Changeset 170 in tracdarcs

Timestamp:

05/08/10 17:14:29 (21 months ago)

Author:

lele@…

Hash name:

20100508151429-97f81-d8e79c8c9d05a79fd74262ee17eda4f47a2164a5

Message:

Keep a per file per revision cache of darcs annotate

Location:

tracdarcs

Files:

• components.py (modified) (3 diffs)
• repository.py (modified) (3 diffs)

tracdarcs/components.py

@@ -108 +108 @@
             c.execute('SELECT repo_id,node_id,rev,content,size '
                       'FROM darcs_cache LIMIT 1')
+            c.execute('SELECT repo_id,node_id,rev '
+                      'FROM darcs_annotate_cache LIMIT 1')
             return False
         except:
@@ -129 +131 @@
         drop_table('darcs_node_changes')
         drop_table('darcs_cache')
+        drop_table('darcs_annotate_cache')
 
         connector = DatabaseManager(self.env)._get_connector()[0]
@@ -161 +164 @@
             Column('content',type=blobtype),
             Column('size',type='int')]
+        ann_cache_table = Table('darcs_annotate_cache', key=('repo_id','node_id','rev','up_to_line'))[
+            Column('repo_id',type='int'),
+            Column('node_id',type='int'),
+            Column('rev',type='int'),
+            Column('up_to_line',type='int'),
+            Column('blame_rev',type='int')]
         c = db.cursor()
-        for t in [rev_table,node_table,change_table,cache_table]:
+        for t in [rev_table,node_table,change_table,cache_table,ann_cache_table]:
             for stmt in connector.to_sql(t):
                 c.execute(stmt)

tracdarcs/repository.py

@@ -372 +372 @@
         """Provide detailed backward history for the content of this Node.
 
-        Retrieve an array of revisions parsing `darcs annotate`.
+        Retrieve an array of revisions parsing `darcs annotate`. Since
+        that is (still) not fast enough for some repository, we write
+        a cache of the information: a future annotate on the same file
+        at the same revision won't reexecute `darcs annotate`.
         """
 
@@ -380 +383 @@
 
         c = self.repos.db.cursor()
+
+        # Since darcs is faster and faster in building the content
+        # of a file for more and more recent changes, compute the
+        # optimal revision to build the cache of
+        crev = self._get_cached_rev()
+        if crev is None:
+            crev = self.rev
+
+        # Check if the annotate cache is already present
+        c.execute('SELECT up_to_line,blame_rev FROM darcs_annotate_cache '
+                  'WHERE repo_id = %s AND node_id = %s AND rev = %s '
+                  'ORDER BY up_to_line', (self.repos.id, self.__node_id, crev))
+        row = c.fetchone()
+        if row is not None:
+            self.__log.debug('Annotate cache hit for %s at rev %s', self.path, crev)
+            revs = []
+            line = 0
+            # Expand the cache, producing a list of revisions, one per line
+            while row is not None:
+                while line<row[0]:
+                    revs.append(row[1])
+                    line += 1
+                row = c.fetchone()
+            return revs
+
+        # No cache, build it
 
         class DarcsXMLAnnotateHandler(ContentHandler):
@@ -424 +453 @@
         parser.close()
 
-        return handler.revisions
+        revs = handler.revisions
+
+        # Write a compressed representation
+
+        self.__log.debug('Writing annotate cache for %s at rev %s', self.path, crev)
+        prev = None
+        for i,rev in enumerate(revs):
+            if prev is not None:
+                if prev != rev:
+                    c.execute('INSERT INTO darcs_annotate_cache (repo_id,node_id,rev,up_to_line,blame_rev) '
+                              'VALUES (%s,%s,%s,%s,%s)',
+                              (self.repos.id, self.__node_id, crev, i, prev))
+                    prev = rev
+                    lastline = i
+            else:
+                prev = rev
+                lastline = 0
+        if lastline != len(revs):
+            c.execute('INSERT INTO darcs_annotate_cache (repo_id,node_id,rev,up_to_line,blame_rev) '
+                      'VALUES (%s,%s,%s,%s,%s)',
+                      (self.repos.id, self.__node_id, crev, len(revs), revs[-1]))
+
+        return revs
 
     def get_properties(self):