Leontief is a C++/Python toolkit to facilitate working with economic Input-Output models.
The overall design principle behind Leontief is to strike a balance between usability and performance when working with and/or developing algorithms and tools related to economic input-output systems. A key such design decision is to use a user-friendly C++ library for vector, matrix, tensor manipulations that offers a higher-level API similar e.g., to that of Matlab.
Leontief helps with a number of typical tasks in Environmentally Extended Input-Output modeling:
- Pre-processing Input-Output tables using command line tools and ingesting IO matrices into a C++ environment for further calculations.
- Solving typical EEIO problems using a library of tested algorithms.
- Investigating ad-hoc research questions by writing new high-level code in the form scripts. In particular emphasize graph-oriented tools and calculations
The current focus of development is around utilizing the FIGARO dataset but all open databases will be integrated in due course.
- Clone the repository into your own work environment.
- The command line scripts included in the leontief/scripts directory should be working immediately on a standard linux distribution (or Windows WSL).
A major part of Leontief is a specialized EEIO C++ library that leverages other lower level C++ libraries. To augment or rebuild the library from source you will need to have first a working C++ toolchain that can compile modern C++.
Conan is recommended as a means to pull in dependencies but is not absolutely required. The current dependency list includes:
- The eigen library providing vector,matrix,tensor containers and linear algebra algorithms
- The poco library providing general purpose facilities
- The catch2 library for testing
For compiling the C++ code, Leontief uses cmake.
There is a suit of catch2 based tests that help verify that everything is installed and running properly.
- pip install conan
- conan install . --output-folder=cmake-build-debug --build=missing -s build_type=Debug
Create a leontief/data subdirectory. Download, move and extract publicly available IO databases into the data directory. Current download scripts:
- exiobase
- figaro
- oecd-icio
The focus of Leontief is on high performance numerical calculations involving large matrices. It does not offer dataframe type functionality for searching or filtering data, this is more economically achieved using Python. Yet some preprocessing is required. The following scripts offer needed functionality:
- Strip the matrix files (e.g., bash strip_exiobase.sh) from label data (we will not need those for numerical C++ calculations).
- TBC
TODO
TODO
- Working with CLI tools to process large matrices is elaborated further on the eLearning platform Open Risk Academy