def update_plots(self):
if self.data.empty:
# Initialize empty dicts and return
for stat in ["sigma", "s10", "s2"]:
dict = {
"%s_x" % stat: [],
"is_git_commit": [],
"date": [],
"hash": [],
"author": [],
"message": [],
stat: [],
"nsamples": [],
"accuracy_threshold": [],
"custom_colors": []
}
if stat == "s10" or stat == "s2":
dict["%s_lower_bound" % stat] = []
self.sources["%s_source" % stat].data = dict
# Boxplot dict
dict = {
"is_git_commit": [],
"date": [],
"hash": [],
"author": [],
"message": [],
"x": [],
"min": [],
"quantile25": [],
"quantile50": [],
"quantile75": [],
"max": [],
"mu": [],
"pvalue": [],
"nsamples": [],
"custom_colors": []
}
self.sources["boxplot_source"].data = dict
return
# Select all data matching current test/var/backend
runs = self.data.loc[[self.widgets["select_test"].value],
self.widgets["select_var"].value,
self.widgets["select_backend"].value]
runs = runs.sort_values(by=["timestamp"])
timestamps = runs["timestamp"]
x_series, x_metadata = helper.gen_x_series(self.metadata,
timestamps.sort_values(),
self.current_n_runs)
# Update source
main_dict = runs.to_dict("series")
main_dict["x"] = x_series
# Add metadata (for tooltip)
main_dict.update(x_metadata)
# Select the last n runs only
n = self.current_n_runs
main_dict = {key: value[-n:] for key, value in main_dict.items()}
# Generate color series for display of failed checks
custom_colors = [True] * len(main_dict["check"])
for i in range(len(main_dict["check"])):
custom_colors[
i] = "#1f77b4" if main_dict["check"][i] else "#cc2b2b"
# Generate ColumnDataSources for the 3 dotplots
for stat in ["sigma", "s10", "s2"]:
dict = {
"%s_x" % stat: main_dict["x"],
"is_git_commit": main_dict["is_git_commit"],
"date": main_dict["date"],
"hash": main_dict["hash"],
"author": main_dict["author"],
"message": main_dict["message"],
stat: main_dict[stat],
"nsamples": main_dict["nsamples"],
"accuracy_threshold": main_dict["accuracy_threshold"],
"custom_colors": custom_colors
}
if stat == "s10" or stat == "s2":
dict["%s_lower_bound" % stat] = main_dict["%s_lower_bound" %
stat]
# Filter outliers if the box is checked
if len(self.widgets["outliers_filtering_compare"].active) > 0:
outliers = helper.detect_outliers(dict[stat])
dict[stat] = helper.remove_outliers(dict[stat], outliers)
dict["%s_x" % stat] = helper.remove_outliers(
dict["%s_x" % stat], outliers)
# Assign ColumnDataSource
self.sources["%s_source" % stat].data = dict
# Generate ColumnDataSource for the boxplot
dict = {
"is_git_commit": main_dict["is_git_commit"],
"date": main_dict["date"],
"hash": main_dict["hash"],
"author": main_dict["author"],
"message": main_dict["message"],
"x": main_dict["x"],
"min": main_dict["min"],
"quantile25": main_dict["quantile25"],
"quantile50": main_dict["quantile50"],
"quantile75": main_dict["quantile75"],
"max": main_dict["max"],
"mu": main_dict["mu"],
"pvalue": main_dict["pvalue"],
"nsamples": main_dict["nsamples"],
"custom_colors": custom_colors
}
self.sources["boxplot_source"].data = dict
# Update x axis
helper.reset_x_range(self.plots["boxplot"],
self.sources["boxplot_source"].data["x"])
helper.reset_x_range(self.plots["sigma_plot"],
self.sources["sigma_source"].data["sigma_x"])
helper.reset_x_range(self.plots["s10_plot"],
self.sources["s10_source"].data["s10_x"])
helper.reset_x_range(self.plots["s2_plot"],
self.sources["s2_source"].data["s2_x"])
def update_plots(self):
if self.data.empty:
# Initialize empty dict and return
dict = {
"value_x": [],
"value": [],
"ieee": [],
"custom_colors": []
}
self.source.data = dict
return
# Select all data matching current test/var/backend
runs = self.data.loc[
[self.widgets["select_deterministic_test"].value],
self.widgets["select_deterministic_var"].value,
self.widgets["select_deterministic_backend"].value]
runs = runs.sort_values(by=["timestamp"])
timestamps = runs["timestamp"]
x_series, x_metadata = helper.gen_x_series(self.metadata,
timestamps.sort_values(),
self.current_n_runs)
# Update source
dict = runs.to_dict("series")
dict["value_x"] = x_series
# Add metadata (for tooltip)
dict.update(x_metadata)
# Select the last n runs only
n = self.current_n_runs
dict = {key: value[-n:] for key, value in dict.items()}
# Generate color series for display of failed checks
dict["custom_colors"] = [True] * len(dict["check"])
for i in range(len(dict["check"])):
dict["custom_colors"][
i] = "#1f77b4" if dict["check"][i] else "#cc2b2b"
# Filter outliers if the box is checked
if len(self.widgets["outliers_filtering_deterministic"].active) > 0:
outliers = helper.detect_outliers(dict["value"])
dict["value"] = helper.remove_outliers(dict["value"], outliers)
dict["value_x"] = helper.remove_outliers(dict["%value_x"],
outliers)
# Series to display IEEE results. This is different than reference_value,
# because the latter has to be 0 when no reference run has been done
# (because the value will be shown in the tooltip)
dict["ieee"] = [0] * len(dict["value"])
for i in range(len(dict["value"])):
if dict["check"][i]:
dict["ieee"][i] = nan
else:
dict["ieee"][i] = dict["reference_value"][i]
self.source.data = dict
# Update x axis
helper.reset_x_range(self.plots["comparison_plot"],
self.source.data["value_x"])
def update_plots(self):
# Select all data matching current test/var/backend
runs = self.data.loc[[self.widgets["select_test"].value],
self.widgets["select_var"].value,
self.widgets["select_backend"].value]
timestamps = runs["timestamp"]
x_series, x_metadata = self.gen_x_series(timestamps.sort_values())
# Update source
main_dict = runs.to_dict("series")
main_dict["x"] = x_series
# Add metadata (for tooltip)
main_dict.update(x_metadata)
# Select the last n runs only
n = self.current_n_runs
main_dict = {key: value[-n:] for key, value in main_dict.items()}
# Generate ColumnDataSources for the 3 dotplots
for stat in ["sigma", "s10", "s2"]:
dict = {
"%s_x" % stat: main_dict["x"],
"is_git_commit": main_dict["is_git_commit"],
"date": main_dict["date"],
"hash": main_dict["hash"],
"author": main_dict["author"],
"message": main_dict["message"],
stat: main_dict[stat],
"nsamples": main_dict["nsamples"],
}
if stat == "s10" or stat == "s2":
dict["%s_lower_bound" % stat] = main_dict["%s_lower_bound" %
stat]
# Filter outliers if the box is checked
if len(self.widgets["outliers_filtering_compare"].active) > 0:
outliers = helper.detect_outliers(dict[stat])
dict[stat] = helper.remove_outliers(dict[stat], outliers)
dict["%s_x" % stat] = helper.remove_outliers(
dict["%s_x" % stat], outliers)
# Assign ColumnDataSource
self.sources["%s_source" % stat].data = dict
# Generate ColumnDataSource for the boxplot
dict = {
"is_git_commit": main_dict["is_git_commit"],
"date": main_dict["date"],
"hash": main_dict["hash"],
"author": main_dict["author"],
"message": main_dict["message"],
"x": main_dict["x"],
"min": main_dict["min"],
"quantile25": main_dict["quantile25"],
"quantile50": main_dict["quantile50"],
"quantile75": main_dict["quantile75"],
"max": main_dict["max"],
"mu": main_dict["mu"],
"pvalue": main_dict["pvalue"],
"nsamples": main_dict["nsamples"]
}
self.sources["boxplot_source"].data = dict
# Update x axis
helper.reset_x_range(self.plots["boxplot"],
self.sources["boxplot_source"].data["x"])
helper.reset_x_range(self.plots["sigma_plot"],
self.sources["sigma_source"].data["sigma_x"])
helper.reset_x_range(self.plots["s10_plot"],
self.sources["s10_source"].data["s10_x"])
helper.reset_x_range(self.plots["s2_plot"],
self.sources["s2_source"].data["s2_x"])
def update_plots(self):
groupby_display = self.widgets["groupby_radio"].labels[
self.widgets["groupby_radio"].active]
groupby = self.factors_dict[groupby_display]
filterby_display = self.widgets["filterby_radio"].labels[
self.widgets["filterby_radio"].active]
filterby = self.factors_dict[filterby_display]
# Groupby and aggregate lines belonging to the same group in lists
if not self.run_data.empty:
groups = self.run_data[self.run_data.index.isin(
[self.widgets["select_filter"].value],
level=filterby)].groupby(groupby)
groups = groups.agg({
"sigma": lambda x: x.tolist(),
"s10": lambda x: x.tolist(),
"s2": lambda x: x.tolist(),
"mu": lambda x: x.tolist(),
# Used for mu weighted average first, then will be replaced
"nsamples": lambda x: x.tolist()
})
# Compute the new distributions, ...
groups = self.data_processing(groups).to_dict("list")
else:
groups = {
"mu": [],
"nsamples": [],
"mu_x": [],
"sigma_x": [],
"sigma_min": [],
"sigma_quantile25": [],
"sigma_quantile50": [],
"sigma_quantile75": [],
"sigma_max": [],
"sigma_mu": [],
"s10_x": [],
"s10_min": [],
"s10_quantile25": [],
"s10_quantile50": [],
"s10_quantile75": [],
"s10_max": [],
"s10_mu": [],
"s2_x": [],
"s2_min": [],
"s2_quantile25": [],
"s2_quantile50": [],
"s2_quantile75": [],
"s2_max": [],
"s2_mu": []
}
# Update source
# Assign each ColumnDataSource, starting with the boxplots
for prefix in ["sigma", "s10", "s2"]:
dict = {
"%s_x" % prefix: groups["%s_x" % prefix],
"%s_min" % prefix: groups["%s_min" % prefix],
"%s_quantile25" % prefix: groups["%s_quantile25" % prefix],
"%s_quantile50" % prefix: groups["%s_quantile50" % prefix],
"%s_quantile75" % prefix: groups["%s_quantile75" % prefix],
"%s_max" % prefix: groups["%s_max" % prefix],
"%s_mu" % prefix: groups["%s_mu" % prefix],
"nsamples": groups["nsamples"]
}
# Filter outliers if the box is checked
if len(self.widgets["outliers_filtering_inspect"].active) > 0:
# Boxplots will be filtered by max then min
top_outliers = helper.detect_outliers(dict["%s_max" % prefix])
helper.remove_boxplot_outliers(dict, top_outliers, prefix)
bottom_outliers = helper.detect_outliers(dict["%s_min" %
prefix])
helper.remove_boxplot_outliers(dict, bottom_outliers, prefix)
self.sources["%s_source" % prefix].data = dict
# Finish with the mu plot
dict = {
"mu_x": groups["mu_x"],
"mu": groups["mu"],
"nsamples": groups["nsamples"]
}
self.sources["mu_source"].data = dict
# Filter outliers if the box is checked
if len(self.widgets["outliers_filtering_inspect"].active) > 0:
mu_outliers = helper.detect_outliers(groups["mu"])
groups["mu"] = helper.remove_outliers(groups["mu"], mu_outliers)
groups["mu_x"] = helper.remove_outliers(groups["mu_x"],
mu_outliers)
# Update plots axis/titles
# Get display string of the last (unselected) factor
factors_dict = self.factors_dict.copy()
del factors_dict[groupby_display]
del factors_dict[filterby_display]
for_all = list(factors_dict.keys())[0]
# Update all display strings for plot title (remove caps, plural)
groupby_display = groupby_display.lower()
filterby_display = filterby_display.lower()[:-1]
for_all = for_all.lower()
self.plots["mu_inspect"].title.text = \
"Empirical average μ of %s (groupped by %s, for all %s)" \
% (filterby_display, groupby_display, for_all)
self.plots["sigma_inspect"].title.text = \
"Standard deviation σ of %s (groupped by %s, for all %s)" \
% (filterby_display, groupby_display, for_all)
self.plots["s10_inspect"].title.text = \
"Significant digits s of %s (groupped by %s, for all %s)" \
% (filterby_display, groupby_display, for_all)
self.plots["s2_inspect"].title.text = \
"Significant digits s of %s (groupped by %s, for all %s)" \
% (filterby_display, groupby_display, for_all)
helper.reset_x_range(self.plots["mu_inspect"], groups["mu_x"])
helper.reset_x_range(self.plots["sigma_inspect"], groups["sigma_x"])
helper.reset_x_range(self.plots["s10_inspect"], groups["s10_x"])
helper.reset_x_range(self.plots["s2_inspect"], groups["s2_x"])