with, say, the Facebook Graph.
Create a client
from agithub.GitHub import GitHub
client = GitHub()
GET
Here's how to do a GET
request, with properly-encoded url parameters:
client.issues.get(filter='subscribed')
That is equivalent to the following:
GET /issues/?filter=subscribed
POST
Here's how to send a request body along with your request:
some_object = {'foo': 'bar'}
client.video.upload.post(body=some_object, tags="social devcon")
This will send the following request, with some_object
serialized as
the request body:*
POST /video/upload?tags=social+devcon
{"foo": "bar"}
The body
parameter is reserved and is used to define the request body to be
POSTed. tags
is an example query parameter, showing that you can pass both
an object to send as the request body as well as query parameters.
* For now, the request body is limited to JSON data; but we plan to add support for other types as well
Parameters
headers
Pass custom http headers in your ruquest with the reserved parameter headers
.
from agithub.GitHub import GitHub
g = GitHub()
headers = {'Accept': 'application/vnd.github.symmetra-preview+json'}
status, data = g.search.labels.get(headers=headers, repository_id=401025, q='¯\_(ツ)_/¯')
print(data['items'][0])
{u'default': False, u'name': u'\xaf\\_(\u30c4)_/\xaf', u'url': u'https://api.github.com/repos/github/hub/labels/%C2%AF%5C_(%E3%83%84)_/%C2%AF', u'color': u'008672', u'node_id': u'MDU6TGFiZWwxMTcwNjYzNTM=', u'score': 43.937515, u'id': 117066353, u'description': u''}
body
If you're using POST
, PUT
, or PATCH
(post()
, put()
, or patch()
),
then you should include the body as the body=
argument. The body is
serialized to JSON before sending it out on the wire.
from agithub.GitHub import GitHub
g = GitHub()
# This Content-Type header is only required in this example due to a GitHub
# requirement for this specific markdown.raw API endpoint
headers={'Content-Type': 'text/plain'}
body = '# This should be my header'
status, data = g.markdown.raw.post(body=body, headers=headers)
print(data)
<h1>
<a id="user-content-this-should-be-my-header" class="anchor" href="#this-should-be-my-header" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>This should be my header</h1>
Example App
-
First, instantiate a
GitHub
object.from agithub.GitHub import GitHub g = GitHub()
-
When you make a request, the status and response body are passed back as a tuple.
status, data = g.users.octocat.get() print(data['name']) print(status)
The Octocat 200
-
If you forget the request method,
agithub
will complain that you haven't provided enough information to complete the request.g.users
<class 'agithub.github.IncompleteRequest'>: /users
-
Sometimes, it is inconvenient (or impossible) to refer to a URL as a chain of attributes, so indexing syntax is provided as well. It behaves exactly the same. In these examples we use indexing syntax because you can't have a python function name
- starting with a digit :
1
- containing a dash (
-
) character :Spoon-Knife
g.repos.github.hub.issues[1].get() g.repos.octocat['Spoon-Knife'].branches.get()
(200, { 'id': '#blah', ... }) (200, [ list, of, branches ])
- starting with a digit :
-
You can also pass query parameter to the API as function parameters to the method function (e.g.
get
).status, data = g.repos.octocat['Spoon-Knife'].issues.get( state='all', creator='octocat') print(data[0].keys()) print(status)
[u'labels', u'number', … , u'assignees'] 200
Notice the syntax here:
<API-object>.<URL-path>.<request-method>(<query-parameters>)
API-object
:g
URL-path
:repos.octocat['Spoon-Knife'].issues
request-method
:get
query-parameters
:state='all', creator='octocat'
-
As a weird quirk of the implementation, you may build a partial call to the upstream API, and use it later.
def following(self, user): return self.user.following[user].get myCall = following(g, 'octocat') if 204 == myCall()[0]: print 'You are following octocat'
You are following octocat
You may find this useful — or not.
-
Finally,
agithub
knows nothing at all about the GitHub API, and it won't second-guess you.g.funny.I.donna.remember.that.one.head()
(404, {'message': 'Not Found'})
The error message you get is directly from GitHub's API. This gives you all of the information you need to survey the situation.
-
If you need more information, the response headers of the previous request are available via the
getheaders()
method.g.getheaders()
[('status', '404 Not Found'), ('x-ratelimit-remaining', '54'), … ('server', 'GitHub.com')]
Note that the headers are standardized to all lower case. So though, in this example, GitHub returns a header of
X-RateLimit-Remaining
the header is returned fromgetheaders
asx-ratelimit-remaining
Error handling
Errors are handled in the most transparent way possible: they are passed on to you for further scrutiny. There are two kinds of errors that can crop up:
-
Networking Exceptions (from the
http
library). Catch these withtry .. catch
blocks, as you otherwise would. -
GitHub API errors. These mean you're doing something wrong with the API, and they are always evident in the response's status. The API considerately returns a helpful error message in the JSON body.
Specific REST APIs
agithub
includes a handful of implementations for specific REST APIs. The
example above uses the GitHub API but only for demonstration purposes. It
doesn't include any GitHub specific functionality (for example, authentication).
Here is a summary of additional functionality available for each distinct REST
API with support included in agithub
. Keep in mind, agithub
is designed
to be extended to any REST API and these are just an initial collection of APIs.
agithub/GitHub.py
GitHub : GitHub Authentication
To initiate an authenticated GitHub
object, pass it your username and password
or a token.
from agithub.GitHub import GitHub
g = GitHub('user', 'pass')
from agithub.GitHub import GitHub
g = GitHub(token='token')
GitHub Pagination
When calling the GitHub API with a query that returns many results, GitHub will
paginate the response, requiring
you to request each page of results with separate API calls. If you'd like to
automatically fetch all pages, you can enable pagination in the GitHub
object
by setting paginate
to True
.
from agithub.GitHub import GitHub
g = GitHub(paginate=True)
status, data = g.repos.octocat['Spoon-Knife'].issues.get()
status, data = g.users.octocat.repos.get(per_page=1)
print(len(data))
8
(added in v2.2.0)
GitHub Rate Limiting
By default, if GitHub returns a response indicating that a request was refused due to rate limiting, agithub will wait until the point in time when the rate limit is lifted and attempt the call again.
If you'd like to disable this behavior and instead just return the error
response from GitHub set sleep_on_ratelimit
to False
.
from agithub.GitHub import GitHub
g = GitHub(sleep_on_ratelimit=False)
status, data = g.repos.octocat['Spoon-Knife'].issues.get()
print(status)
print(data['message'])
403
API rate limit exceeded for 203.0.113.2. (But here's the good news: Authenticated requests get a higher rate limit. Check out the documentation for more details.)
(added in v2.2.0)
GitHub Logging
To see log messages related to GitHub specific features like pagination and rate limiting, you can use a root logger from the Python logging module.
import logging
logging.basicConfig()
logger = logging.getLogger() # The root logger
logger.setLevel(logging.DEBUG)
from agithub.GitHub import GitHub
g = GitHub(paginate=True)
status, data = g.repos.octocat['Spoon-Knife'].issues.get()
DEBUG:agithub.GitHub:No GitHub ratelimit remaining. Sleeping for 676 seconds until 14:22:43 before trying API call again.
DEBUG:agithub.GitHub:Fetching an additional paginated GitHub response page at https://api.github.com/repositories/1300192/issues?page=2
DEBUG:agithub.GitHub:Fetching an additional paginated GitHub response page at https://api.github.com/repositories/1300192/issues?page=3
…
Semantics
Here's how agithub
works, under the hood:
- It translates a sequence of attribute look-ups into a URL; The Python method you call at the end of the chain determines the HTTP method to use for the request.
- The Python method also receives
name=value
arguments, which it interprets as follows:headers=
-
You can include custom headers as a dictionary supplied to the
headers=
argument. Some headers are provided by default (such as User-Agent). If these occur in the supplied dictionary, the default value will be overridden.headers = {'Accept': 'application/vnd.github.loki-preview+json'}
-
body=
- If you're using
POST
,PUT
, orPATCH
(post()
,put()
, andpatch()
), then you should include the body as thebody=
argument. The body is serialized to JSON before sending it out on the wire.
- If you're using
- GET Parameters
- Any other arguments to the Python method become GET parameters, and are tacked onto the end of the URL. They are, of course, url-encoded for you.
- When the response is received,
agithub
looks at its content type to determine how to handle it, possibly decoding it from the given char-set to Python's Unicode representation, then converting to an appropriate form, then passed to you along with the response status code. (A JSON object is de-serialized into a Python object.)
Extensibility
agithub
has been written in an extensible way. You can easily:
-
Add new HTTP methods by extending the
Client
class with new Python methods of the same name (and adding them to thehttp_methods
list). -
Add new default headers to the
_default_headers
dictionary. Just make sure that the header names are lower case. -
Add a new media-type (a.k.a. content-type a.k.a mime-type) by inserting a new method into the
ResponseBody
class, replacing'-'
and'/'
with'_'
in the method name. That method will then be responsible for converting the response body to a usable form — and for callingdecode_body
to do char-set conversion, if required. For example to create a handler for the content-typeapplication/xml
you'd extendResponseBody
and create a new method like thisimport xml.etree.ElementTree as ET class CustomResponseBody(ResponseBody): def __init__(self): super(ChildB, self).__init__() def application_xml(self): # Handles Content-Type of "application/xml" return ET.fromstring(self.body)
And if all else fails, you can strap in, and take 15 minutes to read and become an expert on the code. From there, anything's possible.
License
Copyright 2012–2016 Jonathan Paugh and contributors See COPYING for license details