Empirical is a language for time-series analysis. It has statically typed Dataframes with integrated queries and builtin timestamps.
Read CSV files and infer their types (at compile time!):
>>> let trades = load("trades.csv"), quotes = load("quotes.csv")
Compute a one-minute volume-weighted average price (VWAP):
>>> from trades select vwap = wavg(size, price) by symbol, bar(timestamp, 1m)
Get the most recent quote for each trade:
>>> join trades, quotes on symbol asof timestamp
Get the closest event within three seconds:
>>> join trades, events on symbol asof timestamp nearest within 3s
See the website for more examples and full documentation.
The build is handled by CMake. There are some scripts to help with the process, both for POSIX (*.sh) and Windows (*.bat). These can be called by make or make.bat for an easy workflow.
| POSIX | Windows | |
|---|---|---|
| debug build | make |
make.bat |
| production build | make prod |
make.bat prod |
| run tests | make test |
make.bat test |
| deploy binary | make deploy |
make.bat deploy |
| clean build | make clean |
make.bat clean |
Additionally, POSIX has make website.
While the actual code is all C++, there are code generators that need the JVM (ANTLR) and Python (ASDL). These generators are included in this repo, so there is nothing to download.
The build is statically linked, so no libraries are needed to distribute the resulting binary.
The code for Empirical is structured as follows:
src/- source code for Empiricaltests/- regression testsdoc/- documentationcmake/- CMake filesscripts/- build scriptsthirdparty/- open-source dependencies
See the Contributor's Notes. There is also the Maintainer's Notes for the curious.
Affero General Public License (AGPL) with the Commons Clause.
A proprietary license is available to vendors and customers that need more commercial-friendly terms.