Persistent Objectified Indexed Data
Project description
Persistent Objectification of Indexed Data
Trump is a framework for objectifying data, with the goal of centralizing the responsibility of managing feeds, munging, calculating and validating data, upstream of any application or user requirement.
With a focus on business processes, Trump’s long run goals enable data feeds to be:
Prioritized, flexibly - a symbol can be associated with multiple data source for a variety of reasons including redundancy, calculations, or optionality.
Modified, reliably - a symbol’s data feeds can be changed out, without any changes requiring testing to the downstream application or user.
Verified, systematically - a variety of common data processing checks are performed as the symbol’s data is cached.
Audited, quickly - alerts and reports all become possible to assess integrity or inspect where manual over-rides have been performed.
Aggregated, intelligently - on a symbol by symbol basis, feeds can be combined and used in an extensible number of ways.
Customized, dynamically - extensibility is possible at the templating, munging, aggregation, and validity steps.
Planning
See docs/planning.md for the direction of the project.
Basic Usage
This example dramatically understates the utility of Trump’s long term feature set.
Adding a Symbol
from trump.orm import SymbolManager
from trump.templating import QuandlFT, GoogleFinanceFT, YahooFinanceFT
sm = SymbolManager()
TSLA = sm.create(name = "TSLA",
description = "Tesla Closing Price USD")
TSLA.add_tags(["stocks","US"])
#Try Google First
#If Google's feed has a problem, try Quandl's backup
#If all else fails, use Yahoo's data...
TSLA.add_feed(GoogleFinanceFT("TSLA"))
TSLA.add_feed(QuandlFT("GOOG/NASDAQ_TSLA",fieldname='Close'))
TSLA.add_feed(YahooFinanceFT("TSLA"))
#Optional munging, validity checks and aggregation settings would be
#implemented here...
#All three feeds are cached...
TSLA.cache()
#But only a clean version of the data is served up...
print TSLA.df.tail()
TSLA
dateindex
2015-03-20 198.08
2015-03-23 199.63
2015-03-24 201.72
2015-03-25 194.30
2015-03-26 190.40
sm.finish()
Using a Symbol
from trump.orm import SymbolManager
sm = SymbolManager()
TSLA = sm.get("TSLA")
#optional
TSLA.cache()
print TSLA.df.tail()
TSLA
dateindex
2015-03-20 198.08
2015-03-23 199.63
2015-03-24 201.72
2015-03-25 194.30
2015-03-26 190.40
sm.finish()
Installation
See the latest Installation instructions on ReadTheDocs.org
Requirements
Python 2.7; Support for Python 3.3 or 3.4 is do-able, if there is demand.
A Relational Database Supported by SQLAlchemy should work, however the following is tested: * PostgreSQL 9.4 * Persistent SQLite (ie, file-based). Certain features of Trump, wouldn’t make sense with an in-memory implementation)
Dependencies
Pandas (Tested with >= 15.2)
SQLAlchemy (Tested with >= 0.9)
Smuggle (Tested with >= 0.2.0)
Data Source Dependencies
Documentation
Read the latest on ReadTheDocs.org
Communication
Questions, Bugs, Ideas, Requests or just say “Hi” -> GitHub Issues, InvTech@equitable.ca, or jeffrey.mclarty@gmail.com
Contribute Code -> New Branch + GitHub Pull Request
Chat -> Gitter
License
BSD-3 clause. See the actual License.
Background
The prototype for Trump was built at Equitable Life of Canada in 2014 by Jeffrey McLarty, CFA and Derek Vinke, CFA. Jeffrey McLarty currently leads the Open Source initiative.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.