@@ -19,39 +19,61 @@ Documentation
1919-------------
2020http://raccoon.readthedocs.io/en/latest/
2121
22+ Source location
23+ ~~~~~~~~~~~~~~~
24+ Hosted on GitHub: https://github.com/rsheftel/raccoon
25+
2226Overview
2327--------
2428Raccoon is a lightweight DataFrame and Series implementation inspired by the phenomenal Pandas package for the one use
25- case where Pandas is known to be sub-optimal: DataFrames that grow in size by rows frequently in the code. Additionally
26- Raccoon DataFrames can be parametrized to be sorted so that additions to the DataFrame keep the index in sorted order
27- to speed inserts and retrievals.
29+ case where Pandas is known to be sub-optimal: DataFrames and Series that grow in size by rows frequently in the code.
30+ Additionally Raccoon DataFrames and Series can be parametrized to be sorted so that additions to the DataFrame keep the
31+ index in sorted order to speed inserts and retrievals.
32+
33+ A simple speed comparison of Raccoon vs Pandas for typical functionality is located in the documentation.
2834
2935Inspiration
30- -----------
31- Pandas DataFrames are excellent multi-purpose data structures for data management and analysis. One of the use cases
32- I had was to use DataFrames as a type of in-memory database table. The issue was that this required lots of growing
33- the rows of the DataFrame, something that is known to be slow in Pandas. The reason it is slow in Pandas is that the
34- underlying data structure is numpy which does a complete copy of the data when the size of the array grows.
36+ ~~~~~~~~~~~
37+ Pandas DataFrames and Series are excellent multi-purpose data structures for data management and analysis. One of the
38+ use cases I had was to use DataFrames as a type of in-memory database table. The issue was that this required lots of
39+ growing the rows of the DataFrame, something that is known to be slow in Pandas. The reason it is slow in Pandas is that
40+ the underlying data structure is numpy which does a complete copy of the data when the size of the array grows.
3541
3642Functionality
37- -------------
43+ ~~~~~~~~~~~~~
3844Raccoon implements what is needed to use the DataFrame as an in memory store of index and column data structure
3945supporting simple and tuple indexes to mimic the hierarchical indexes of Pandas. The methods included are primarily
4046about setting values of the data frame, growing and appending the data frame and getting values from the data frame.
4147The raccoon DataFrame is not intended for math operations like pandas and only limited basic math methods are included.
4248
49+ Underlying Data Structure
50+ ~~~~~~~~~~~~~~~~~~~~~~~~~
51+ Raccoon uses the standard built in lists as its default underlying data structure. There is an option on object
52+ construction to use any other drop-in replacement for lists. For example the fast blist package
53+ http://stutzbachenterprises.com/blist/ could be used as a list replacement for the underlying data structure.
54+
4355Why Raccoon?
44- ------------
56+ ~~~~~~~~~~~~
4557According to wikipedia some scientists believe the panda is related to the raccoon
4658
4759Contributing
48- ------------
60+ ~~~~~~~~~~~~
4961Contribution in the form of pull requests are welcome. Use pytest to run the test suite. Be sure any new additions
5062come with accompanying tests.
5163
5264Future
53- ------
65+ ~~~~~~
5466This package serves the needs it was originally created for. Any future additions by myself will be driven by my own
5567needs, but it is completely open source to I encourage anyone to add on and expand.
5668
5769My hope is that one day Pandas solves the speed problem with growing DataFrames and this package becomes obsolete.
70+
71+ Python Version
72+ ~~~~~~~~~~~~~~
73+ Raccoon requires Python 3.4 or greater. Python 2.7 support was eliminated as of version 3.0. If you need to use raccoon
74+ with Python 2.7 use any version less than 3.0
75+
76+ Helper scripts
77+ ~~~~~~~~~~~~~~
78+ There is helper function to generate these docs from the source code. On windows cd into the docs directory and
79+ execute make_docs.bat from the command line. To run the test coverage report run the coverage.sh script.
0 commit comments