from .glueplotly import GluePlotly from plotly.graph_objects import FigureWidget from ipywidgets import IntText, Dropdown, BoundedIntText from IPython.display import clear_output import numpy as np import pandas as pd import networkx as nx import math import re class GlueNetworkPlotly (GluePlotly): default_size_marker = 5 #'planar_layout' 'bipartite_layout' layout_options = ['spring_layout', 'circular_layout', 'kamada_kawai_layout', 'random_layout', 'shell_layout', 'spectral_layout'] default_layout_type = layout_options[0] focused_size_marker = 3 options = {} G = None def __init__(self, data, dimensions, **kwargs): GluePlotly.__init__(self, data, dimensions, **kwargs) self.DefaultLayoutTitles("", "", "") self.options['line_width'] = BoundedIntText(description = 'Lines width:', value = 1, min=0, max=10) self.options['line_width'].observe(lambda v:self.UpdateTraces({'line.width':v['new']}), names='value') self.options['marker_size'] = BoundedIntText(description = 'Markers size:', value = 3, min=0, max=15) self.options['marker_size'].observe(lambda v:self.UpdateTraces({'marker.size':v['new']}), names='value') self.options['layout_type'] = Dropdown(description = 'Layout type:', value = self.default_layout_type, options=self.layout_options) self.options['layout_type'].observe(lambda v:self.updateRender(), names='value') self.options['layout_k'] = BoundedIntText(description = 'Layout k:', value = 5, min=1,max=10) self.options['layout_k'].observe(lambda v:self.updateRender(), names='value') self.options['layout_iter'] = BoundedIntText(description = 'Layout Iter:', value = 100, min=20,max=250) self.options['layout_iter'].observe(lambda v:self.updateRender(), names='value') self.DefaultLegend('v', 1.02, 1.0); self.updateRender() def createFigureWidget(self, x_id, y_id): traces = [] color = "#444444" df = self.data.to_dataframe() df_merge = df.merge(df, on=y_id) dfo = pd.crosstab(df_merge[x_id + '_x'], df_merge[x_id + '_y']) idx = dfo.columns.union(dfo.index) dfo = dfo.reindex(index = idx, columns=idx, fill_value=0) self.G = nx.convert_matrix.from_pandas_adjacency(dfo) if self.options['layout_type'].value == 'bipartite_layout': pos = nx.bipartite_layout(self.G) elif self.options['layout_type'].value == 'circular_layout': pos = nx.circular_layout(self.G) elif self.options['layout_type'].value == 'kamada_kawai_layout': pos = nx.kamada_kawai_layout(self.G) elif self.options['layout_type'].value == 'planar_layout': pos = nx.planar_layout(self.G) elif self.options['layout_type'].value == 'random_layout': pos = nx.random_layout(self.G) elif self.options['layout_type'].value == 'shell_layout': pos = nx.shell_layout(self.G) elif self.options['layout_type'].value == 'spectral_layout': pos = nx.spectral_layout(self.G) elif self.options['layout_type'].value == 'shell_layout': pos = nx.shell_layout(self.G) else: pos = nx.spring_layout(self.G, k=self.options['layout_k'].value, iterations=self.options['layout_iter'].value) degree = dict(nx.degree(self.G)) n_x = [0 for i in range(len(self.G.nodes()))] n_y = [0 for i in range(len(self.G.nodes()))] n_t = ['' for i in range(len(self.G.nodes()))] n_s = [0 for i in range(len(self.G.nodes()))] for i, node in enumerate(self.G.nodes()): n_x[i] = pos[node][0] n_y[i] = pos[node][1] n_t[i] = node n_s[i] = degree[node] max_degree = max(n_s) min_degree = min(n_s) n_s = [math.ceil((n-min_degree)/(max_degree-min_degree)*self.focused_size_marker*self.options['marker_size'].value) + self.options['marker_size'].value for n in n_s] node_trace = { 'type' : "scatter", 'name': self.data.label + "_" + x_id, 'x' : n_x, 'y' : n_y, 'text' : n_t, 'mode' : 'markers', 'hoverinfo' :'text', 'marker' : { 'size' : n_s, 'color' : self.data.get_component(x_id).color, 'line' : { 'width' : self.options['line_width'].value, 'color' : self.data.get_component(x_id).color, } }, 'showlegend' : True, } e_x = [0 for i in range(2*len(self.G.edges()))] e_y = [0 for i in range(2*len(self.G.edges()))] for i, edge in enumerate(self.G.edges()): e_x[2*i] = pos[edge[0]][0] e_y[2*i] = pos[edge[0]][1] e_x[2*i+1] = pos[edge[1]][0] e_y[2*i+1] = pos[edge[1]][1] color = self.data.get_component(y_id).color color = re.match(r"rgb\((\d+),\s*(\d+),\s*(\d+)\)" , color) color = (int(color[1]),int(color[2]),int(color[3])) color = '#%02x%02x%02x' % color color = 'rgba'+ str(self.getDeltaColor(color, 0.2)) edge_trace = { 'type' : "scatter", 'name': self.data.label + "_" + y_id, 'x' : e_x, 'y' : e_y, 'line' : { 'width': self.options['line_width'].value, 'color' : color, }, 'hoverinfo' : 'none', 'mode' : 'lines', 'showlegend' : True, } if self.only_subsets == False: traces.append(edge_trace) for sset in self.data.subsets: if hasattr(sset,"disabled") == False or sset.disabled == False: color = sset.style.color dfs = df[sset.to_mask()] df_merge = dfs.merge(dfs, on=y_id) dfs = pd.crosstab(df_merge[x_id + '_x'], df_merge[x_id + '_y']) idx = dfs.columns.union(dfs.index) dfs = dfs.reindex(index = idx, columns=idx, fill_value=0) G = nx.convert_matrix.from_pandas_adjacency(dfs) e_x = [0 for i in range(2*len(G.edges()))] e_y = [0 for i in range(2*len(G.edges()))] for i, edge in enumerate(G.edges()): e_x[2*i] = pos[edge[0]][0] e_y[2*i] = pos[edge[0]][1] e_x[2*i+1] = pos[edge[1]][0] e_y[2*i+1] = pos[edge[1]][1] trace = { 'type' : "scatter", 'name': str(sset.label) + "_" + str(y_id), 'x' : e_x, 'y' : e_y, 'line' : { 'width': self.options['line_width'].value, 'color' : color }, 'hoverinfo' : 'none', 'mode' : 'lines', 'showlegend' : True, } traces.append(trace) traces.append(node_trace) layout = { 'title' : self.options['title'].value, 'margin' : { 'l':self.margins['left'].value, 'r':self.margins['right'].value, 'b':self.margins['bottom'].value, 't':self.margins['top'].value }, 'xaxis': { 'autorange' : True, 'zeroline': True, 'title' : self.options['xaxis'].value, 'autorange' : True , 'zeroline' : False, 'showticklabels' : False, 'showline' : True, 'showgrid' : False, }, 'yaxis': { 'autorange':True, 'zeroline': True, 'title' : self.options['yaxis'].value, 'autorange' : True , 'zeroline' : False, 'showticklabels' : False, 'showline' : True, 'showgrid' : False, }, 'showlegend': self.margins['showlegend'].value, 'legend' : { 'orientation' : self.margins['legend_orientation'].value, 'x' : self.margins['legend_xpos'].value, 'y' : self.margins['legend_ypos'].value } } return FigureWidget({ 'data': traces, 'layout': layout }) def updateRender(self): self.plotly_fig = self.createFigureWidget(self.dimensions[0], self.dimensions[1]) if self.only_subsets == False: self.plotly_fig.data[0].on_selection(lambda x,y,z : self.setSubset(x,y,z), True) if self.on_selection_callback is not None: self.plotly_fig.data[0].on_selection(self.on_selection_callback, True) GluePlotly.display(self) def updateSelection(self, ids): #self.parent.printInDebug(ids) self.plotly_fig.data[0].update( selectedpoints=ids ) def setSubset(self,trace,points,selector): if(self.parent is not None): self.parent.updateSelection(points.point_inds)