Escape from spreadsheet models!
- 4 June 2019: modelx v0.0.22 is released. See modelx v0.0.22 (4 June 2019) release notes for details.
- 24 March 2019: spyder-modelx v0.0.9 is released. See Release Notes for details.
- 24 March 2019: modelx v0.0.21 is released. See Release Notes for details.
- 2 February 2019: modelx v0.0.20 is released. See Release Notes for details.
- 13 January 2019: spyder-modelx v0.0.8 is released. See Release Notes for details.
- 13 January 2019: modelx v0.0.19 is released. See Release Notes for details.
- 31 December 2018: spyder-modelx v0.0.7 is released. See Release Notes for details.
- 31 December 2018: modelx v0.0.18 is released. See Release Notes for details.
What is modelx?¶
modelx is a Python package to build object-oriented models containing formulas and values to carry out complex calculations. You can think of it as a hierarchical and multidimensional extension of spreadsheet, but there’s so much more to it!
modelx comes with features that enable users to interactively develop, run and scrutinize complex models in smart ways:
- Only little Python knowledge required
- Model composed of a tree of Spaces containing Cells
- Cells containing formulas and data
- Dynamic name binding for evaluating formulas within a Space
- Space inheritance
- Dynamic parametrized spaces created interactively
- GUI as Spyder plugin (spyder-modelx)
- Cells graph to track cells interdependency (Under development)
- Saving to / loading from files
- Conversion to Pandas objects
- Reading from Excel files
Who is modelx for?¶
modelx is designed to be domain agnostic.
The modelx was created by actuary, and its primary use is to develop actuarial projection models. lifelib (https://lifelib.io) is a library of actuarial models that are built on top of modelx.
However, modelx is intentionally designed to eliminate domain specific features so that potential audience for modelx can be wider than actuaries, whoever needs to develop complex models of any sorts that are too much to deal with by spreadsheets.
How modelx works¶
modelx exposes its API functions and classes such as Model, Space and Cells to its users, and the users build their models from those classes, by defining calculation formulas in the form of Python functions and associating those calculations with Cells objects.
Below is a very simple working example in which following operations are demonstrated:
- a new model is created,
- and in the model, a new space is created,
- and in the space, a new cells is created , which is associated with the Fibonacci series.
from modelx import * model, space = new_model(), new_space() @defcells def fibo(n): if n == 0 or n == 1: return n else: return fibo(n - 1) + fibo(n - 2)
To get a Fibonacci number for, say 10, you can do:
>>> fibo(10) 55 >>> fibo.series n 0 0 1 1 2 1 3 2 4 3 5 5 6 8 7 13 8 21 9 34 10 55 Name: fibo, dtype: int64
Refer to lifelib (https://lifelib.io) fo more complex examples.
Python and modelx¶
Aside from modelx being a Python package and written entirely in Python, modelx utilizes Python in that it lets users define formulas by writing Python functions and converting it to modelx formulas. However, there is a critical difference between how Python functions are interpreted by Python and how modelx formulas are interpreted by modelx.
Python employs lexical scoping, i.e. the namespace in which function code is executed is determined by textual context. The global namespace of a function is the module that the function is defined in. In contrast, the evaluation of modelx formulas is based on dynamic scoping. Each Cells belongs to a space, and the space has associated namespace (a mapping of names to objects). The formula associated with the cells is evaluated in that namespace. So, what module a formula is defined (in the form of a Python function) does not affect the result of formula evaluation. It is what space the cells belongs to that affects the result.
Copyright 2017-2019, Fumito Hamamura
modelx is free software; you can redistribute it and/or modify it under the terms of GNU Lesser General Public License v3 (LGPLv3).
Contributions, productive comments, requests and feedback from the community are always welcome. Information on modelx development is found at Github https://github.com/fumitoh/modelx
modelx is in its early alpha-release stage, and its specifications are subject to changes without consideration on backward compatibility. The source files of you models may need to be modified manually, if there are updates that break backward compatibility in newer versions of modelx.
Likewise, model files saved with one version may not load with a newer version. When updating modelx to a newer version, make sure you rebuild model files saved using older versions of modelx from their source code.
modelx was originally conceived and written by Fumito Hamamura and it was first released in October 2017.
- What’s New
- Spyder plugin
- Technical Guide