Added tests for surrogate-based indexed NL fuzzy joins, where the top-level join is unecessary and optimized away.
git-svn-id: https://asterixdb.googlecode.com/svn/branches/asterix_fuzzy_perf@891 eaa15691-b419-025a-1212-ee371bd00084
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-jaccard-inline.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-jaccard-inline.aql
new file mode 100644
index 0000000..9f6e66c
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-jaccard-inline.aql
@@ -0,0 +1,31 @@
+/*
+ * Description : Fuzzy self joins a dataset, DBLP, based on the similarity-jaccard function of its titles' 3-gram tokens.
+ * DBLP has a 3-gram index on title, and we expect the join to be transformed into an indexed nested-loop join.
+ * We test the inlining of variables that enable the select to be pushed into the join for subsequent optimization with an index.
+ * We expect the top-level equi join introduced because of surrogate optimization to be removed, since it is not necessary.
+ * Success : Yes
+ */
+
+drop dataverse test if exists;
+create dataverse test;
+use dataverse test;
+
+create type DBLPType as closed {
+ id: int32,
+ dblpid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create index ngram_index on DBLP(title) type ngram(3);
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ngram-jaccard-inline.adm";
+
+for $a in dataset('DBLP')
+for $b in dataset('DBLP')
+let $jacc := similarity-jaccard(gram-tokens($a.title, 3, false), gram-tokens($b.title, 3, false))
+where $jacc >= 0.5f and $a.id < $b.id
+return {"atitle": $a.title, "btitle": $b.title, "jacc": $jacc}
\ No newline at end of file