# HG changeset patch
# User Paul Boddie <paul@boddie.org.uk>
# Date 1317420064 -7200
# Node ID 6397ceae2ddaa3aa1ea4c2d28a03338d580ff979
# Parent  baee3f51e8f51bee6308dec3182d74f6a0e33b8a
Added support for duplicate keys so that an index will always refer to the first
record of any group of records sharing such a key, thus ensuring that use of the
index does not cause records to be missed (because some of them occur before the
referenced record).

diff -r baee3f51e8f5 -r 6397ceae2dda simplex.py
--- a/simplex.py	Fri Sep 30 23:46:27 2011 +0200
+++ b/simplex.py	Sat Oct 01 00:01:04 2011 +0200
@@ -65,9 +65,24 @@
     l = []
     pos = 0
 
+    current_key = None
+    start_pos = 0
+
     for i, record in enumerate(f.get_records()):
+        key = f.get_key(record)
+
+        # Where duplicate keys are permitted, the first record employing the key
+        # must be available as an index entry. Otherwise, records preceding the
+        # one referenced by the entry may have the same key and be missed when
+        # seeking using the index.
+
+        if key != current_key:
+            current_key = key
+            start_pos = pos
+
         if i % interval == 0:
-            l.append((f.get_key(record), pos))
+            l.append((current_key, start_pos))
+
         pos += len(record)
 
     return l