This feature is a big one! I can't tell you how many times someone has asked me: "I have a set of prices on my computer, can I use it in Quantopian?" or "Do you have this specific other data source in Quantopian?" Until today, the answer was no!
Fetcher permits you to import any time-series data in a CSV format into Quantopian. The data can be additional information about a stock already in Quantopian, like if you import the short interest for a list of stocks, or it can be totally stand-alone data like the price of palladium.
Fetcher imports the data into a pandas dataframe. Once in the dataframe it is sorted by datetime. Fetcher permits you to manipulate and format the imported dataframe both before and after the timesort. Once the formatting is done, the dataframe is then used as a datasource just like the existing price data that Quantopian uses.
This is incredibly powerful. Until today, you've only been able to code your algorithms against signals that you find in the price and volume stock data. Now, you can code your algorithm against any data source you can get your hands on! I particularly want to point out http://www.quandl.com/. They've put together 4 million datasets that you can use for free that are easily accessed through Quantopian. The sample algo below loads the prices of palladium and gold from Quandl, as an example.
For the full explanation of the feature, you can read the help here and here.
Here is a quick primer. The generic code that you use for Fetcher:
fetch_csv(url, pre_func=None, post_func=None, date_column='date',
date_format='%m/%d/%y', timezone='UTC', symbol=None, **kwargs)
Here is a simple implementation for a file shared on Dropbox:
fetch_csv('https://dl.dropbox.com/u/1/history.csv',
date_col='date',
date_format='%m/%d/%Y')
Here are the first few lines of that CSV, so you can see how the formatting works:
date,symbol,days_to_cover
2/28/2013,MTH,3.43719
2/15/2013,MTH,2.841933
1/31/2013,MTH,3.197315