1 files changed, 125 insertions, 0 deletions
diff --git a/src/mongo/db/exec/oplogstart.h b/src/mongo/db/exec/oplogstart.h
new file mode 100644
index 00000000000..159e8c82a2a
--- /dev/null
+++ b/src/mongo/db/exec/oplogstart.h
@@ -0,0 +1,125 @@
+/**
+ *    Copyright (C) 2013 10gen Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/db/diskloc.h"
+#include "mongo/db/exec/collection_scan.h"
+#include "mongo/db/exec/plan_stage.h"
+#include "mongo/db/matcher/expression.h"
+#include "mongo/util/timer.h"
+
+namespace mongo {
+
+    class NamespaceDetails;
+
+    /**
+     * OplogStart walks a collection backwards to find the first object in the collection that
+     * matches the query.  It's used by replication to efficiently find where the oplog should be
+     * replayed from.
+     *
+     * The oplog is always a capped collection.  In capped collections, documents are oriented on
+     * disk according to insertion order.  The oplog inserts documents with increasing timestamps.
+     * Queries on the oplog look for entries that are after a certain time.  Therefore if we
+     * navigate backwards, the last document we encounter that satisfies our query (over the
+     * timestamp) is the first document we must scan from to answer the query.
+     *
+     * Why isn't this a normal reverse table scan, you may ask?  We could be correct if we used a
+     * normal reverse collection scan.  However, that's not fast enough.  Since we know all
+     * documents are oriented on disk in insertion order, we know all documents in one extent were
+     * inserted before documents in a subsequent extent.  As such we can skip through entire extents
+     * looking only at the first document.
+     *
+     * Why is this a stage?  Because we want to yield, and we want to be notified of DiskLoc
+     * invalidations.  :(
+     */
+    class OplogStart : public PlanStage {
+    public:
+        // Does not take ownership.
+        OplogStart(const string& ns, MatchExpression* filter, WorkingSet* ws);
+        virtual ~OplogStart();
+
+        virtual StageState work(WorkingSetID* out);
+        virtual bool isEOF();
+
+        virtual void invalidate(const DiskLoc& dl, InvalidationType type);
+        virtual void prepareToYield();
+        virtual void recoverFromYield();
+
+        // PS. don't call this.
+        virtual PlanStageStats* getStats() { return NULL; }
+
+        // For testing only.
+        void setBackwardsScanTime(int newTime) { _backwardsScanTime = newTime; }
+        bool isExtentHopping() { return _extentHopping; }
+        bool isBackwardsScanning() { return _backwardsScanning; }
+    private:
+        // Copied verbatim.
+        static DiskLoc prevExtentFirstLoc(NamespaceDetails* nsd, const DiskLoc& rec);
+
+        StageState workBackwardsScan(WorkingSetID* out);
+
+        void switchToExtentHopping();
+
+        StageState workExtentHopping(WorkingSetID* out);
+
+        // If we're backwards scanning we just punt to a collscan.
+        scoped_ptr<CollectionScan> _cs;
+
+        // What's our current DiskLoc?  Set by both collscan and extent hopping.
+        // Only written by collscan, read and written by extent hopping.
+        DiskLoc _curloc;
+
+        // Have we done our heavy init yet?
+        bool _needInit;
+
+        // Our first state: going backwards via a collscan.
+        bool _backwardsScanning;
+
+        // Our second state: hopping backwards extent by extent.
+        bool _extentHopping;
+
+        // Our final state: done.
+        bool _done;
+
+        NamespaceDetails* _nsd;
+
+        // We only go backwards via a collscan for a few seconds.
+        Timer _timer;
+
+        // WorkingSet is not owned by us.
+        WorkingSet* _workingSet;
+
+        string _ns;
+        
+        MatchExpression* _filter;
+
+        static int _backwardsScanTime;
+    };
+
+}  // namespace mongo