Changeset 673 for hodgestar/PythonCode/SvnDots

Timestamp:

10/24/09 10:48:28 (3 years ago)

Author:

hodgestar

Message:

Add custom path ordering optimization (reduces distance between consecutive commits by the same user).

Location:

hodgestar/PythonCode/SvnDots

Files:

• fa-commits-date.png (modified) (previous)
• fa-commits-rev.png (modified) (previous)
• svndots.py (modified) (5 diffs)

hodgestar/PythonCode/SvnDots/svndots.py

@@ -6 +6 @@
 import math
 import random
+import os
+import logging
 
 class Rev(object):
@@ -250 +252 @@
     """Plot a list of revision objects"""
     import matplotlib.pyplot as plt
+    import pylab
+
+    common = len(os.path.commonprefix(paths))
+    paths = [p[common:] for p in paths]
+    for user, points in data.items():
+        data[user] = [(d[0], d[1][common:]) for d in points]
 
     pathdict = dict(zip(paths, range(len(paths))))
@@ -257 +265 @@
     if is_date:
         plotter = plt.plot_date
+        xgetter = lambda r: r.date
     else:
         plotter = plt.plot
+        xgetter = lambda r: r.revid
 
     for user in users:
         points = data[user]
-        x = [p[0] for p in points]
-        y = [pathdict[p[1]]+0.2*userdict[user]-0.4 for p in points]
+        x = [xgetter(p[0]) for p in points]
+        y = [pathdict[p[1]]+0.1*userdict[user]-0.2 for p in points]
+        plotter(x, y,
+            usercolor[user] + '-',
+            alpha=0.25,
+            linewidth=1.0,
+        )
         plotter(x, y,
             usercolor[user] + linestyle,
             label=user,
-            markersize=4.5,
+            markersize=8.0,
             linewidth=1.0,
+            picker=4.0,
         )
 
-    plt.yticks([y for y in range(len(paths))], paths, size='xx-small')
-    plt.grid(color='grey', linestyle='-', alpha=0.25, ydata=[y for y in range(0,len(paths)+1,2)], linewidth=20.0)
+    plt.yticks([y for y in range(len(paths))], paths)
+    plt.grid(color='grey', linestyle='-', alpha=0.25, ydata=[y for y in range(0,len(paths)+1,2)], linewidth=24.0)
     plt.ylabel('File')
     plt.xlabel(is_date and 'Date' or 'Revision')
     plt.ylim(-0.5,len(paths)-0.5)
     plt.legend(loc=0)
+
+    fig = plt.figure(1)
+    xmin, xmax = plt.xlim()
+    ymin, ymax = plt.ylim()
+    ann = plt.annotate("default", (xmin + (xmax-xmin)*0.05, ymin + (ymax-ymin)*0.05),
+        backgroundcolor="yellow",
+        bbox=dict(boxstyle="round", fc="0.8"),
+        visible=False,
+    )
+    #ann.set_animated(True)
+    current_rev = [-1]
+
+    def onpick(event):
+        line = event.artist
+        ind = event.ind[0]
+        x, y = line.get_xdata()[ind], line.get_ydata()[ind]
+        user = line.get_label()
+        rev = data[user][ind][0]
+        comment = list(rev.comments[user])[0]
+        text = "%s [%s]: %s" % (user, rev.revid, comment)
+        text = text[:150]
+
+        if ann.get_visible() and current_rev[0] == rev.revid:
+            # print "II:", rev.revid
+            ann.set_visible(False)
+            current_rev[0] = -1
+        else:
+            # print "VV:", rev.revid
+            ann.set_text(text)
+            ann.set_visible(True)
+            current_rev[0] = rev.revid
+
+        pylab.draw()
+
+    cid = fig.canvas.mpl_connect('pick_event', onpick)
+
     plt.show()
+
+def minimize_rev_separation(revs, paths):
+    """Attempt to minimize the maximum distance between paths in
+       a revision.
+       """
+    import numpy
+
+    pathindex = dict(zip(paths, range(len(paths))))
+    revindex = []
+    userindex = {}
+    for rev in revs:
+        for r_user, r_paths in rev.contributions.items():
+            idxs = [pathindex[r] for r in r_paths if r in pathindex]
+            if idxs:
+                revindex.append(idxs)
+                userindex.setdefault(r_user,[])
+                userindex[r_user].extend(idxs)
+    userindex = userindex.values()
+
+    allindex = userindex
+
+    def func(x):
+        tot = 0.0
+        for idxs in allindex:
+            places = x.take(idxs)
+            tot += sum((places[1:] - places[:-1])**2)
+        return tot
+
+    x0 = numpy.array(range(len(paths)), dtype=float)
+    best_tot = func(x0)
+    path_min = x0.copy()
+    logging.info("Initial objective function value: %s", best_tot)
+    for i in range(1000):
+        random.shuffle(x0)
+        tot = func(x0)
+        if tot < best_tot:
+            best_tot = tot
+            path_min = x0.copy()
+    logging.info("Final objective function value:", best_tot)
+
+    path_pairs = zip(paths, path_min)
+    path_pairs.sort(key=lambda p: p[1])
+    return [p[0] for p in path_pairs]
 
 def main(args):
@@ -292 +387 @@
                       default=42,
                       help="set random module seed.")
+    parser.add_option("-f", "--filter", dest="filter",
+                      default=r".*\.py$",
+                      help="regular expression filter for file names")
 
     (options, args) = parser.parse_args()
@@ -312 +410 @@
     path_vecs = {}
 
+    path_revs = {}
+
+    path_re = re.compile(options.filter)
+
     for i, rev in enumerate(revs):
         for r_user, r_paths in rev.contributions.iteritems():
-            r_paths = [p for p in r_paths if p.endswith('.py')]
+            r_paths = [p for p in r_paths if path_re.match(p)]
             users.add(r_user)
             paths.update(r_paths)
 
             data.setdefault(r_user, [])
-
-            if options.bydate:
-                data[r_user].extend((rev.date, p) for p in r_paths)
-            else:
-                data[r_user].extend((rev.revid, p) for p in r_paths)
+            data[r_user].extend((rev, p) for p in r_paths)
 
             for r_path in r_paths:
-                path_vecs.setdefault(r_path, {})
-                path_vecs[r_path].setdefault(r_user, 0.0)
-                path_vecs[r_path][r_user] += 1.0
+                path_revs.setdefault(r_path, [])
+                path_revs[r_path].append(i)
+
+#            for r_path in r_paths:
+#                path_vecs.setdefault(r_path, {})
+#                path_vecs[r_path].setdefault(r_user, 0.0)
+#                path_vecs[r_path][r_user] += 1.0
 
     users = sorted(users)
-    paths = sorted(paths)
-
-    path_vecs = path_vecs.items()
-    path_names = [pv[0] for pv in path_vecs]
-    path_dicts = [pv[1] for pv in path_vecs]
-    path_vectors = [Vector([u in d and d[u] or 0.0 for u in users]) for d in path_dicts]
-    path_vectors = [v*(1.0/sum(v)) for v in path_vectors]
-
-    means, clusters = k_means(path_vectors, len(users)+2, 100, Vector.euclidian_distance)
-
-    clusters.sort(key=lambda c: len(c))
-
-    ordered_paths = []
-    print "Clusters:"
-    for cluster in clusters:
-        print "\t", [path_names[i] for i in cluster]
-        ordered_paths.extend(path_names[i] for i in cluster)
+
+    # paths = sorted(paths, key=lambda p: (len(path_revs[p]), max(path_revs[p])))
+    # ordered_paths = paths
+
+    ordered_paths = minimize_rev_separation(revs, paths)
+
+#    path_vecs = path_vecs.items()
+#    path_names = [pv[0] for pv in path_vecs]
+#    path_dicts = [pv[1] for pv in path_vecs]
+#    path_vectors = [Vector([u in d and d[u] or 0.0 for u in users]) for d in path_dicts]
+#    path_vectors = [v*(1.0/sum(v)) for v in path_vectors]
+#
+#    means, clusters = k_means(path_vectors, len(users)+2, 100, Vector.euclidian_distance)
+#
+#    clusters.sort(key=lambda c: len(c))
+#
+#    ordered_paths = []
+#    print "Clusters:"
+#    for cluster in clusters:
+#        print "\t", [path_names[i] for i in cluster]
+#        ordered_paths.extend(path_names[i] for i in cluster)
 
     plotrevs(users, ordered_paths, data, options.bydate, options.linestyle)