pints issue 783 plot histogram

README.md

@@ -31,12 +31,12 @@ Some details:
 
 ### Creating plots, running analysis
 
-- Use `./funk plot test_name` to run a plot, or `./funk plot --all` to run all plots
-- Use `./funk analyse test_name` to check if a test passed or failed, or `./funk analyse --all` to check all tests
+- Use `./funk plot test_name` to run a plot, or `./funk plot --all` to run all plots.
+- Use `./funk analyse test_name` to check if a test passed or failed, or `./funk analyse --all` to check all tests.
 
 
 ### Installing
 
-- Functional testing requires Python 3.4 or later
-- When cloning, make sure to add the `--recursive` switch
+- Functional testing requires Python 3.4 or later.
+- When cloning, make sure to add the `--recursive` switch; `git reset --hard` within each submodule might be needed to get it working.
 - To install, use `python3 -m pip install -r requirements.txt`. This makes sure you have all the dependencies you know.

pfunk/plot.py

@@ -17,6 +17,76 @@
 import pfunk
 
 
+def histogram(results, variable, title, xlabel, threshold=None):
+    """
+    Creates and returns a histogram plot for a variable (over commits).
+    """
+    fig = plt.figure(figsize=(11, 4.5))
+    plt.suptitle(title + ' : ' + pfunk.date())
+
+    # Left plot: Variable per commit for all data as 1 histogram
+    plt.subplot(1, 2, 1)
+    fig.autofmt_xdate()
+    x, y, u, m, s = pfunk.gather_statistics_per_commit(results, variable)
+    if len(x) == 0:
+        plt.text(0.5, 0.5, 'No data')
+    else:
+        y = np.asarray(y)
+        last_commit = [i for i, k in enumerate(x) if k == u[-1]]
+        mask_lc = np.ones(y.shape, dtype=bool)
+        mask_lc[last_commit] = False
+
+        n, _, _ = plt.hist(y[mask_lc], bins='auto', color='#607c8e', alpha=0.7,
+                label='All %s commits' % len(u))
+
+        if len(last_commit) > 25:
+            plt.hist(y[last_commit], bins='auto', color='#0504aa', alpha=0.5,
+                    label='Last commit')
+        else:
+            stem_height = [np.max(n)] * len(last_commit)
+            ml, sl, bl = plt.stem(y[last_commit], stem_height,
+                    label='Last commit')
+            plt.setp(ml, color='#0504aa', alpha=0.5)
+            plt.setp(sl, color='#0504aa', alpha=0.5)
+            plt.setp(bl, visible=False)
+    plt.ylabel('Frequency (total %s runs)' % len(y))
+    plt.xlabel(xlabel)
+    plt.legend()
+
+    # Right plot: Same, to most recent data.
+    plt.subplot(1, 2, 2)
+    n_commits = 12
+    fig.autofmt_xdate()
+    x, y, u, m, s = pfunk.gather_statistics_per_commit(results, variable,
+            n=n_commits)
+    if len(x) == 0:
+        plt.text(0.5, 0.5, 'No data')
+    else:
+        y = np.asarray(y)
+        last_commit = [i for i, k in enumerate(x) if k == u[-1]]
+        mask_lc = np.ones(y.shape, dtype=bool)
+        mask_lc[last_commit] = False
+
+        n, _, _ = plt.hist(y[mask_lc], bins='auto', color='#607c8e', alpha=0.7,
+                label='Last %s commits' % n_commits)
+
+        if len(last_commit) > 10:
+            plt.hist(y[last_commit], bins='auto', color='#0504aa', alpha=0.5,
+                    label='Last commit')
+        else:
+            stem_height = [np.max(n)] * len(last_commit)
+            ml, sl, bl = plt.stem(y[last_commit], stem_height,
+                    label='Last commit')
+            plt.setp(ml, color='#0504aa', alpha=0.5)
+            plt.setp(sl, color='#0504aa', alpha=0.5)
+            plt.setp(bl, visible=False)
+    plt.ylabel('Frequency (total %s runs)' % len(y))
+    plt.xlabel(xlabel)
+    plt.legend()
+
+    return fig
+
+
 def variable(results, variable, title, ylabel, threshold=None):
     """
     Creates and returns a default plot for a variable vs commits.

pfunk/tests/mcmc_banana.py

@@ -157,6 +157,14 @@ def _plot(self, results):
             'Kullback-Leibler divergence', 3 * self._pass_threshold)
         )
 
+        # Figure: KL histogram
+        figs.append(pfunk.plot.histogram(
+            results,
+            'kld',
+            'Banana w. ' + self._method,
+            'Kullback-Leibler divergence')
+        )
+
         # Figure: KL over time
         figs.append(pfunk.plot.convergence(
             results,
@@ -176,4 +184,12 @@ def _plot(self, results):
             'Effective sample size')
         )
 
+        # Figure: ESS histogram
+        figs.append(pfunk.plot.histogram(
+            results,
+            'ess',
+            'Banana w. ' + self._method,
+            'Effective sample size')
+        )
+
         return figs