url summary: show right and wrong parse counts for each regex

Previously this command only showed total counts for each regular
expression.  This doesn't give you a sense of which regexes are working
well and which ones are not.  We now display the number of right, wrong,
and total URL parses per regex.

It's easier to see where we might improve the URL parsing with this
change.

1 files changed, 30 insertions, 13 deletions

diff --git a/lib/spack/spack/cmd/url.py b/lib/spack/spack/cmd/url.py
index f1ce050a90..a24fcc575c 100644
--- a/lib/spack/spack/cmd/url.py
+++ b/lib/spack/spack/cmd/url.py
@@ -160,10 +160,13 @@ def url_summary(args):
     correct_versions = 0
 
     # Collect statistics on which regexes were matched and how often
-    name_regex_dict    = dict()
-    name_count_dict    = defaultdict(int)
-    version_regex_dict = dict()
-    version_count_dict = defaultdict(int)
+    name_regex_dict     = dict()
+    right_name_count    = defaultdict(int)
+    wrong_name_count    = defaultdict(int)
+
+    version_regex_dict  = dict()
+    right_version_count = defaultdict(int)
+    wrong_version_count = defaultdict(int)
 
     tty.msg('Generating a summary of URL parsing in Spack...')
 
@@ -189,9 +192,11 @@ def url_summary(args):
             try:
                 version, vs, vl, vi, vregex = parse_version_offset(url)
                 version_regex_dict[vi] = vregex
-                version_count_dict[vi] += 1
                 if version_parsed_correctly(pkg, version):
                     correct_versions += 1
+                    right_version_count[vi] += 1
+                else:
+                    wrong_version_count[vi] += 1
             except UndetectableVersionError:
                 pass
 
@@ -199,9 +204,11 @@ def url_summary(args):
             try:
                 name, ns, nl, ni, nregex = parse_name_offset(url, version)
                 name_regex_dict[ni] = nregex
-                name_count_dict[ni] += 1
                 if name_parsed_correctly(pkg, name):
                     correct_names += 1
+                    right_name_count[ni] += 1
+                else:
+                    wrong_name_count[ni] += 1
             except UndetectableNameError:
                 pass
 
@@ -216,24 +223,34 @@ def url_summary(args):
     tty.msg('Statistics on name regular expressions:')
 
     print()
-    print('    Index  Count  Regular Expression')
+    print('    Index   Right   Wrong   Total   Regular Expression')
     for ni in sorted(name_regex_dict.keys()):
-        print('    {0:>3}: {1:>6}   r{2!r}'.format(
-            ni, name_count_dict[ni], name_regex_dict[ni]))
+        print('    {0:>5}   {1:>5}   {2:>5}   {3:>5}   r{4!r}'.format(
+            ni,
+            right_name_count[ni],
+            wrong_name_count[ni],
+            right_name_count[ni] + wrong_name_count[ni],
+            name_regex_dict[ni])
+        )
     print()
 
     tty.msg('Statistics on version regular expressions:')
 
     print()
-    print('    Index  Count  Regular Expression')
+    print('    Index   Right   Wrong   Total   Regular Expression')
     for vi in sorted(version_regex_dict.keys()):
-        print('    {0:>3}: {1:>6}   r{2!r}'.format(
-            vi, version_count_dict[vi], version_regex_dict[vi]))
+        print('    {0:>5}   {1:>5}   {2:>5}   {3:>5}   r{4!r}'.format(
+            vi,
+            right_version_count[vi],
+            wrong_version_count[vi],
+            right_version_count[vi] + wrong_version_count[vi],
+            version_regex_dict[vi])
+        )
     print()
 
     # Return statistics, only for testing purposes
     return (total_urls, correct_names, correct_versions,
-            name_count_dict, version_count_dict)
+            right_name_count, right_version_count)
 
 
 def url_stats(args):