Basic wrapper to return datasets from the Quandl website as Pandas dataframe objects with a timeseries index, or as a numpy array. This allows interactive manipulation of the results via IPython or storage of the datasets using Pandas I/O functions. You will need a familarity with pandas to get the most out of this.
See the Quandl API for more information.
Usage is simple and mirrors the functionality found at Quandl/API.
A request with a full list of options would be the following.
python
data = Quandl.get('PRAGUESE/PX', authtoken='xxxxxx', trim_start='2001-01-01',
trim_end='2010-01-01', collapse='annual',
transformation='rdiff', rows=4, returns='numpy')
All options beyond specifying the dataset (PRAUGESE/PX) are optional,though it is helpful to specify an authtoken at least once to increase download limits.
You can then view the dataframe with data.head().
See the pandas documentation for a wealth of options on data manipulation.
Authtokens are saved as pickled files in the local directory so it is unnecessary to enter them more than once,
unless you change your working directory. To replace simply save the new token or delete the authtoken.p file.
Using the search function you can search Quandl, specifying what sources to search, and what page you wish to show.
Quandl.search(query = "Search terms", source = "Source you wish to search", page = 1)
An example of searching for datasets having to do with oil:
python
import Quandl
datasets = Quandl.search('OIL')
datasets[0]
will output:
{u'code': u'NSE/OIL',
u'colname': [u'Date',
u'Open',
u'High',
u'Low',
u'Last',
u'Close',
u'Total Trade Quantity',
u'Turnover (Lacs)'],
u'desc': u'Historical prices for Oil India Limited (OIL), (ISIN: INE274J01014), National Stock Exchange of India.',
u'freq': u'daily',
u'name': u'Oil India Limited'}
An example of creating a pandas time series for IBM stock data, with a weekly frequency:
python
data = Quandl.get('GOOG/NYSE_IBM', collapse='weekly')
data.head()
will output
No authentication tokens found,usage will be limited
Returning Dataframe for GOOG/NYSE_IBM
Open High Low Close Volume
Date
2013-03-28 209.83 213.44 209.74 213.30 3752999
2013-03-15 215.38 215.90 213.41 214.92 7937244
2013-03-08 209.85 210.74 209.43 210.38 3700986
2013-03-01 200.65 202.94 199.36 202.91 3309434
2013-02-22 199.23 201.09 198.84 201.09 3107976
You can now upload your own data to Quandl through the Python package.
At this time the only accepted format is a date indexed Pandas DataSeries.
Things to do before you upload:
Then call the following to push the data:
python
Quandl.push(data, code='TEST', name='Test', desc='test')
All parameters but desc are necessary
If you wish to override the existing set at code TEST add override=True.
Uploading a pandas DataSeries with random data:
```python import pandas import numpy
index = ['Dec 12 2296', 'Dec 21 1998', 'Oct 9 2000', 'Oct 19 2001', 'Oct 30 2003', 'Nov 12 2003'] data = pandas.DataFrame(numpy.random.randn(6, 3), index=index, columns=['D', 'B', 'C']) print Quandl.push(data, code='F32C', name='Test', desc='test', authtoken='xxxxxx') ```
Will print the link to your newly uploaded data.
The IPython notebook is an excellent python environment for interactive data work. Spyder is also a superb IDE for analysis and more numerical work.
I would suggest downloading the data in raw format in the highest frequency possible and preforming any data manipulation in pandas itself.
See this link for more information about timeseries in pandas.
Please send any questions, comments, or any other inquires about this package to Chris@quandl.com.
The stable version of Quandl can be installed with pip:
pip install Quandl
Pandas :: https://code.google.com/p/pandas/
dateutil (should be installed as part of pandas) :: http://labix.org/python-dateutil