OanPy: Python Bindings for OANDA Trading API

This open source package provides an extension module for Python that can be used to access to the programmatic API for the OANDA FX broker. The entire API's functionality is supported, including live market data updates and order placement.

The ability to do fast prototyping of trading strategies directly using a dynamic language like Python is very exciting. You can leverage numerical libraries like NumPy in order to perform analysis on market data directly from the API, or you can write custom trading GUIs leveraging toolkits, e.g., PyQt. Our API works in multi-threaded environments, and is compatible with select()/epoll() reactor-based networking libraries (and with Twisted).

How it Works

In order to use these Python bindings, you will need to independently obtain a license to the Linux C++ API from OANDA, and then run a Makefile to build and install your Python extension module, linked with this API.

Once again, this package does NOT provide the include files nor the binary library that OANDA makes available to its subscribers. You need to obtain this separately (i.e., it costs money).

With it, you will be able to import all of the functionality provided by the OANDA API with a simple import, for example:

# Import the OANDA classes.
from oanda import *

...

# Login to my OANDA account.
fxclient = FXTrade()
try:
    fxclient.login(username, password)
except OAException, e:
    raise SystemExit("Could not login: %s" % e)

# Get my account object.
user = fxclient.getUser()
account = user.getAccounts()[0]

...

# Place a trade.
order = MarketOrder()
order.base, order.quote = 'EUR', 'USD'
order.units = 10000
trades = account.execute(order)
for trade in trades:
    print trade

...

The classes and functions names correspond directly to the documentation that OANDA will provide for your API library. This package's documentation only describes difference between the official API and our mapping into Python.

Documentation

The documentation can be found here.

Source Code

A Mercurial repository can be found at:

http://github.com/blais/oanpy

Caveats

Some information about the data feed (as of 2009):

The OANDA API can be used to write trading automatons, but is not a “high-frequency” trading API by today's standards: trade confirmations take one to several seconds--contrast this with connections to today's main exchanges where you can achieve a turnaround time in the order of a few hundreds of microseconds. Note that this is perfectly fine if your holding periods are on the order of a few minutes, or even a few tens of seconds, or if you're using this to automate some sort of currency hedging.
You cannot buy at the bid nor sell at the offer, i.e., OANDA forces you to cross the spread in order to execute all orders, in particular, limit orders. Their "limit orders" are slightly different: they simply sit on their books until the spread happens to cross to your order. This is a market maker, remember, you're not trading directly with others, you're trading with OANDA's level 1 book.
The market data feed sometimes sends market data out-of-order: about 0.5% of all updates have timestamps that are either incorrect or not flushed in sequence. The server we provide has an option to filter those out.
The C++ API's network layer flushes its orders in blocks, and they are not transmitted as soon as they are available on the remote side. You might want to use the Java API if you want faster updates... but given the delays involved in "normal" operation, I doubt it's worth the effort.
OANDA indicated via its forums that its Java API is more stable than its C++ API. This is most likely because their developers are more comfortable with Java--whatever that may mean to you. Note: the Java API is fairly clean and it is not unreasonable to consider writing your code directly that API instead (if you speak Java).

Supported Platforms: Linux

Currently, we support only the Linux platform (both Ubuntu 9.04 and Gentoo are known to work).

If you are interested in support for Windows or Mac and are willing to pay for it by the hour, please contact the author.

Licensing

Copyright (C) 2009  Furius Enterprise / 6114750 Canada Inc.
Licensed under the terms of the GNU Lesser General Public License, version 3.
See http://furius.ca/oanpy/LICENSE for details.

Notes:

The GNU LGPL allows you to write private/secret programs against the library. Using this open source library does NOT imply that you need to share your code with others. You can use the library to write top-secret trading strategies to take over the world if you like :-)
The license has a liability clause in it; thus Furius is not responsible for any loss which you may incur as a result of using this code. It is entirely your own responsibility to insure that the code you're using works as you'd expect.
There are no guarantees that this API will get updated if the OANDA API changes. You may have to do this yourself if you come to depend on it. (However, it would probably be easy to do so, given you have all the source code).

Author

Can be contacted here.