You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by jb...@apache.org on 2013/09/01 04:36:17 UTC
[2/7] git commit: Optimize name query performance in wide rows patch
by rbranson; reviewed by jbellis for CASSANDRA-5966
Optimize name query performance in wide rows
patch by rbranson; reviewed by jbellis for CASSANDRA-5966
Project: http://git-wip-us.apache.org/repos/asf/cassandra/repo
Commit: http://git-wip-us.apache.org/repos/asf/cassandra/commit/b14273b4
Tree: http://git-wip-us.apache.org/repos/asf/cassandra/tree/b14273b4
Diff: http://git-wip-us.apache.org/repos/asf/cassandra/diff/b14273b4
Branch: refs/heads/trunk
Commit: b14273b4352ac814946d8e1b9f9be479ddae8d0b
Parents: 1a50215
Author: Jonathan Ellis <jb...@apache.org>
Authored: Sat Aug 31 20:35:19 2013 -0500
Committer: Jonathan Ellis <jb...@apache.org>
Committed: Sat Aug 31 20:35:19 2013 -0500
----------------------------------------------------------------------
CHANGES.txt | 1 +
.../db/columniterator/SSTableNamesIterator.java | 37 ++++++++++++++++----
2 files changed, 31 insertions(+), 7 deletions(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/cassandra/blob/b14273b4/CHANGES.txt
----------------------------------------------------------------------
diff --git a/CHANGES.txt b/CHANGES.txt
index 777d0d1..1943217 100644
--- a/CHANGES.txt
+++ b/CHANGES.txt
@@ -1,4 +1,5 @@
1.2.10
+ * Optimize name query performance in wide rows (CASSANDRA-5966)
* Upgrade metrics-core to version 2.2.0 (CASSANDRA-5947)
* Add snitch, schema version, cluster, partitioner to JMX (CASSANDRA-5881)
* Fix CqlRecordWriter with composite keys (CASSANDRA-5949)
http://git-wip-us.apache.org/repos/asf/cassandra/blob/b14273b4/src/java/org/apache/cassandra/db/columniterator/SSTableNamesIterator.java
----------------------------------------------------------------------
diff --git a/src/java/org/apache/cassandra/db/columniterator/SSTableNamesIterator.java b/src/java/org/apache/cassandra/db/columniterator/SSTableNamesIterator.java
index df28c46..40934d4 100644
--- a/src/java/org/apache/cassandra/db/columniterator/SSTableNamesIterator.java
+++ b/src/java/org/apache/cassandra/db/columniterator/SSTableNamesIterator.java
@@ -153,7 +153,7 @@ public class SSTableNamesIterator extends SimpleAbstractColumnIterator implement
List<OnDiskAtom> result = new ArrayList<OnDiskAtom>();
if (indexList.isEmpty())
{
- readSimpleColumns(file, columns, result);
+ readSimpleColumns(sstable.metadata, file, columns, result);
}
else
{
@@ -175,27 +175,37 @@ public class SSTableNamesIterator extends SimpleAbstractColumnIterator implement
iter = result.iterator();
}
- private void readSimpleColumns(FileDataInput file, SortedSet<ByteBuffer> columnNames, List<OnDiskAtom> result) throws IOException
+ private void readSimpleColumns(CFMetaData metadata,
+ FileDataInput file,
+ SortedSet<ByteBuffer> columnNames,
+ List<OnDiskAtom> result)
+ throws IOException
{
+ AbstractType<?> comparator = metadata.comparator;
OnDiskAtom.Serializer atomSerializer = cf.getOnDiskSerializer();
+ ByteBuffer maximalColumnName = columnNames.last();
int count = file.readInt();
- int n = 0;
+
for (int i = 0; i < count; i++)
{
OnDiskAtom column = atomSerializer.deserializeFromSSTable(file, sstable.descriptor.version);
+ ByteBuffer columnName = column.name();
+
if (column instanceof IColumn)
{
- if (columnNames.contains(column.name()))
+ if (columnNames.contains(columnName))
{
result.add(column);
- if (++n >= columns.size())
- break;
}
}
else
{
result.add(column);
}
+
+ // Already consumed all of this block that's going to have columns that apply to this query.
+ if (comparator.compare(columnName, maximalColumnName) >= 0)
+ break;
}
}
@@ -231,6 +241,12 @@ public class SSTableNamesIterator extends SimpleAbstractColumnIterator implement
{
long positionToSeek = basePosition + indexInfo.offset;
+ // SortedSet.subSet() is end-exclusive, so we special-case that
+ // if it's one of the columns we're looking for
+ ByteBuffer maximalColumnName = columnNames.contains(indexInfo.lastName)
+ ? indexInfo.lastName
+ : columnNames.subSet(indexInfo.firstName, indexInfo.lastName).last();
+
// With new promoted indexes, our first seek in the data file will happen at that point.
if (file == null)
file = createFileDataInput(positionToSeek);
@@ -238,13 +254,20 @@ public class SSTableNamesIterator extends SimpleAbstractColumnIterator implement
OnDiskAtom.Serializer atomSerializer = cf.getOnDiskSerializer();
file.seek(positionToSeek);
FileMark mark = file.mark();
+
// TODO only completely deserialize columns we are interested in
while (file.bytesPastMark(mark) < indexInfo.width)
{
OnDiskAtom column = atomSerializer.deserializeFromSSTable(file, sstable.descriptor.version);
+ ByteBuffer columnName = column.name();
+
// we check vs the original Set, not the filtered List, for efficiency
- if (!(column instanceof IColumn) || columnNames.contains(column.name()))
+ if (!(column instanceof IColumn) || columnNames.contains(columnName))
result.add(column);
+
+ // Already consumed all of this block that's going to have columns that apply to this query.
+ if (comparator.compare(columnName, maximalColumnName) >= 0)
+ break;
}
}
}