Division by zero in convergence test if yMax == yMin,

[arurke]

In analysis_metric(), if all elements in "y-axis" data are the same (i.e. min and max are identical), it leads to a division by zero in the convergence test, see https://github.com/romain-jacob/triscale/blob/master/helpers.py#L66 and two lines above.

The issue can easily be reproduced:

x = np.arange(0, 100)
y_same_value = np.full(len(x), 100)
df = pd.DataFrame(
    {'x': x,
     'y': y_same_value})

triscale.analysis_metric( 
    df,
    metric = {'measure': 50},
    convergence = {'expected': True})

>  triscale/helpers.py:66: RuntimeWarning: invalid value encountered in true_divide

An intuitive solution is to simply state that the data is converged in such cases with identical elements, but perhaps I am missing something about the statistics so I'll leave the PR to someone else 😅

[romain-jacob]

Thanks for reporting!

Indeed, if all elements are the same, the empirical evidence for "has converged" is as strong as it gets :-) I'll add a test case for that.

[arurke]

On further inspection, the problem occurs for metrics only if:

1. All data is equal AND bounds are not set, or
2. Bounds are set by user to equal values (which raises other questions)

The issue is present for all invocations of convergence_test(). E.g. for KPIs it occurs if all values are equal (as bounds are always set to max(), min()).