PR for #18: Update package structure for compatability with Shiny (#21)

* WIP for #18

* lots of changes for #18

* a working shiny app for #18

* draft 1 for #18

* fix for #18

* makes deployable for #18

* removes app for #18

* updates version for #18

* removes fixed test for #18

* unbumps version for #18

Author: MosesStewart

examples/examples.py

@@ -1,8 +1,8 @@
 """ Test examples """
 import sys, os, re
-sys.path.append('/'.join(re.split('/|\\\\', os.path.dirname( __file__ ))[0:-1]) + '/src/BootstrapReport')
+sys.path.append('/'.join(re.split('/|\\\\', os.path.dirname( __file__ ))[0:-1]))
 import timeit
-from main import ObjectOfInterest
+from src.BootstrapReport import ObjectOfInterest
 import pandas as pd
 
 def gamma_example():

src/BootstrapReport/helpers.py

@@ -209,9 +209,8 @@ def select_bandwidth(rot_se, max_grid, min_grid, num_gridpoints,
 
     return implied_normal_pdf, best_bandwidth, best_bandwidth_index, avg_tvd_list, expansion_count
 
-def plot_min_crossings(outfile, optimal_path, crossings, alpha, replicates, estimate, std, upper_cb, lower_cb, left_upper_cb, **kwargs):
+def plot_min_crossings(optimal_path, crossings, alpha, replicates, estimate, std, upper_cb, lower_cb, left_upper_cb, plt_set):
     """ Create the plot for `get_crossings` of ObjectOfInterest
-    :param outfile: destination for the plot
     :param optimal_path: nx2 numpy.array. First column contains x values, and second column contains y values of optimal path.
     :param crossings: number of crossings, or changes in direction
     :param alpha: 1 - alpha = confidence level for number of crossings 
@@ -222,19 +221,18 @@ def plot_min_crossings(outfile, optimal_path, crossings, alpha, replicates, esti
     :param lower_cb: function that returns the difference between the lower confidence bound and the cdf of the implied normal
     :param left_upper_cb: function that returns the left-handed limit of the difference between the upper confidence bound 
         and the cdf of the implied normal
+    :param plt_set: dictionary of plot options
     """
-    plt_set = {'fontsize': 18, 'legend_fontsize': 20, 'labelsize': 28, 'linecolor': '#f9665e', 'bandcolor': '#a8d9ed', 'linewidth': 2, 'dpi': 100}
-    for key, value in kwargs.items():
-        plt_set[key] = value
-    mpl.rcParams['axes.spines.bottom'] = False
-    plt.rcParams.update({'font.size': plt_set['fontsize']})
-    plt.rcParams.update({'legend.fontsize': plt_set['legend_fontsize']})
-    plt.rcParams.update({'axes.labelsize': plt_set['labelsize']})
+    plt.rc('font', size = plt_set['fontsize'])
+    plt.rc('legend', fontsize = plt_set['legend_fontsize'])
+    plt.rc('axes', labelsize = plt_set['labelsize'])
+    fig, ax = plt.subplots(figsize = (10, 10))
+    ax.spines.bottom.set_visible(False)
 
     num_rep, x, y = len(replicates), optimal_path[-1, 0], optimal_path[-1, 1]
     plot_data = ('Num. crossings = %d' % crossings)
     props = dict(boxstyle = 'round, pad = 0.75, rounding_size = 0.3', facecolor = 'white', alpha = 0.7)
-    plt.plot(optimal_path[:, 0], optimal_path[:, 1], color = plt_set['linecolor'], label = 'Optimal path', linewidth = plt_set['linewidth'])
+    ax.plot(optimal_path[:, 0], optimal_path[:, 1], color = plt_set['linecolor'], label = 'Optimal path', linewidth = plt_set['linewidth'])
 
     zero_null_rej = stats.norm.ppf(np.sqrt(np.log(2/alpha)/(2 * num_rep)), loc = estimate, scale = std)
     if zero_null_rej < replicates[0]:
@@ -258,24 +256,22 @@ def plot_min_crossings(outfile, optimal_path, crossings, alpha, replicates, esti
     else:
         postline = np.array([[replicates[-1], optimal_path[-1, 1]], [end + 0.5 * std, 0], [end + std, 0]])
 
-    plt.plot(preline[:, 0], preline[:, 1], color = plt_set['linecolor'], linewidth = plt_set['linewidth'])
-    plt.plot(postline[:, 0], postline[:, 1], color = plt_set['linecolor'], linewidth = plt_set['linewidth'])
+    ax.plot(preline[:, 0], preline[:, 1], color = plt_set['linecolor'], linewidth = plt_set['linewidth'])
+    ax.plot(postline[:, 0], postline[:, 1], color = plt_set['linecolor'], linewidth = plt_set['linewidth'])
     x_axis = np.concatenate([np.linspace(replicates[r], replicates[r + 1], 10) for r in range(num_rep - 1)], axis = None)
     x_axis = np.append(np.linspace(preline[0, 0], replicates[0], 15), x_axis)
     x_axis = np.append(x_axis, np.linspace(replicates[-1], postline[-1, 0], 15))
     upper_band, lower_band = [upper_cb(x) for x in x_axis], [lower_cb(x) for x in x_axis]
     lbound, rbound = x_axis[0] + 0.25 * std, x_axis[-1] - 0.25 * std
     topbound = np.max(upper_band) + 0.025 * (np.max(upper_band) - np.min(lower_band))
     botbound = np.min(lower_band) - 0.025 * (np.max(upper_band) - np.min(lower_band))
-    plt.xlim(lbound, rbound)
-    plt.ylim(botbound, topbound)
-    plt.hlines(y = botbound, xmin = lbound, xmax  = replicates[0], color = 'k', lw = 2, linestyle = (1.5, (1.5, 1)))
-    plt.hlines(y = botbound, xmin = replicates[0], xmax  = replicates[-1], color = 'k', lw = 2)
-    plt.hlines(y = botbound, xmin = replicates[-1], xmax = rbound, color = 'k', lw = 2, linestyle = (1.5, (1.5, 1)))
-    plt.fill_between(x_axis, lower_band, upper_band, color = plt_set['bandcolor'], label = 'Confidence band', alpha = 0.25)
-    plt.legend(edgecolor = 'k', loc = 'upper left')
-    plt.text(rbound - 0.04 * (rbound - lbound), botbound + 0.07 * (topbound - botbound), plot_data, 
+    ax.set_xlim(lbound, rbound)
+    ax.set_ylim(botbound, topbound)
+    ax.hlines(y = botbound, xmin = lbound, xmax  = replicates[0], color = 'k', lw = 2, linestyle = (1.5, (1.5, 1)))
+    ax.hlines(y = botbound, xmin = replicates[0], xmax  = replicates[-1], color = 'k', lw = 2)
+    ax.hlines(y = botbound, xmin = replicates[-1], xmax = rbound, color = 'k', lw = 2, linestyle = (1.5, (1.5, 1)))
+    ax.fill_between(x_axis, lower_band, upper_band, color = plt_set['bandcolor'], label = 'Confidence band', alpha = 0.25)
+    ax.legend(edgecolor = 'k', loc = 'upper left')
+    ax.text(rbound - 0.04 * (rbound - lbound), botbound + 0.07 * (topbound - botbound), plot_data, 
                 fontsize = plt_set['legend_fontsize'], verticalalignment = 'bottom', horizontalalignment='right', bbox = props)
-    plt.savefig(outfile, transparent = True, dpi = plt_set['dpi'])
-    plt.clf()
-    mpl.rcParams.update(mpl.rcParamsDefault)
+    return fig

src/BootstrapReport/main.py

@@ -110,6 +110,10 @@ def get_crossings(self, alpha = 0.05, outfile = None, **kwargs):
         :param alpha: 1 - alpha = confidence level for confidence bands
         :param outfile: path to output figure displaying algorithm
         """
+        plt_set = {'fontsize': 18, 'legend_fontsize': 20, 'labelsize': 28, 'linecolor': '#f9665e', 'bandcolor': '#a8d9ed', 'linewidth': 2, 'dpi': 100}
+        for key, value in kwargs.items():
+            plt_set[key] = value
+        
         rep = self.replicates/self.se
         est = self.estimate/self.se
         num_rep, crossings = len(rep), 0
@@ -167,8 +171,11 @@ def left_upper_cb(x):
         self.crossings = crossings
 
         if not outfile == None:
-            helpers.plot_min_crossings(outfile, optimal_path, self.crossings, alpha, rep, est,
-                                       1, upper_cb, lower_cb, left_upper_cb, **kwargs)
+            fig = helpers.plot_min_crossings(optimal_path, self.crossings, alpha, rep, est,
+                                             1, upper_cb, lower_cb, left_upper_cb, plt_set)
+            if type(outfile) == str:
+                fig.savefig(outfile, transparent = True, dpi = plt_set['dpi'])
+            return fig
 
     def pp_plot(self, confidence_band = True, alpha = 0.05, outfile = None, **kwargs):
         """ create the pp plot
@@ -180,6 +187,7 @@ def pp_plot(self, confidence_band = True, alpha = 0.05, outfile = None, **kwargs
         plt_set = {'fontsize': 18, 'legend_fontsize': 20, 'labelsize': 28, 'pointsize': 7, 'pointcolor': '#f9665e', 'bandcolor': '#a8d9ed', 'dpi': 100}
         for key, value in kwargs.items():
             plt_set[key] = value
+
         num_replicates = len(self.replicates)
 
         replicates_eval_normcdf = norm.cdf(self.replicates, self.estimate, self.se)
@@ -197,28 +205,27 @@ def pp_plot(self, confidence_band = True, alpha = 0.05, outfile = None, **kwargs
              'Neg. distance = %.3f' % self.neg_dist))
         props = dict(boxstyle = 'round, pad = 0.75, rounding_size = 0.3', facecolor = 'white', alpha = 0.86)
 
-        plt.rcParams.update({'font.size': plt_set['fontsize']})
-        plt.rcParams.update({'legend.fontsize': plt_set['legend_fontsize']})
-        plt.rcParams.update({'axes.labelsize': plt_set['labelsize']})
-        
-        plt.figure(figsize=(10, 10))
+        plt.rc('font', size = plt_set['fontsize'])
+        plt.rc('legend', fontsize = plt_set['legend_fontsize'])
+        plt.rc('axes', labelsize = plt_set['labelsize'])
+        fig, ax = plt.subplots(figsize = (10, 10))
+
         if confidence_band == True:
-            plt.fill_between(dkw_xgrid, dkw_lbound, dkw_ubound, color = plt_set['bandcolor'], label = 'Confidence band', alpha = 0.35)
-        plt.scatter(replicates_eval_normcdf, replicate_ecdf, s = plt_set['pointsize'],
+            ax.fill_between(dkw_xgrid, dkw_lbound, dkw_ubound, color = plt_set['bandcolor'], label = 'Confidence band', alpha = 0.35)
+        ax.scatter(replicates_eval_normcdf, replicate_ecdf, s = plt_set['pointsize'],
                     c = plt_set['pointcolor'], label='Bootstrap replicates')
-        plt.xlabel("CDF of normal distribution")
-        plt.ylabel("CDF of bootstrap distribution")
-        plt.legend(edgecolor = 'k', loc = 'upper left')
-        plt.axline((0, 0), (1, 1), color="black", linestyle=(0, (5, 5)))
-        plt.text(0.52, 0.06, plot_data, fontsize = plt_set['legend_fontsize'], \
+        ax.set_xlabel("CDF of normal distribution")
+        ax.set_ylabel("CDF of bootstrap distribution")
+        ax.legend(edgecolor = 'k', loc = 'upper left')
+        ax.axline((0, 0), (1, 1), color="black", linestyle=(0, (5, 5)))
+        ax.text(0.52, 0.06, plot_data, fontsize = plt_set['legend_fontsize'], \
             verticalalignment = 'bottom', horizontalalignment='left', bbox = props)
-        plt.ylim(0, 1)
-        plt.xlim(0, 1)
+        ax.set_ylim(0, 1)
+        ax.set_xlim(0, 1)
 
         if not outfile == None:
-            plt.savefig(outfile, transparent = True, dpi = plt_set['dpi'])
-        plt.clf()
-        mpl.rcParams.update(mpl.rcParamsDefault)
+            fig.savefig(outfile, transparent = True, dpi = plt_set['dpi'])
+        return fig
 
 
     def density_plot(self, bounds = None, bandwidth = None, outfile = None, **kwargs):
@@ -230,6 +237,7 @@ def density_plot(self, bounds = None, bandwidth = None, outfile = None, **kwargs
         plt_set = {'fontsize': 18, 'legend_fontsize': 20, 'labelsize': 28, 'linecolor': '#f9665e', 'linewidth': 1, 'dpi': 100}
         for key, value in kwargs.items():
             plt_set[key] = value
+
         if not bandwidth:
             bandwidth = self.best_bandwidth_value
         if bounds != None:
@@ -238,25 +246,24 @@ def density_plot(self, bounds = None, bandwidth = None, outfile = None, **kwargs
             lbound, ubound = self.replicates[0] - 2*self.best_bandwidth_value, self.replicates[-1] + 2*self.best_bandwidth_value
 
         pdf_from_kde = helpers.get_kde(self.replicates, bandwidth)
-
-        xgrid = np.linspace(lbound, ubound, len(self.replicates) * 100)
+        xgrid = np.linspace(lbound, ubound, plt_set['dpi'] * 2)
         density = [pdf_from_kde(x) for x in xgrid]
 
-        plt.rcParams.update({'font.size': plt_set['fontsize']})
-        plt.rcParams.update({'legend.fontsize': plt_set['legend_fontsize']})
-        plt.rcParams.update({'axes.labelsize': plt_set['labelsize']})
+        plt.rc('font', size = plt_set['fontsize'])
+        plt.rc('legend', fontsize = plt_set['legend_fontsize'])
+        plt.rc('axes', labelsize = plt_set['labelsize'])
+        fig, ax = plt.subplots()
 
-        plt.xlim(lbound, ubound)
-        plt.xlabel('Value of object of interest')
-        plt.ylabel('Density')
-        plt.plot(xgrid, density, linewidth = plt_set['linewidth'], color = plt_set['linecolor'])
-        plt.plot([self.replicates[0], self.replicates[-1]], [0.0001, 0.0001], '|k', markeredgewidth = 1, label = 'Range of bootstrap replicates')
-        plt.legend(loc = 'best', fontsize = 'x-small', markerscale = 0.75)
+        ax.set_xlim(lbound, ubound)
+        ax.set_xlabel('Value of object of interest')
+        ax.set_ylabel('Density')
+        ax.plot(xgrid, density, linewidth = plt_set['linewidth'], color = plt_set['linecolor'])
+        ax.plot([self.replicates[0], self.replicates[-1]], [0.0001, 0.0001], '|k', markeredgewidth = 1, label = 'Range of bootstrap replicates')
+        ax.legend(loc = 'best', fontsize = 'x-small', markerscale = 0.75)
 
         if not outfile == None:
-            plt.savefig(outfile, transparent = True, dpi = plt_set['dpi'])
-        plt.clf()
-        mpl.rcParams.update(mpl.rcParamsDefault)
+            fig.savefig(outfile, transparent = True, dpi = plt_set['dpi'])
+        return fig
 
     def get_tv_min(self, init_values = None, optimization_bounds = None, bounds_of_integration = np.inf):
         """

tests/test_BootstrapReport.py

@@ -42,9 +42,6 @@ def test_crossings():
     test_replicates = pd.read_csv('examples/gamma_replicates.csv')['replicate_value'].values
     estimate, standard_error = 0, 1
     test = ObjectOfInterest(estimate = estimate, se = standard_error, replicates = test_replicates)
-    # Test function makes it to `savefig` when plotting outfile
-    with pytest.raises(AttributeError):
-        test.get_crossings(outfile = True)
     test = ObjectOfInterest(estimate = estimate, se = standard_error, \
                             replicates = norm.ppf(np.linspace(0.01, 1, 100), loc = estimate, scale = standard_error))
     test.get_crossings()