modelx v0.1.0 (1 December 2019)#

This release introduces major spec changes and enhancements, as well as a few minor bug fixes.


Input and calculated values

A Cells holds values either input by the user, or calculated by its formula. The former are now called input values, and distinguished from tha latter:

def foo(x): return x * foo(x-1)

foo[0] = 5
foo[5]   # Calculated buy the formula

foo[0] in this example holds an input value of 5, while foo[5] holds a calculated value of 120. To check whether a node of a Cells has an input value, is_input() is now available.:

>>> foo.node(0).is_input()
>>> foo.node(5).is_input()

Change in Cells.clear and introduction of Cells.clear_all and Cells.clear_at

With the introduction of the input value concept, Cells.clear() method now only clears calculated values, but not input values. Cells.clear_all() method is introduced for clearing both input and calculated values. Another new method Cells.clear_at() is introduced for clearing the value of a specific Cells node, whether it’s an input value or a calculated value.

Recalculation upon input change

Before version 0.1.0, when a new input value is assigned to a node of a Cells, the values of the nodes depending on the input cells are cleared. From version 0.1.0, those values are recalculated. Continuing the previous example:

>>> foo.series
0      1
1      1
2      2
3      6
4     24
5    120
Name: foo, dtype: int64

>>> foo[0] = 2
>>> foo.series
0      2
1      2
2      4
3     12
4     48
5    240
Name: foo, dtype: int64

Models available as modelx attirbutes

Only available with Python 3.7 or newer

In addition to get_models(), models can also be acquired through modelx attribute access:

>>> m1 = mx.get_models["Model1"]
>>> m2 = mx.Model1
>>> m1 is m2

defcells and new_space creates new model/space if no current model/space is available

Previously, if the current model and/or the current space don’t exist, the user needs to create a model and/or a space manually. From version 0.1.0, if the current model and/or the current space are missing when calling defcells() and new_space() API decorator/functions, a new model and/or a new space are created automatically, and are assigned to the current model and/or the current space, so the user does not need to create them explicitly beforehand.

New serializer for reading/writing models

Serializer is an internal part of modelx to implement the functionality of reading models from and writing models to files in a human-readable format. The initial version of modelx serializer (version 1) was introduced in modelx v0.0.22. The version 1 serializer had limited functionality.

modelx v0.1.0 comes with almost completely rewritten serializer (version 2). The version 2 serializer reflects the following improvements.

  • Output files with “.py” extension are syntactically correct Python scripts, and also importable.

  • The new format of the output files is more readable.

  • When writing a model to an existing directory, the existing directory is renamed with the prefix of “_BAK1”.

  • Any picklable objects as references can be stored in binary data files.

  • Input values of Cells are also stored. Input values can be any picklable objects.

  • Object identities among the stored picklable objects are preserved.

read_model() of modelx v0.1.0 still supports reading files written in the version 1 format.

Other enhancements

  • defcells() can now be used to update existing cells formulas, in addition to creating a new cells.

  • maxlen parameter is added to start_stacktrace(), to specify the maximum length of traces to be kept (GH13).

Backward Incompatible Changes#

Bug Fixes#

  • cur_space() is set to None when the current model is deleted.

  • The current space attribute of a models is now saved and restored by save() method and open_model().

  • Fix the error raised from a package supporting IPython’s auto complete feature when used in Windows command prompt console.

  • Fix the error raised when attempting to delete refs by the del statement.

  • Fix missing source file info from restored cells.

  • Fix crash with Python 3.8 on Windows due to stack overflow.