penny 0.3.4

Inspect csv files and lists of data

Penny
=====

Inspect csv files and lists of data to find out the truth!
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

.. figure:: http://www.martianwatches.com/wp-content/uploads/2013/10/InspectorGadget.jpg
:alt: alt tag

alt tag
Uncle Gadget was great and all, but when it came to real detective work,
we all know Penny did the heavy lifting. Hence, Penny, the Python module
that inspects stuff. Feed it rows or columns from a dataset, and get
information about the column types -- including whether or not a given
column represents a category or date. Penny also finds column headers
(waaaay more reliably than the ``Sniffer`` class in to the standard
``csv`` module).

Why?
~~~~

If you're working with a few datasets, it's easy to figure out which
columns are supposed to be dates, integers and even categories just by
looking at the raw csv files. But if you need to programmatically deal
with lots of datasets, this gets tedious fast.

Setup
~~~~~

Grab the package.

::

pip install penny

Or grab the code from GitHub.

::

git clone https://github.com/gati/penny
cd penny
pip install -r requirements.txt

Getting Started
~~~~~~~~~~~~~~~

Guess the headers of a csv file.

.. code:: python

from penny.headers import get_headers

with open('your-awesome-file.csv') as csvfile:
has_header, headers = get_headers(csvfile)

# Prints True/False depending on whether or not headers were found
print has_header

# Prints column headers or placeholders if real headers weren't found
print headers # ['Example Header A', 'Example Header B']

Guess the data type of a column in your dataset.

.. code:: python

from penny.inspectors import column_types_probabilities

fileobj = open('your-awesome-file.csv')
rows = list(csv.reader(fileobj))

# Get the values from column 0
column_0 = [x[0] for x in rows]
probs = column_types_probabilities(column_0)

# Prints something like {'date': 1, 'int': .75, 'category': 0 ...}
print probs

Or get type guesses for all the rows in your dataset at once.

.. code:: python

from penny.inspectors import rows_types_probabilities

fileobj = open('your-awesome-file.csv')
rows = list(csv.reader(fileobj))
probs = rows_types_probabilities(rows)

Penny checks for a lot of data "types," not just the standard ``int``,
``str``, etc. Here's the list (for now):

- **date** something ``dateutil.parser`` can parse into a ``datetime``
object
- **int** a whole number
- **bool** y/n or yes/no or something true/falsey
- **float** a number with a decimal
- **category** something you might want to group records by
- **id** unique for each row
- **coord** a float that might be a latitude or longitude
- **coord\_pair** string that looks like "coord,coord"
- **city** one of the world's 80,000 largest cities
- **region** smaller than a country, bigger than a city. state,
province, etc
- **country** a country name on the `ISO 3166
list <http://en.wikipedia.org/wiki/ISO_3166-1#Current_codes>`__
- **address** a full address you could geocode with a service like
Google Maps

Last but not least, you can also inspect a column for a single type.

.. code:: python

from penny.list_check import column_probability_for_type

fileobj = open('your-awesome-file.csv')
rows = list(csv.reader(fileobj))

# Get the values from column 0
column_0 = [x[0] for x in rows]
prob = column_probability_for_type(column_0, 'date')

# Prints something like 0.78
print prob

Contributing & Credits
~~~~~~~~~~~~~~~~~~~~~~

This is a work in progress, so pull request at will. Some of this work
was inspired by `messytables <https://github.com/okfn/messytables>`__,
which looks great for xls files but wasn't quite what I needed. Thanks
to `Chris Albon <http://twitter.com/chrisalbon>`__ for putting together
a `repo of useful test
datasets <https://github.com/chrisalbon/Variable-Type-Identification-Test-Datasets>`__.

Questions, concerns, devoted fan mail to
[@jonathonmorgan](http://twitter.com/jonathonmorgan) on Twitter.