feat(eda):add categorical interaction in create_report

dataprep/eda/configs.py

@@ -1109,9 +1109,14 @@ class Interactions(BaseModel):
     """
     enable: bool, default True
         Whether to create this element
+    cat_enable: bool, default False
+        where enable categorical column in interactions. By default it is False,
+        which means only numerical interactions are computed. If set to True, cat-cat
+        and cat-num interactions will be computed.
     """
 
     enable: bool = True
+    cat_enable: bool = False
 
 
 class Correlations(BaseModel):

dataprep/eda/correlation/render.py

@@ -57,7 +57,7 @@ def render_correlation(itmdt: Intermediate, cfg: Config) -> Any:
     elif itmdt.visual_type == "correlation_scatter":
         visual_elem = render_scatter(itmdt, plot_width, plot_height, cfg)
     elif itmdt.visual_type == "correlation_crossfilter":
-        visual_elem = render_crossfilter(itmdt, plot_width, plot_height)
+        visual_elem = render_crossfilter(itmdt, plot_width, plot_height, cfg)
     else:
         raise NotImplementedError(f"Unknown visual type {itmdt.visual_type}")
 
@@ -372,25 +372,34 @@ def render_scatter(
 
 
 ######### Interactions for report #########
-def render_crossfilter(itmdt: Intermediate, plot_width: int, plot_height: int) -> column:
+def render_crossfilter(
+    itmdt: Intermediate, plot_width: int, plot_height: int, cfg: Config
+) -> column:
     """
     Render crossfilter scatter plot with a regression line.
     """
 
     # pylint: disable=too-many-locals, too-many-function-args
-    df = itmdt["data"]
-    df["__x__"] = df[df.columns[0]]
-    df["__y__"] = df[df.columns[0]]
-    source_scatter = ColumnDataSource(df)
-    source_xy_value = ColumnDataSource({"x": [df.columns[0]], "y": [df.columns[0]]})
-    var_list = list(df.columns[:-2])
+
+    if cfg.interactions.cat_enable:
+        all_cols = itmdt["all_cols"]
+    else:
+        all_cols = itmdt["num_cols"]
+
+    scatter_df = itmdt["scatter_source"]
+    # all other plots except for scatter plot, used for cat-cat and cat-num interactions.
+    other_plots = itmdt["other_plots"]
+    scatter_df["__x__"] = scatter_df[scatter_df.columns[0]]
+    scatter_df["__y__"] = scatter_df[scatter_df.columns[0]]
+    source_scatter = ColumnDataSource(scatter_df)
+    source_xy_value = ColumnDataSource({"x": [scatter_df.columns[0]], "y": [scatter_df.columns[0]]})
+    var_list = list(all_cols)
 
     xcol = source_xy_value.data["x"][0]
     ycol = source_xy_value.data["y"][0]
 
     tooltips = [("X-Axis: ", "@__x__"), ("Y-Axis: ", "@__y__")]
-
-    fig = Figure(
+    scatter_fig = Figure(
         plot_width=plot_width,
         plot_height=plot_height,
         toolbar_location=None,
@@ -399,10 +408,12 @@ def render_crossfilter(itmdt: Intermediate, plot_width: int, plot_height: int) -
         x_axis_label=xcol,
         y_axis_label=ycol,
     )
-    scatter = fig.scatter("__x__", "__y__", source=source_scatter)
+    scatter = scatter_fig.scatter("__x__", "__y__", source=source_scatter)
 
     hover = HoverTool(tooltips=tooltips, renderers=[scatter])
-    fig.add_tools(hover)
+    scatter_fig.add_tools(hover)
+
+    fig_all_in_one = column(scatter_fig, sizing_mode="stretch_width")
 
     x_select = Select(title="X-Axis", value=xcol, options=var_list, width=150)
     y_select = Select(title="Y-Axis", value=ycol, options=var_list, width=150)
@@ -413,19 +424,31 @@ def render_crossfilter(itmdt: Intermediate, plot_width: int, plot_height: int) -
             args=dict(
                 scatter=source_scatter,
                 xy_value=source_xy_value,
-                x_axis=fig.xaxis[0],
+                fig_all_in_one=fig_all_in_one,
+                scatter_plot=scatter_fig,
+                x_axis=scatter_fig.xaxis[0],
+                other_plots=other_plots,
             ),
             code="""
         let currentSelect = this.value;
         let xyValueData = xy_value.data;
         let scatterData = scatter.data;
-
         xyValueData['x'][0] = currentSelect;
-        scatterData['__x__'] = scatterData[currentSelect];
-
-        x_axis.axis_label = currentSelect;
-        scatter.change.emit();
         xy_value.change.emit();
+
+        const children = []
+        let ycol = xyValueData['y'][0];
+        let col = currentSelect + '_' + ycol
+        if (col in other_plots) {
+            children.push(other_plots[col])
+        }
+        else {
+            scatterData['__x__'] = scatterData[currentSelect];
+            x_axis.axis_label = currentSelect;
+            scatter.change.emit();
+            children.push(scatter_plot)
+        }
+        fig_all_in_one.children = children;        
         """,
         ),
     )
@@ -435,25 +458,39 @@ def render_crossfilter(itmdt: Intermediate, plot_width: int, plot_height: int) -
             args=dict(
                 scatter=source_scatter,
                 xy_value=source_xy_value,
-                y_axis=fig.yaxis[0],
+                fig_all_in_one=fig_all_in_one,
+                scatter_plot=scatter_fig,
+                y_axis=scatter_fig.yaxis[0],
+                other_plots=other_plots,
             ),
             code="""
-        let currentSelect = this.value;
+        let ycol = this.value;
         let xyValueData = xy_value.data;
         let scatterData = scatter.data;
-
-        xyValueData['y'][0] = currentSelect;
-        scatterData['__y__'] = scatterData[currentSelect];
-
-        y_axis.axis_label = currentSelect;
-        scatter.change.emit();
+        xyValueData['y'][0] = ycol;
         xy_value.change.emit();
+
+        const children = []
+        let xcol = xyValueData['x'][0];
+        let col = xcol + '_' + ycol;
+        if (col in other_plots) {
+            children.push(other_plots[col])
+        }
+        else {
+            scatterData['__y__'] = scatterData[ycol];
+            y_axis.axis_label = ycol;
+            scatter.change.emit();
+            children.push(scatter_plot)
+        }
+        fig_all_in_one.children = children;        
         """,
         ),
     )
 
-    fig = column(row(x_select, y_select, align="center"), fig, sizing_mode="stretch_width")
-    return fig
+    interaction_fig = column(
+        row(x_select, y_select, align="center"), fig_all_in_one, sizing_mode="stretch_width"
+    )
+    return interaction_fig
 
 
 # ######### Interactions for report #########

dataprep/eda/create_report/formatter.py

@@ -13,7 +13,7 @@
 from ..configs import Config
 from ..correlation import render_correlation
 from ..correlation.compute.overview import correlation_nxn
-from ..distribution import render
+from ..distribution import compute, render
 from ..utils import _calc_line_dt
 from ..distribution.compute.overview import calc_stats
 from ..distribution.compute.univariate import calc_stats_dt, cont_comps, nom_comps
@@ -146,16 +146,21 @@ def _format_variables(df: EDAFrame, cfg: Config, data: Dict[str, Any]) -> Dict[s
 def _format_interaction(data: Dict[str, Any], cfg: Config) -> Dict[str, Any]:
     """Format of Interaction section"""
     res: Dict[str, Any] = {}
-    if len(data["num_cols"]) > 0:
-        # interactions
-        if cfg.interactions.enable:
-            res["has_interaction"] = True
-            itmdt = Intermediate(data=data["scat"], visual_type="correlation_crossfilter")
-            rndrd = render_correlation(itmdt, cfg)
-            rndrd.sizing_mode = "stretch_width"
-            res["interactions"] = components(rndrd)
-        else:
-            res["has_interaction"] = False
+    # interactions
+    if cfg.interactions.enable:
+        res["has_interaction"] = True
+        itmdt = Intermediate(
+            scatter_source=data["interaction.scatter_source"],
+            other_plots=data["interaction.other_plots"],
+            num_cols=data["num_cols"],
+            all_cols=data["all_cols"],
+            visual_type="correlation_crossfilter",
+        )
+        rndrd = render_correlation(itmdt, cfg)
+        rndrd.sizing_mode = "stretch_width"
+        res["interactions"] = components(rndrd)
+    else:
+        res["has_interaction"] = False
     return res
 
 
@@ -372,25 +377,44 @@ def basic_computations(
     cfg
         The config dict user passed in. E.g. config =  {"hist.bins": 20}
         Without user's specifications, the default is "auto"
-    """  # pylint: disable=too-many-branches
+    """
+    # pylint: disable=too-many-branches, protected-access, too-many-locals
 
     variables_data = _compute_variables(df, cfg)
     overview_data = _compute_overview(df, cfg)
     data: Dict[str, Any] = {**variables_data, **overview_data}
 
     df_num = df.select_num_columns()
+    num_columns = df_num.columns
+    cat_columns = [col for col in df.columns if col not in num_columns]
     data["num_cols"] = df_num.columns
+    data["all_cols"] = df.columns
     # interactions
     if cfg.interactions.enable:
         if cfg.scatter.sample_size is not None:
             sample_func = lambda x: x.sample(n=min(cfg.scatter.sample_size, x.shape[0]))
         else:
             sample_func = lambda x: x.sample(frac=cfg.scatter.sample_rate)
-        data["scat"] = df_num.frame.map_partitions(
+        data["interaction.scatter_source"] = df_num.frame.map_partitions(
             sample_func,
             meta=df_num.frame,
         )
 
+        data["interaction.other_plots"] = {}
+        if cfg.interactions.cat_enable:
+            other_plots = {}
+            curr_cfg = Config.from_dict(display=["Box Plot", "Nested Bar Chart"])
+            for cat_col in cat_columns:
+                for other_col in df.columns:
+                    if (cat_col == other_col) or (cat_col + "_" + other_col in other_plots):
+                        continue
+                    # print(f"cat col:{cat_col}, other col:{other_col}")
+                    imdt = compute(df._ddf, cat_col, other_col, cfg=curr_cfg)
+                    box_plot = render(imdt, curr_cfg)["layout"][0]
+                    other_plots[cat_col + "_" + other_col] = box_plot
+                    other_plots[other_col + "_" + cat_col] = box_plot
+            data["interaction.other_plots"] = other_plots
+
     # correlations
     if cfg.correlations.enable:
         data.update(zip(("cordx", "cordy", "corrs"), correlation_nxn(df_num, cfg)))

dataprep/tests/eda/test_config.py

@@ -52,3 +52,4 @@ def test_sanity_compute_3(simpledf: dd.DataFrame) -> None:
 
 def test_report(simpledf: dd.DataFrame) -> None:
     create_report(simpledf, display=["Overview", "Interactions"])
+    create_report(simpledf, display=["Interactions"], config={"interactions.cat_enable": True})