improve modularity

__init__.py

@@ -1,2 +1,3 @@
 from .models import ActiveExplore
+from .apps import ExploreML
 from .data import loadActiveH5

apps/__init__.py

@@ -0,0 +1 @@
+from .explore_predict import ExploreML

apps/explore_predict.py

@@ -0,0 +1,47 @@
+from exploreML.models import ActiveExplore
+from exploreML.models import PredictMap
+from bokeh.layouts import column, row
+
+class ExploreML:
+
+
+    def __init__(self,grid_shape=(1,1),**kwargs):
+
+        self.grid_shape = grid_shape
+        self.layouts = []
+        #maybe map_p_high/low
+
+
+    def addActiveExplore(self,data_dict,sampling_dict,**kwargs):
+
+        active = ActiveExplore(data_dict, sampling_dict, **kwargs)
+
+        self.layouts += [active.layout]
+        self.toolbar = active.toolbar
+        self.clust_dict = active.clust_dict # Index clusters for heatmap
+        self.clust_methods = active.clust_methods
+        self.upper_source = active.upper_source
+        self.upper_dict = active.upper_dict
+        self.row_name = active.row_name
+        self.col_name = active.col_name
+        self.sample_sliders = active.sample_sliders
+        self.plot_size = active.plot_size
+        self.radio_button_group = active.radio_button_group
+        self.toggle = active.toggle
+        self.data_toggle = active.data_toggle
+        self.active_dim = active.active_dim
+
+
+    def addPredictMap(self,pred_df,map_dict,**kwargs):
+
+        predict = PredictMap(pred_df, map_dict, toolbar=self.toolbar,
+                             clust_dict=self.clust_dict,clust_methods=self.clust_methods,
+                             upper_source=self.upper_source,upper_dict=self.upper_dict,
+                             row_name=self.row_name,col_name=self.col_name,
+                             radio_button_group=self.radio_button_group,toggle=self.toggle,
+                             sample_sliders=self.sample_sliders,plot_size=self.plot_size,
+                             data_toggle=self.data_toggle,active_dim=self.active_dim,**kwargs)
+        self.layouts += [predict.layout]
+
+    def Layout(self):
+        self.layout = row(*self.layouts)

models/__init__.py

@@ -1 +1,2 @@
 from .active_explore import ActiveExplore
+from .predict_map import PredictMap

models/active_explore.py

@@ -2,15 +2,15 @@
 import pandas as pd
 import seaborn as sns
 import math
-import os
 from bokeh.io import show, save
 from bokeh.plotting import Figure, output_file, show, output_notebook
 from bokeh.layouts import column, row
 from bokeh.models import ColorBar, LinearColorMapper, BasicTicker, CustomJS, ColumnDataSource,\
  Toggle, Slider, RadioButtonGroup, Select, Legend, ColorPicker, Panel, Tabs, RangeSlider,HoverTool
+from bokeh.models.widgets import Div
 from bokeh.palettes import all_palettes
 from scipy.cluster.hierarchy import linkage, dendrogram
-from exploreML.models.custom_tools import ResetTool
+import itertools
 
 with open('exploreML/models/active_explore_js/slider_callback.js','r') as f:
     slider_callback_js = f.read()
@@ -45,15 +45,15 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
                  row_name='dim1', col_name='dim2', val_name='entry_value',hide_heatmap_labels=False,
                  heatmap_colors = 'default', n_colors=10, inds_colors=[1,2,4,5,6,7,8],
                  color_palette='Category10',plot_size=700, line_width=500, line_height=300,
-                 name='active explorer', url='active_explorer.html', plot_location='below'):
+                 name='active explorer', url='active_explorer.html', plot_location='below',
+                 file_output=True):
 
         self.name = name
         self.numLinePlots = num_line_plots
         self.is_sym = is_sym
-
-        # Output file
-        output_file(filename=url, title=name)
-        # output_notebook()
+        self.row_name = row_name
+        self.col_name = col_name
+        self.plot_size = plot_size
 
         # Load data
         M = data_dict['M']
@@ -81,6 +81,7 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
         sampler_color = dict(zip(sampling_names, sampler_palette[:n_samplers]))
 
         active_dim = sampling_dfs[0].shape[0]
+        self.active_dim = active_dim
         if self.is_sym:
             sampling_methods = [self._makeSymAL(sampler, row_coord, col_coord, active_x, batch_col)\
                                 for sampler in sampling_dfs]
@@ -94,6 +95,7 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
         # Precompute the indices for the different clustering methods
         clust_dict = {method:self._makeClustIndex(M, method) for method in clust_methods}
         self.clust_dict = clust_dict
+        self.clust_methods = clust_methods
 
         # Active learning data
         sampling_data = {sample:self._addIndexCols(df, df_index, df_cols, meta_vars)\
@@ -108,6 +110,7 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
         # Column source data of mask for each clustering method
         upper_dict = {method:self._makeGImatrix(self._makeMaskUpper(M.iloc[clust_dict[method][0],clust_dict[method][1]]),meta_vars)\
                    .dropna(axis=0).to_dict(orient='list') for method in clust_methods}
+        self.upper_dict = upper_dict
 
         # Collecting the quantitative columns from the sampling data
         quant_options = sampling_methods[0].describe().T.query('std > 0').index.to_list()
@@ -121,6 +124,7 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
                         for sampler,data in sampling_data.items()}
 
         upper_source = ColumnDataSource(data=upper_dict[init_clust])
+        self.upper_source = upper_source
 
         # Initialize sources
         # GI column source data
@@ -235,19 +239,14 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
         with open('exploreML/models/active_explore_js/radio_call.js','r') as f:
             radio_call_js = f.read()
 
-        radio_call = CustomJS(args=dict(methods=clust_methods,clust_dict=clust_dict,plot=p,
-                                        up_source=upper_source, up_dict=upper_dict),code=radio_call_js)
-
-        sample_sliders = {sampler:Slider(start=0, end=active_dim, value=0, step=1,\
+        # TODO: Make step a variable
+        sample_sliders = {sampler:Slider(start=0, end=active_dim, value=0, step=120,\
                           title=title,max_width=300)\
                           for sampler,title in zip(sampling_names,sampling_titles)}
 
-        radio_button_group = RadioButtonGroup(labels=[x.capitalize() for x in clust_methods], active=0,max_width=300)
-
         slider_js = {sampler:sample_sliders[sampler].js_on_change('value', self._slider_callback(sampler, active_sources, sampling_sources, symMult))\
                      for sampler in sampling_names}
-
-        radio_button_group.js_on_click(radio_call)
+        self.sample_sliders = sample_sliders
 
         toggle = Toggle(label="Lower Triangle (Toggle)", button_type="primary",max_width=300)
         toggle.js_on_click(CustomJS(args=dict(plot=upperFig),code="""
@@ -256,6 +255,7 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
             plot.visible = cb_obj.active;
 
         """))
+        self.toggle = toggle
 
         #default, primary, success, warning, danger, light
         #Toggle to show and hide training examples
@@ -270,6 +270,7 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
             data_toggle_js = f.read()
 
         data_toggle = Toggle(label="Show All (Toggle)", button_type="primary",max_width=300)
+        self.data_toggle = data_toggle
         data_toggle.js_on_click(CustomJS(args=dict(sliders=sample_sliders, active_dim=\
                                                    active_dim),
         code=data_toggle_js))
@@ -282,19 +283,32 @@ def __init__(self, data_dict, sampling_dict, is_sym=False, num_line_plots=2,
         train_picker = ColorPicker(width=50,color='white')
         train_picker.js_link('color', trainFig.glyph, 'fill_color')
 
-        sliders = column(list(sample_sliders.values()))
+        sliders = column(list(sample_sliders.values()),name='slider_col')
 
         linePlots = [column(*line) for line in list(zip(line_selects,line_tabs))]
 
-        heatmap_layout = row(column(radio_button_group,toggle,sliders,data_toggle,
-                             row(train_toggle,train_picker),range_slider,select_colorbar,width=340),
-                             p)
+        radio_call = CustomJS(args=dict(methods=clust_methods,clust_dict=clust_dict,plot=p,
+                                up_source=upper_source, up_dict=upper_dict),code=radio_call_js)
+
+        radio_button_group = RadioButtonGroup(labels=[x.capitalize() for x in clust_methods], active=0, max_width=300)
+        radio_button_group.js_on_click(radio_call)
+        self.radio_button_group = radio_button_group
+
+        toolbar = column(radio_button_group,toggle,sliders,data_toggle,
+                             row(train_toggle,train_picker),
+                             range_slider,select_colorbar,
+                             width=310,name='toolbar', margin=(0,10,0,10))
+        heatmap_layout = row(toolbar,p)
+
+        layout = self._make_layout(heatmap_layout,linePlots,plot_location)
+
+        self.toolbar = toolbar
+        self.layout = layout
 
-        layout = self._make_layout(heatmap_layout,linePlots, plot_location)
+        if file_output:
 
-        # show(layout)
-        # self.layout = layout
-        save(layout)
+            output_file(filename=url, title=name)
+            save(layout)
 
     def _make_layout(self, heatmap_layout, linePlots, plot_location):
 
@@ -313,7 +327,8 @@ def _make_layout(self, heatmap_layout, linePlots, plot_location):
             layout = column(line_layout, heatmap_layout)
 
         elif plot_location == 'below': # Below
-            layout = column(heatmap_layout,line_layout)
+            # layout = column(heatmap_layout,Div(height=80),line_layout)
+            layout = column(heatmap_layout,Div(height=20),line_layout)
 
         return layout
 
@@ -331,32 +346,8 @@ def _line_callback(self, Fig, Fig2, Figs1, Figs2, samplerCol_meta):
                                       plot2=Figs2,
                                       col_meta=samplerCol_meta,
                                       yaxis=Fig.yaxis[0],
-                                      yaxis2=Fig2.yaxis[0]), code="""
-                    console.log('select: value=' + this.value, this.toString())
-
-                    var select = cb_obj.value;
-                    const keys = Object.keys(plot);
-                    var keysLength = keys.length;
-
-                    for (var i = 0; i < keysLength; i++) {
-                      plot[keys[i]].glyph.y.field = select;
-                      plot[keys[i]].data_source.change.emit();
-
-                      plot2[keys[i]].glyph.y.field = select;
-                      plot2[keys[i]].data_source.change.emit();
-                    }
-
-                    var mx = col_meta[select].max + col_meta[select].max * 0.05;
-                    var mn = col_meta[select].min - col_meta[select].max * 0.05;
-
-                    fig.y_range.end = mx;
-                    fig.y_range.start = mn;
-                    yaxis.axis_label = select;
-
-                    fig2.y_range.end = mx;
-                    fig2.y_range.start = mn;
-                    yaxis2.axis_label = select;
-                                      """)
+                                      yaxis2=Fig2.yaxis[0]),
+                                      code=line_callback_js)
 
         return callback
 

models/active_explore_js/line_callback.js

@@ -12,10 +12,13 @@ for (var i = 0; i < keysLength; i++) {
   plot2[keys[i]].data_source.change.emit();
 }
 
-fig.y_range.end = col_meta[select].max;
-fig.y_range.start = col_meta[select].min;
+var mx = col_meta[select].max + col_meta[select].max * 0.05;
+var mn = col_meta[select].min - col_meta[select].max * 0.05;
+
+fig.y_range.end = mx;
+fig.y_range.start = mn;
 yaxis.axis_label = select;
 
-fig2.y_range.end = col_meta[select].max;
-fig2.y_range.start = col_meta[select].min;
+fig2.y_range.end = mx;
+fig2.y_range.start = mn;
 yaxis2.axis_label = select;

models/active_explore_js/radio_call_plots.js

@@ -0,0 +1,43 @@
+console.log('radio_button_group: active=' + this.active, this.toString())
+
+//Store radioVal for clustering methods
+let radioVal = cb_obj.active;
+
+let method = methods[radioVal];
+let x_range = clust_dict[method][3];
+let y_range = [].concat(clust_dict[method][2]).reverse();
+
+plot1.x_range.factors = x_range;
+plot1.y_range.factors = y_range;
+
+plot2.x_range.factors = x_range;
+plot2.y_range.factors = y_range;
+
+let cdsLists = up_dict[method];
+
+//Reorder mask
+//Assign mask source to variable
+var data = up_source.data;
+var keys = Object.keys(data);
+var keysLength = keys.length;
+
+// Reorder mask
+var mask_map = new Map();
+var new_map = new Map();
+for (var i = 0; i < keysLength; i++) {
+
+  //Assign mask source to map variable
+  mask_map.set(keys[i], data[keys[i]]);
+
+  // Reassign CDS
+  new_map.set(keys[i], cdsLists[keys[i]]);
+
+  // Clear old mask CDS
+  mask_map.get(keys[i]).splice(0, mask_map.get(keys[i]).length);
+
+  // Add new mask cds
+  mask_map.set(keys[i], mask_map.get(keys[i]).push(...new_map.get(keys[i])));
+
+}
+
+up_source.change.emit();