egentry / clustered_sne Goto Github PK

If you restart a run which already has had SNe occur, but with no *_overview.dat file present, then the *_overview.dat file might list too few SNe under the field "Number of SNe"

In particular, I've restarted an 11 SNe run after 1 SNe has gone off, and the *_overview.dat is created saying:
Number of SNe: 10
which breaks my RunSummary (python object) constructor.

(Error message at bottom. First noticed by Ben Wibking)

Things work fine when I run on an up-to-date Python 3.5 installation, but fail when I run with a Python 2.7 installation and the default Anaconda 4.0.0 packages.

TypeError                                 Traceback (most recent call last)
/Users/egentry/Dropbox/data/clustered_SNe/analysis/../../clustered_SNe/analysis/visualize_helpers.pyc in single_run(data_dir, id)
    241         raise FileNotFoundError("No directory found named: "+ data_dir)
    242 
--> 243     run_summary = RunSummary(data_dir=data_dir, id=id)
    244     sedov_solution = SedovSolution(E_0,
    245                                    run_summary.overview.background_density,

/Users/egentry/Dropbox/data/clustered_SNe/analysis/../../clustered_SNe/analysis/parse.pyc in __init__(self, data_dir, id)
    169             M_tot[k]    = df_tmp.M_int.iloc[-1]
    170             Z_tot[k]    = np.sum(df_tmp.Mass * df_tmp.Z)
--> 171             zones[k]    = df_tmp.shape[0]
    172 
    173             over_dense = np.where(df_tmp.Density > overview.background_density * 1.0001)[0]

TypeError: 'int' object does not support item assignment`

In version 6.0.0 of ipywidgets, my single_run function no longer works properly with ipywidgets interactive functionality. This can most clearly be seen in my analysis/visualize.ipynb notebook.

Previously I had a hackish solution for first choosing a simulation and loading it interactively, and then plotting a specific view interactively. This shouldn't be done in a single interact function, because I don't need to reparse the entire simulation every time I want to change my axis limits. So I had nested interact calls, which would try to clear previous widgets whenever I chose a new simulation. Unfortunately interact gives a slightly different output now, which breaks my previous assumptions. My fault for designing things in a way that they didn't recommend.

For now, you can get around this by simple re-running the entire cell that contains the interact(single_run) statement. This should clear all previous widgets, at the cost of losing all the custom options you've configured.

For a medium-term solution, I'm guessing you could fix single_run to point at the correct element in children, and clear that instead (possibly breaking backwards compatibility w.r.t. ipywidgets.

For a long-term solution, you should design widgets that actually interact with each other in the way that ipywidgets assumed. This is fairly complex for the datatypes I want to interact with, so I don't plan to implement this fix.