Current implementation does not allow for easy varying of system size while keeping initial condition proportions constant

This should also enable additional functionality, in particular:

• plotting difference of two state variables
• displaying bifurcation view as part of part of multi view controller (issue #13) - internal PyDStool plotting does not seem well-behaved and does not plot to current axes

The user can select (through widget) the color to assign to each state (also called reactant).
The color will be used in the plots (e.g., network graph, temporal dynamics)

_fig_formatting_2D and _fig_formatting_3D to look up and redner LaTeX variable names correctly

For example, add possibility to visualise network, temporal dynamics and final distribution after a multiagent simulation.
Possible organisation of the figures in tabs.

To enable variable-population models, and constant background flows

Data is generated but it is not (yet) visualised.
Possible visualisations (to be decided):

• histograms [+ variance bars when it is result from multiple runs]
• points [+ boxplots when it is result from multiple runs]

Safari issue, or jupyter notebook version issue, possibly? - wanted to check if others see this; e.g. for this command:

mmt.MuMoTdefault.setTimeDefaults(initTime=10, step=1)
net = model1.multiagent()

running at 2560 x 1600 widgets past Particle speed cannot be seen / reached

Overlay results from various description levels in one single plot.

Should be capturing to log file, or using error widget

Parameters reported in logs lag behind latest versions plotted, according to widget values.

The PyDStool-based bifurcation function is not analysing the Pais model as expected. To test:

• run notebook MuMoT_PyDStest.ipynb
• in widgets set Delta = 0, mu = +ve
• expected pitchfork bifurcation not detected

Similar to Mathematica Manipulate, print out a string describing the parameterisation of the model, suitable for cut and paste to create a standalone view

Controllers can be parameterised with these, so they just need to be chosen appropriately.

MA and SSA views/controllers manage figure with bugs.
For example, running two times
net = model1.multiagent()
will have the widgets interfering between different cells.

The views must use the standard functions: _buildFig and _fig_formatting_2D

In views, etc.

In particular, for setting system size check all reactants appear in assignment, with negative sign, and system size parameter sign is positive

There is the possibility to download the simulation data file (e.g. through MuMoTmultiagentController.downloadTimeEvolution()), however the current data format is very basic and not properly formatted and should be improved.

Views currently assume presence of a controller to set their parameter values. Expert users should be able to directly instantiate view objects, setting parameter values on creation, without going via a controller.

Sometimes when moving sliders a second figure window is generated below the first. Moving the slider again the second window is updated correctly. However, this only happens for some examples and not for others.

Currently views on models (bifurcation plots are the only current view) are integrated with models; these will be separated out so multiple independent views on the same model are possible

E.g. stream (and all other functions) need to accept as arguments not only
'Psi_{1}'
but also
'\Psi_{1}'
and
'\Psi_1'

Either by analytic solution for fixed points, or numerical integration. Trying the former before the latter may introduce unacceptable delays for intractable systems of equations.

Otherwise, the axes are adapted on data

Vector and stream plots need to display an error message and exit before constructing widgets, when system size has not been set

Causes strange behaviour, as represented by showODEs()

Under some Python / Jupyter installations, vector length in 3d vector views appears fixed and uniform.

At the moment the multiagent particle simulation have periodic boundary conditions (torus).
Add walls and either:

• let the particle bounce on the walls (as molecules do)
• let the particle get stuck to the wall (as simple robots do)

For bifurcation parameter at least - unable to check LaTeX state variables due to issue #36

bifurcation view will not work with variable names having underscores, due to _ and {} characters breaking internal PyDS variable naming rules

Suggest a simple function to hash all variable names coming from models to pure alphabetic strings, for passing to PyDS

if model2 is derived from model1 via a call to substitute, then model1.show() can show reactions whose rates reflect the substitution made to create model2

When only visualisation options are changed (e.g. "show communication links" in multiagent sim.), MuMoT should replot without re-computing all data

In places the code uses error classes like SyntaxError, in others custom error classes, in places assert, in other places print and return

The HTML widget for displaying error messages interactively is not always displayed, and is not updated with new error messages

Add "Play", "Pause", "One-Step" buttons to start, pause and advance 1-step at a time the simulation

In particular, whitespace should not be required for a valid model specification

At the moment MA and SSA views use a custom _getStoichiometry function.
For consistency, in the project, we should use a single Stoichiometry function.

Too many issues to list at the moment! But the first critical one is to complete a proof-of-concept for interactive bifurcation analysis of a described model using PyDSTool. The widgets update the parameters of the model but the bifurcation analysis raises an exception for reasons that are not currently clear.

E.g. in showODEs(), showReactants()

Add smart parsing of input string.
Add generation of input string from simulation output (to be reused with copy/paste for next command)

2d and 3d field-based plots (stream and vector plots) should not display / compute areas outside the simplex

Plotting a 3+d system when system size has not been set, but system size is non-constant, requires error handling

however, substituting not by an expression but by another symbol, for example U = X would work

3d stream plot will need to be implemented by hand, by numerically integrating the equations from selected starting points, as I have not found a library implementation so far

Widgets currently constructed in arbitrary order, which will be confusing for users; sort alphanumerically