Corduroy provides a python-friendly wrapper around CouchDB’s HTTP-based API. Behind the scenes it hooks into the asynchronous i/o routines from your choice of Tornado or the Requests & Gevent modules.
Using corduroy you can query the database without blocking your server’s event loop, making it ideal for CouchApp micro-middleware or scripted batch operations.
As a real world(ish) example of working with Corduroy, consider this pair of Tornado event handlers which update a url-specifed document then query a view. The first uses explicit callbacks to resume execution after each response from the database is received:
db = Database('players') class RankingsUpdater(tornado.web.RequestHandler): @tornado.web.asynchronous def post(self, player_id): self.new_score = int(self.request.body) db.get(player_id, callback=self.got_player) def got_player(doc, status): doc.score = self.new_score db.save(doc, callback=self.saved_player) def saved_player(conflicts, status): db.view('leaderboard/highscores', callback=self.got_highscores) def got_highscores(rows, status): self.write(json.dumps(rows)) self.finish()
An alternative syntax is available (when using Tornado) through the use of the @relax
decorator. Instead of defining callbacks for each database operation, the library can be called as
part of a yield
expression.
Tornado’s generator
module will intercept these yield
s and provide a callback automatically. The result
is code that looks quite sequential but will still execute asynchronously:
class RankingsUpdater(tornado.web.RequestHandler): @relax def post(self, player_id): # update this player's score doc = yield db.get(player_id) doc.score = int(self.request.body) yield db.save(doc) # return the new rankings highscores = yield db.view('leaderboard/highscores') self.write(json.dumps(highscores)) self.finish()
For a gentle introduction to Corduroy (and CouchDB in general), take a look at the Guide. Documentation for all of Corduroy’s module-level classes can be found in the Reference section.
Corduroy can be found on PyPi and can be installed
with your choice of pip
or easy_install
.
Download corduroy-0.9.1.tar.gz or clone the repository.
tar xzf corduroy-0.9.1.tar.gz cd corduroy-0.9.1 python setup.py install
If you’re writing a Tornado app, Corduroy can use its pure-python HTTP client by installing with:
pip install corduroy tornado
Or if you’d prefer the libcurl-based client (which supports pooling and other niceties), use:
pip install corduroy tornado pycurl
If pycurl complains (I’m looking at you, OS X), try:
env ARCHFLAGS="-arch x86_64" pip install pycurl
Gevent users can install with:
pip install corduroy requests gevent
The library can also be used with plain-old blocking i/o:
pip install corduroy requests
Corduroy is released under the BSD license. Use it freely and in good health.
Corduroy is derived from Christopher Lenz’s excellent couchdb-python module and inherits much of its API (and most of its test cases) from that codebase. It is also indebted to Eric Naeseth’s mind-expanding Swirl library which first acquainted me with the idea of using generators to simulate sequential code.