plot_repr – Transform test results into PlotTemplate

Module containing all available methods to convert a test result in a table to be converted in rst.

valjean.javert.plot_repr.dimensions_from_array(array_shape)

Check if array is consistent with 1D plot.

Checks are done on presence of non-trivial and trivial dimensions, trivial being a dimension with size equal to one, non-trivial being for a dimension with at least two values.

Parameters:

array_shape (tuple(int)) – shape of the values array

Returns:

indices of the non-trivial dimensions in the shape tuple

Return type:

tuple(int)

valjean.javert.plot_repr.dimensions_and_bins(bins, array_shape)

Determine the dimensions of the result from the collections.OrderedDict of bins. It is expected to be the “only non-trivial” dimensions.

Parameters:

• bins (collections.OrderedDict(str, numpy.ndarray)) – bins coming from the results dataset

• array_shape (tuple(int)) – shape of the values array from the results dataset

Returns:

dimensions and bins to be used (non-trivial ones), or None if all the dimensions are trivial

Return type:

collections.OrderedDict

valjean.javert.plot_repr.trim_range(bins, max_ratio=1000)

Adapt bins range when extreme bins are very large.

This function suggests reasonable ranges for the given bin axes by trimming the extreme bins if they are too wide. For each axis, the extreme bins are trimmed if their width is at least max_ratio times larger than the width of the neighbouring bin (the first bin is compared to the second one, and the last bin is compared to the second-last one). If there is no need to change the previous limits a tuple with initial limits is returned for the considered dimension. A boolean is associated to the tuple to precise if the limits have be changed.

The trimmed ranges extend a bit over the clipped bins, so that their content is still visible in the plots.

Parameters:

• bins (list(numpy.ndarray)) – bins of the PlotTemplate.

• max_ratio (float) – the bin size ratio above which bins will be truncated.

Returns:

new limits for all dimensions. The list has one item per dimension; the bool indicates whether the axis was actually trimmed.

Return type:

list(tuple(float, float, bool))

Examples:

>>> bins = [np.array([-1e4, 0, 2, 4, 1e4]),
...         np.array([-1e2, 0, 2, 4, 1e4]),
...         np.array([-1e2, 0, 2, 4, 1e2])]
>>> trim_range(bins)
[(-0.2, 4.2, True), (-100.0, 9.2, True), (-100.0, 100.0, False)]

valjean.javert.plot_repr.fit_curve_ranges(curves, threshold=0.0)

Return a set of best-fit limits for the given curves.

This function suggests axis limits for the given curves that make all curves fill the plot area. The limits are determined by looking at the region where any of the curves exceed the given threshold.

Parameters:

• curves (list(CurveElements)) – list of curves

• threshold (float) – threshold for the detection

Returns:

suggested limits for all dimensions

Return type:

list((float, float))

Example:

>>> bins = [np.array([-1.0, 0.0, 1.0]),
...         np.array([0.0, 5.0, 10.0, 15.0])]
>>> curve1 = CurveElements(np.array([[1.0, 2.0, 3.0],
...                                  [2.0, 3.0, 4.0]]),
...                       bins=bins, legend='curve1')
>>> curve2 = CurveElements(np.array([[1.0, 4.0, 3.0],
...                                  [2.0, 3.0, 2.0]]),
...                       bins=bins, legend='curve2')
>>> fit_curve_ranges([curve1, curve2], threshold = 3.5)
[(-1.0, 1.0), (5.0, 15.0)]

valjean.javert.plot_repr.ranges_union(ranges, union_axis=0, minmax_axis=- 1)

Return the union of the given ranges, axis by axis.

Parameters:

• ranges (list or numpy.ndarray) – an n-dimensional array of range bounds. The array shape can be anything, as long as one of the dimensions has length 2; this axis is assumed to contain the (min_range, max_range) pair.

• union_axis (anything that NumPy understands as an axis.) – the axis along which the union will be computed. If unspecified, the first axis will be assumed.

• minmax_axis (int) – the length-2 axis along which the range minimum and maximum are stored. If unspecified, the last axis will be assumed.

Returns:

an array of union bounds, with the same shape as the input array except for the suppression of union_axis.

Return type:

numpy.ndarray

Examples:

>>> ranges_union([[(-1, 1)]])
array([[-1,  1]])

>>> ranges_union([[(-5, 5)], [(-3, 8)]])
array([[-5,  8]])

>>> ranges_union([[(-5, 5), (-1, 1)],
...               [(-3, 8), (-2, 2)]])
array([[-5,  8],
       [-2,  2]])

>>> ranges_union([[(-5, 5), (-1, 1)],
...               [(-3, 8), (-2, 2)],
...               [(-7, 1), (0, 6)]])
array([[-7,  8],
       [-2,  6]])

valjean.javert.plot_repr.curve_limits(curve, threshold=0.0)

Return the set of bounds over each axis where the given curve exceeds the given threshold.

For example, the following curve exceeds the threshold value of 50 in the central bins, which corresponds to 2.0<x<4.0 and -0.5<y<0.5:

>>> bins = [np.linspace(0.0, 6.0, num=7),
...         np.linspace(-2.5, 2.5, num=6)]
>>> curve = CurveElements(np.array([[  1,   1,   1,   1,   1],
...                                 [  1,  10,  30,  10,   1],
...                                 [  1,  10, 100,  10,   1],
...                                 [  1,  10, 100,  10,   1],
...                                 [  1,  10,  30,  10,   1],
...                                 [  1,   1,   1,   1,   1]]),
...                      bins=bins, legend='curve')
>>> curve_limits(curve, threshold=50)
[(2.0, 4.0), (-0.5, 0.5)]

Changing the threshold to 20 modifies the range for the first axis, but not for the second one.

>>> curve_limits(curve, threshold=20)
[(1.0, 5.0), (-0.5, 0.5)]

The function also works if the number of bin edges is equal to the number of values along a given direction (as opposed to the number of values plus one):

>>> bins = [np.array([-1.0, 1.0]), np.array([0.0, 5.0, 10.0, 15.0])]
>>> curve = CurveElements(np.array([[1.0, 2.0, 3.0],
...                                 [2.0, 3.0, 4.0]]),
...                      bins=bins, legend='curve')
>>> curve_limits(curve, threshold = 3.5)
[(1.0, 1.0), (10.0, 15.0)]

valjean.javert.plot_repr.pad_range(limits, log, padding=0.05)

Pad the given limits.

This function adds a bit of padding to the input limit range. If log is False, the new limits will be (limits[0] - padding * delta, limits[1] + padding * delta), where delta = limits[1] - limits[0]. In log scale (log=True), the padded limits are (limits[0] * exp(-padding * log_delta), limits[1] * exp(padding * log_delta), with log_delta = log(limits[1] / limits[0]).

Parameters:

• limits ((float, float)) – a pair of floats.

• log (bool) – whether we are in log scale.

• padding (float) – the amount of padding to insert.

Returns:

new limits

Return type:

(float, float)

valjean.javert.plot_repr.post_treatment(templates, result)

Post-treatment of plots after template generate.

For example add names from test result if not done before, suppress zero bins at range edges, etc.

valjean.javert.plot_repr.build_plot_template_with_dim(curves, axnames)

Organise curves in SubPlotElements.

If curves are 1D they will all go in the same SubPlotElements, if there are in 2 dimensions each curve will have its SubPlotElements.

Parameters:

• curves (list(CurveElements)) – list of curves

• axnames (list(str)) – list of axis names

Return type:

list(SubPlotElements)

valjean.javert.plot_repr.repr_datasets_values(result)

Representation of the datasets values from test results obtained from a child test of TestDataset.

Examples:

• TestResultEqual;

• TestResultApproxEqual;

• TestResultStudent.

Parameters:

result (TestResult) – test result from test on datasets

Return type:

list(PlotTemplate)

If the dimension cannot be determined or if the dimension is greater than 2 an empty list is returned.

valjean.javert.plot_repr.repr_testresultequal(result, _verbosity=Verbosity.DEFAULT)

Represent the equal test result as a plot.

Parameters:

result (TestResultEqual) – a test result.

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_testresultapproxequal(result, _verbosity=Verbosity.DEFAULT)

Represent the approx equal test result as a plot.

Parameters:

result (TestResultApproxEqual) – a test result.

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_testresultstudent(result, verbosity=Verbosity.DEFAULT)

Plot the Student results according to verbosity.

Parameters:

result (TestResultStudent) – a test result.

Returns:

Representation of a TestResultStudent as a plot.

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_student_intermediate(result)

Represent the Student test result as a plot.

By default two PlotTemplate are returned in order to get a top plot representing the two series of values and the bottom plot representing Student’s t-statistic.

Parameters:

result (TestResultStudent) – a test result.

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_student_full_details(result)

Represent the Student test result as a plot.

By default two PlotTemplate are returned in order to get a top plot representing the two series of values and the bottom plot representing Student’s t-statistic.

Parameters:

result (TestResultStudent) – a test result.

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_student_tstud(result)

Represent the t distribution from a Student test result as a plot.

Parameters:

result (TestResultStudent) – a test result.

Return type:

list(PlotTemplate)

Note

if we have a member units in dataset the axis names would be constructed like name + units['name'].

If the dimension cannot be determined an empty list is returned.

valjean.javert.plot_repr.repr_student_pvalues(result)

Representation of p-values from Student test result as a plot.

Parameters:

result (TestResultStudent) – a test result.

Return type:

list(PlotTemplate)

If p-values were not calculated, no PlotTemplate is built, so an empty list is returned.

valjean.javert.plot_repr.repr_testresultmetadata(_result, _verbosity=Verbosity.DEFAULT)

Plot metadata test -> no plot done.

Returns:

empty list

valjean.javert.plot_repr.repr_testresultexternal(_result, _verbosity=Verbosity.DEFAULT)

Plot external test -> no plot done.

If plots are required they are already done. Their representation in the report is done by ExternalRepresenter.

Returns:

empty list

valjean.javert.plot_repr.repr_testresultfailed(_result, _verbosity=Verbosity.DEFAULT)

Plot failed test -> no plot done.

Returns:

empty list

valjean.javert.plot_repr.repr_testresultstats(result, status_ok, label)

Represent result from statistical test as a plot (pie chart probably).

Null results are omitted from the table.

Parameters:

result (TestResult) – result from a statistical test

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_testresultstatstasks(result, _verbosity=Verbosity.DEFAULT)

Represent result from statistical test on tasks as a plot (pie chart).

Parameters:

result (TestResultStatsTasks) – result from tasks statistics

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_testresultstatstests(result, _verbosity=Verbosity.DEFAULT)

Represent result from statistical test on tests as a plot (pie chart).

Parameters:

result (TestResultStatsTests) – result from tasks statistics

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_statstestsby2labels(result)

Plot statistics on the tests classified by 2 labels.

Caution: the tests are summed over the second label, only the first one will be written on the plot. This is for summaries.

Return type:

list(PlotTemplate)

valjean.javert.plot_repr.repr_testresultstatstestsbylabels(result, _verbosity=Verbosity.DEFAULT)

Plot statistics on the tests classified by labels if 1 or labels.

Return type:

list(PlotTemplate)