It fetches price data from Oanda using its python bindings for REST API v20.

To use the fetcher, you need python 3 with packages v20 and pandas installed. They can be installed individually or by

pip install -r requirements.txt

Configuration parameters for communicating with Oanda servers are stored in a config file. A sample config file may look like:

[DEFAULT]
hostname = api-fxpractice.oanda.com
port = 443
ssl = true
token = XXXXXXXXXXXXXXXXXXXXXXXXXXXXX
datetime_format = RFC3339

where token is required to grant access to the REST API. See developer page of Oanda website to see how to sign up and get a free access token.

After preparing a config file, you can run fetcher to get price data. For instance,

python3 fetch_oandata.py EUR_USD 2020-04-05 2020-05-16 -g M30 -c /path/to/config -o /path/to/csv

fetches the price of EURO in USD within the given period with granularity of 30 minutes. The result is stored in the given CSV file. By default the mid price of the instrument is fetched. To obtain bid or ask prices, -p B or -p A can be added to the list of arguments, respectively. To see the supported granularity and a complete list of arguments, run

python3 fetch_oandata -h

If the given period is long and granularity is fine, i.e. historical price data contain lots of candle sticks, fetching them will sometimes get failed. To mitigate the problem, the period is split into several sub-intervals, each holding a fewer number of candle sticks. The number of those sub-intervals can be set by --split or -s. Adding -s 10 to the list of arguments above, for example, splits the period into 10 sub-intervals of roughly the same size and fetch the prices for each sub-interval individually before merging them together. If it is not set, it will be computed according to the given period and granularity.

Fetch of price data for each sub-interval can nevertheless fail for other reasons. In this case, the fetcher retries again. The number of attempts before reporting failure can be set via -r.