latest versions of dependent packages cause problems with MuMoT figures

[joefresna]

When updating all packages, manual configuration of PyDSTool is necessary (because otherwise it won't work with SciPy 1.0.0)
https://stackoverflow.com/questions/48251201/import-of-package-pydstool-doesnt-work
In particular, it's necessary to modify two files:
lib/python3.5/site-packages/PyDSTool/__init__.py (line 75):
change

if vernums[1] < 5:

in

if vernums[0] < 1 and vernums[1] < 5:

and file lib/python3.5/site-packages/PyDSTool/parseUtils.py (line 46-47), change:

'h2vp', 'sph_jn', 'sph_yn', 'sph_jnyn', 'sph_in',
'sph_kn', 'sph_inkn', 'riccati_jn', 'riccati_yn',

to

'h2vp', 'spherical_jn', 'spherical_yn', 'spherical_in',
'spherical_kn', 'riccati_jn', 'riccati_yn',

After the update, the dependent libraries in their last version causes problems to MuMoT figures:

• Plot size is small
• A new figure is drawn each time a widget is changed (instead of updating the same figure)

[joefresna]

Maybe, working with the latest version of libraries, we can solve the issue of the empty figure when a cell is run twice.

[jarmarshall]

Using MumotEnv as described in the readme seems to do the trick for now

[joefresna]

we still have issues of double figures when the same cell is run twice and I think that checking what's wrong with newer packages (where their bugs have been possibly patched) can help us in understanding what we are doing wrong.
Especially ipywidgets is still in active development and bugs are patched daily.

[jarmarshall]

With the latest commit and the MumotEnv.yml I cannot reproduce the double figures issue - please provide a minimal working example.

I think we need to be pragmatic about finishing the first release, and moving on to community development or development funded by other means. To do that we need a stable environment that works for current functionality.

[joefresna]

The double-figure issue still happens to me (and I think Thomas too) using MumotEnv.yml on every cell that produces a plot.
Run any cell, then change test (e.g. a parameter) in that cell and run it again. It generates an empty plot. We tried various time to find the cause of it but without success.

[jarmarshall]

I chose a random cell in the user manual and tried to follow your procedure - I had no issues. Please upload a minimal working example.

[joefresna]

Here
SmallTest-Copy1.ipynb.zip

[jarmarshall]

OK - this must be something internal to Jupyter I think - strangely clearing the cell output from the menu bar, then re-running, still exhibits the problem. I agree about fixing this in the future by allowing updated libraries, but we will have to approach that goal slowly - did you make a pull request with your changes on the PyDSTool repository yet, @joefresna ?

[joefresna]

Not yet but it's on my list.

[tbose1]

The double-plotting of figures when repeatedly executing commands should now be solved with commit 473e975.
I made a little hack in the _buildFig() function. I renamed the old version to _buildFigOLD() which is still in the file, so we can quickly convert back should we encounter problems with the new version of _buildFig(). (though I tested it and it seems to work with all views and sequences of executing and re-executing the plotting commands)

[jarmarshall]

OK to close?

[joefresna]

I realised I don't know how to make a Pull Request to a git-hub repository where I don't have permissions. 😰

Anyhow, I found that changes similar to the one I wanted to propose have been recently refused:
robclewley/pydstool#132

[jarmarshall]

It looks like this pull request makes various Python 3 compatibility changes including the scipy versioning issue, and is still open...
robclewley/pydstool#133
we need to try and get in touch with whoever is most recently active in maintaining the repository...

[jarmarshall]

Also, I think pull request #69 may address this - if not, in there there is info on how to add a specific GitHub repo branch as a dependency

[willfurnass]

Fixed via #338