Debug

Debug#

We provide a debugger module to facilitate the fine-tuning of the knowledge base of a Draco instance. You can learn more about the applications of this API in the applications section.

DracoDebug#

class draco.debug.DracoDebug(specs, draco=__DEFAULT_DRACO__)#

Class to aid the debugging of visualization specifications against the knowledge base of a Draco instance. Methods exposed by this class give insight into the constraints and weights defined in the Draco instance and how they are violated by the supplied visualization specifications.

__init__(specs, draco=__DEFAULT_DRACO__)#

Initializes a new debugger instance.

Parameters:

  • specs (dict[str, Union[Iterable[str], str]]) – dict of specifications to analyze, keys representing the spec name and values the corresponding ASP declarations.

  • draco (Draco) – The Draco instance to be debugged.

property chart_preferences: DataFrame#

Returns a DataFrame with four columns:

  • chart_name: Specification name, a key of self.specs

  • pref_name: Preference (feature) name,

    element of self.feature_names

  • pref_description: Preference (feature) description.

  • count: The number of times the spec

    identified by chart_name violates pref_name.

  • weight: The weight associated with pref_name.

The returned DataFrame always has len(self.specs) * len(self.feature_names) rows.

Returns:

The above-described pandas DataFrame.

property feature_descriptions: dict[str, str]#
Returns:

a dict mapping feature names to their descriptions

property feature_names: set[str]#
Returns:

a set of the names of the features (constraints) of the underlying Draco instance

DracoDebugPlotter#

class draco.debug.DracoDebugPlotter(chart_preferences)#

Class to aid the plotting of Draco debug data produced by DracoDebug.chart_preferences.

__init__(chart_preferences)#

Initializes a new plotter instance, to be used to create charts visualizing the Draco debug data.

Parameters:

chart_preferences (DataFrame) – the DataFrame returned by DracoDebug.chart_preferences

create_chart(*, cfg=None, violated_prefs_only=False, plot_size=None)#

Creates a vertically concatenated chart made up of an aligned bar chart visualizing feature weights and a heatmap visualizing the number of times each spec violates each feature.

Parameters:

  • cfg (DracoDebugChartConfig | ChartConfig | None) – the configuration based on which the chart title and sorting is set. A default configuration will be used of this is not specified, featuring an alphabetical sort by feature name and spec name.

  • violated_prefs_only (bool) – whether to only include features that are violated by at least one spec. Defaults to False.

  • plot_size (tuple[float, float] | None) – the size of the plot described as a tuple of (width, height).

Return type:

VConcatChart

Returns:

the above-described Altair chart