Produce proper commit objects when iterating over all commits

The iter_individual_commits method that allowed iterating over
all the commits stored in the database was producing the commits
as they are stored in the database. This meant that the line_counts
and test_counts attributes weren't populated, which reduced the
utility of the regression checking.

Instead of database-level commits, produce the processed commit
objects that have the line and test counts computed. This added
two new requirements:
1) Implement __eq__ and __hash__ for authors so that the
comparison in regression checking works for the count
dictionaries.
2) Load the Author relationship of Commit eagerly so that the
author of a commit is available afterwards.

Author: Jaakko Kangasharju

githammer/dbtypes.py

@@ -74,6 +74,12 @@ def name(self):
         else:
             return None
 
+    def __eq__(self, other):
+        return self.canonical_name == other.canonical_name and self.aliases == other.aliases
+
+    def __hash__(self):
+        return hash(self.canonical_name)
+
     def __repr__(self):
         return self.name
 
@@ -89,7 +95,7 @@ class Commit(Base):
     commit_time_utc_offset = Column(Integer, nullable=False)
     parent_ids = Column(JSONType)
 
-    author = relationship('Author', back_populates='commits')
+    author = relationship('Author', back_populates='commits', lazy='joined')
 
     def __init__(self, **kwargs):
         super(Commit, self).__init__(**kwargs)

githammer/hammer.py

@@ -343,5 +343,5 @@ def iter_individual_commits(self):
         self._fail_unless_database_exists()
         session = self._Session()
         for commit in session.query(Commit).order_by(Commit.commit_time):
-            yield commit
+            yield self._shas_to_commits.get(commit.hexsha)
         session.close()