def build_summary_stats_json(report):
"""Creates a dict from the summary statistics in the report.
Parameters
----------
report: :py:class:`solarforecastarbiter.datamodel.Report`
Returns
-------
str
The json representing the summary statistics. Will be a string
representing an empty json array if the report does not have a
computed raw_report.
Raises
------
ValueError
If report.raw_report is populated but no
report.raw_report.metrics have `is_summary == True`
indicating that the report was made without
summary statistics.
"""
if getattr(report, 'raw_report') is not None:
df = plotly_figures.construct_metrics_dataframe(
list(
filter(lambda x: getattr(x, 'is_summary', False),
report.raw_report.metrics)),
rename=plotly_figures.abbreviate)
if df.empty:
raise ValueError('No summary statistics in report.')
return df.to_json(orient="records")
else:
return "[]"
def test_construct_metrics_dataframe_with_rename(report_with_raw):
metrics = report_with_raw.raw_report.metrics
df = figures.construct_metrics_dataframe(metrics,
rename=figures.abbreviate)
report_params = report_with_raw.report_parameters
original_names = [
fxobs.forecast.name for fxobs in report_params.object_pairs
]
abbreviated = list(map(figures.abbreviate, original_names))
assert np.all(df['abbrev'] == np.repeat(
np.array(abbreviated, dtype=object),
len(report_params.metrics) * len(report_params.categories)))
def build_metrics_json(report):
"""Creates a dict from the metrics results in the report.
Parameters
----------
report: :py:class:`solarforecastarbiter.datamodel.Report`
Returns
-------
str
The json representing the report metrics. The string will be a string
representing an empty json array if the report does not have a
computed raw_report.
"""
if getattr(report, 'raw_report') is not None:
df = plotly_figures.construct_metrics_dataframe(
report.raw_report.metrics, rename=plotly_figures.abbreviate)
return df.to_json(orient="records")
else:
return "[]"
def test_construct_metrics_dataframe(report_with_raw):
report = report_with_raw
metrics = report.raw_report.metrics
df = figures.construct_metrics_dataframe(metrics)
names = df['name']
abbrev = df['abbrev']
categories = df['category']
metrics = df['metric']
values = df['value']
report_params = report.report_parameters
expected_length = (len(report_params.metrics) *
len(report_params.categories) *
len(report_params.object_pairs))
assert all([len(v) == expected_length for k, v in df.items()])
original_names = [
fxobs.forecast.name for fxobs in report_params.object_pairs
]
assert np.all(names == np.repeat(
np.array(original_names),
len(report_params.metrics) * len(report_params.categories)))
assert np.all(names == abbrev)
assert np.all(metrics == np.tile(
np.repeat(np.array(report_params.metrics, dtype=object),
len(report_params.categories)),
len(report_params.object_pairs)))
assert np.all(categories == np.tile(
np.array(report_params.categories),
len(report_params.metrics) * len(report_params.object_pairs)))
# this could maybe use value variance, but asserting the dataframe process
# did not mangle values for now
assert (values == 2).all()
def test_construct_metric_dataframe_no_values():
# Iterative metrics datafame creation just builds an empty dataframe
# with correct columns if no MetricResults are found in the metrics tuple
df = figures.construct_metrics_dataframe(())
assert df['index'].size == 0
assert 'abbrev' in df