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-edit-distance-inline.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-edit-distance-inline.aql
new file mode 100644
index 0000000..78a285a
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-edit-distance-inline.aql
@@ -0,0 +1,31 @@
+/*
+ * Description : Fuzzy self joins a dataset, DBLP, based on the edit-distance function of its authors.
+ * DBLP has a 3-gram index on authors, 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(authors) type ngram(3);
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ngram-edit-distance-inline.adm";
+
+for $a in dataset('DBLP')
+for $b in dataset('DBLP')
+let $ed := edit-distance($a.authors, $b.authors)
+where $ed < 3 and $a.id < $b.id
+return {"aauthors": $a.authors, "bauthors": $b.authors, "ed": $ed}
\ No newline at end of file
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-edit-distance.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-edit-distance.aql
new file mode 100644
index 0000000..b05c0d1
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-edit-distance.aql
@@ -0,0 +1,39 @@
+/*
+ * Description : Fuzzy joins two datasets, DBLP and CSX, based on the edit-distance function of their authors.
+ * DBLP has a 3-gram index on authors, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 type CSXType as closed {
+ id: int32,
+ csxid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create dataset CSX(CSXType) partitioned by key id;
+
+create index ngram_index on DBLP(authors) type ngram(3);
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ngram-edit-distance.adm";
+
+for $a in dataset('DBLP')
+for $b in dataset('CSX')
+where edit-distance($a.authors, $b.authors) < 3 and $a.id < $b.id
+return {"aauthors": $a.authors, "bauthors": $b.authors}
\ No newline at end of file
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-fuzzyeq-edit-distance.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-fuzzyeq-edit-distance.aql
new file mode 100644
index 0000000..87ebe4a
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-fuzzyeq-edit-distance.aql
@@ -0,0 +1,42 @@
+/*
+ * Description : Fuzzy joins two datasets, DBLP and CSX, based on ~= using edit distance of their authors.
+ * DBLP has a 3-gram index on authors, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 type CSXType as closed {
+ id: int32,
+ csxid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create dataset CSX(CSXType) partitioned by key id;
+
+create index ngram_index on CSX(authors) type ngram(3);
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ngram-fuzzyeq-edit-distance.adm";
+
+set simfunction 'edit-distance';
+set simthreshold '3';
+
+for $a in dataset('DBLP')
+for $b in dataset('CSX')
+where $a.authors ~= $b.authors and $a.id < $b.id
+return {"aauthors": $a.authors, "bauthors": $b.authors}
\ No newline at end of file
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-fuzzyeq-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-fuzzyeq-jaccard.aql
new file mode 100644
index 0000000..77aa691
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-fuzzyeq-jaccard.aql
@@ -0,0 +1,42 @@
+/*
+ * Description : Fuzzy joins two datasets, DBLP and CSX, based on ~= using Jaccard of their 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 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 type CSXType as closed {
+ id: int32,
+ csxid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create dataset CSX(CSXType) partitioned by key id;
+
+create index ngram_index on DBLP(title) type ngram(3);
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ngram-fuzzyeq-jaccard.adm";
+
+set simfunction 'jaccard';
+set simthreshold '0.5f';
+
+for $a in dataset('DBLP')
+for $b in dataset('CSX')
+where gram-tokens($a.title, 3, false) ~= gram-tokens($b.title, 3, false) and $a.id < $b.id
+return {"atitle": $a.title, "btitle": $b.title}
\ No newline at end of file
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
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-jaccard.aql
new file mode 100644
index 0000000..1979201
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ngram-jaccard.aql
@@ -0,0 +1,40 @@
+/*
+ * Description : Fuzzy joins two datasets, DBLP and CSX, based on the similarity-jaccard function of their 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 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 type CSXType as closed {
+ id: int32,
+ csxid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create dataset CSX(CSXType) 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.adm";
+
+for $a in dataset('DBLP')
+for $b in dataset('CSX')
+where similarity-jaccard(gram-tokens($a.title, 3, false), gram-tokens($b.title, 3, false)) >= 0.5f
+ and $a.id < $b.id
+return {"atitle": $a.title, "btitle": $b.title}
\ No newline at end of file
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-edit-distance-inline.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-edit-distance-inline.aql
new file mode 100644
index 0000000..25cdc53
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-edit-distance-inline.aql
@@ -0,0 +1,38 @@
+/*
+ * Description : Fuzzy self joins a dataset, Customers, based on the edit-distance function of its interest lists.
+ * Customers has a keyword index on interests, 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: [string],
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_olist-edit-distance-inline.adm";
+
+for $a in dataset('Customers')
+for $b in dataset('Customers')
+let $ed := edit-distance($a.interests, $b.interests)
+where $ed <= 2 and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests, "ed": $ed}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-edit-distance.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-edit-distance.aql
new file mode 100644
index 0000000..e5dec4f
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-edit-distance.aql
@@ -0,0 +1,38 @@
+/*
+ * Description : Fuzzy joins two datasets, Customer and Customer2, based on the edit-distance function of their interest lists.
+ * Customers has a keyword index on interests, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: [string],
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create dataset Customers2(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_olist-edit-distance.adm";
+
+for $a in dataset('Customers')
+for $b in dataset('Customers2')
+where edit-distance($a.interests, $b.interests) <= 2 and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-fuzzyeq-edit-distance.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-fuzzyeq-edit-distance.aql
new file mode 100644
index 0000000..ae24f13
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-fuzzyeq-edit-distance.aql
@@ -0,0 +1,41 @@
+/*
+ * Description : Fuzzy joins two datasets, Customer and Customer2, based on ~= using edit distance of their interest lists.
+ * Customers has a keyword index on interests, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: [string],
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create dataset Customers2(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_olist-fuzzyeq-edit-distance.adm";
+
+set simfunction 'edit-distance';
+set simthreshold '3';
+
+for $a in dataset('Customers')
+for $b in dataset('Customers2')
+where $a.interests ~= $b.interests and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-fuzzyeq-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-fuzzyeq-jaccard.aql
new file mode 100644
index 0000000..f7df554
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-fuzzyeq-jaccard.aql
@@ -0,0 +1,41 @@
+/*
+ * Description : Fuzzy joins two datasets, Customer and Customer2, based on ~= using Jaccard of their interest lists.
+ * Customers has a keyword index on interests, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: [string],
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create dataset Customers2(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_olist-fuzzyeq-jaccard.adm";
+
+set simfunction 'jaccard';
+set simthreshold '0.7f';
+
+for $a in dataset('Customers')
+for $b in dataset('Customers2')
+where $a.interests /*+ indexnl */ ~= $b.interests and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-jaccard-inline.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-jaccard-inline.aql
new file mode 100644
index 0000000..61816b8
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-jaccard-inline.aql
@@ -0,0 +1,38 @@
+/*
+ * Description : Fuzzy self joins a dataset, Customers, based on the similarity-jaccard function of its interest lists.
+ * Customers has a keyword index on interests, 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: [string],
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_olist-jaccard-inline.adm";
+
+for $a in dataset('Customers')
+for $b in dataset('Customers')
+let $jacc := /*+ indexnl */ similarity-jaccard($a.interests, $b.interests)
+where $jacc >= 0.7f and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests, "jacc": $jacc }
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-jaccard.aql
new file mode 100644
index 0000000..d16776a
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/olist-jaccard.aql
@@ -0,0 +1,38 @@
+/*
+ * Description : Fuzzy joins two datasets, Customer and Customer2, based on the similarity-jaccard function of their interest lists.
+ * Customers has a keyword index on interests, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: [string],
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create dataset Customers2(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_olist-jaccard.adm";
+
+for $a in dataset('Customers')
+for $b in dataset('Customers2')
+where /*+ indexnl */ similarity-jaccard($a.interests, $b.interests) >= 0.7f and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-fuzzyeq-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-fuzzyeq-jaccard.aql
new file mode 100644
index 0000000..0d4872b
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-fuzzyeq-jaccard.aql
@@ -0,0 +1,41 @@
+/*
+ * Description : Fuzzy joins two datasets, Customer and Customer2, based on ~= using Jaccard of their interest sets.
+ * Customers has a keyword index on interests, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: {{string}},
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create dataset Customers2(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ulist-fuzzyeq-jaccard.adm";
+
+set simfunction 'jaccard';
+set simthreshold '0.7f';
+
+for $a in dataset('Customers')
+for $b in dataset('Customers2')
+where $a.interests /*+ indexnl */ ~= $b.interests and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-jaccard-inline.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-jaccard-inline.aql
new file mode 100644
index 0000000..c89fdc3
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-jaccard-inline.aql
@@ -0,0 +1,38 @@
+/*
+ * Description : Fuzzy self joins a dataset, Customers, based on the similarity-jaccard function of its interest sets.
+ * Customers has a keyword index on interests, 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: {{string}},
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ulist-jaccard-inline.adm";
+
+for $a in dataset('Customers')
+for $b in dataset('Customers')
+let $jacc := /*+ indexnl */ similarity-jaccard($a.interests, $b.interests)
+where $jacc >= 0.7f and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests, "jacc": $jacc}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-jaccard.aql
new file mode 100644
index 0000000..72d0341
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/ulist-jaccard.aql
@@ -0,0 +1,38 @@
+/*
+ * Description : Fuzzy joins two datasets, Customer and Customer2, based on the similarity-jaccard function of their interest sets.
+ * Customers has a keyword index on interests, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 AddressType as closed {
+ number: int32,
+ street: string,
+ city: string
+}
+
+create type CustomerType as closed {
+ cid: int32,
+ name: string,
+ age: int32?,
+ address: AddressType?,
+ interests: {{string}},
+ children: [ { name: string, age: int32? } ]
+}
+
+create dataset Customers(CustomerType) partitioned by key cid;
+
+create dataset Customers2(CustomerType) partitioned by key cid;
+
+create index interests_index on Customers(interests) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_ulist-jaccard.adm";
+
+for $a in dataset('Customers')
+for $b in dataset('Customers2')
+where /*+ indexnl */ similarity-jaccard($a.interests, $b.interests) >= 0.7f and $a.cid < $b.cid
+return {"ainterests": $a.interests, "binterests": $b.interests}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-fuzzyeq-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-fuzzyeq-jaccard.aql
new file mode 100644
index 0000000..99a3c79
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-fuzzyeq-jaccard.aql
@@ -0,0 +1,42 @@
+/*
+ * Description : Fuzzy joins two datasets, DBLP and CSX, based on ~= using Jaccard of their titles' word tokens.
+ * DBLP has a keyword index on title, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 type CSXType as closed {
+ id: int32,
+ csxid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create dataset CSX(CSXType) partitioned by key id;
+
+create index keyword_index on DBLP(title) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_word-fuzzyeq-jaccard.adm";
+
+set simfunction 'jaccard';
+set simthreshold '0.5f';
+
+for $a in dataset('DBLP')
+for $b in dataset('CSX')
+where word-tokens($a.title) ~= word-tokens($b.title) and $a.id < $b.id
+return {"atitle": $a.title, "btitle": $b.title}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-jaccard-inline.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-jaccard-inline.aql
new file mode 100644
index 0000000..24c66ac
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-jaccard-inline.aql
@@ -0,0 +1,31 @@
+/*
+ * Description : Fuzzy self joins a dataset, DBLP, based on the similarity-jaccard function of its titles' word tokens.
+ * DBLP has a keyword 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 keyword_index on DBLP(title) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_word-jaccard-inline.adm";
+
+for $a in dataset('DBLP')
+for $b in dataset('DBLP')
+let $jacc := similarity-jaccard(word-tokens($a.title), word-tokens($b.title))
+where $jacc >= 0.5f and $a.id < $b.id
+return {"atitle": $a.title, "btitle": $b.title, "jacc": $jacc}
diff --git a/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-jaccard.aql b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-jaccard.aql
new file mode 100644
index 0000000..ef91b89
--- /dev/null
+++ b/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join-noeqjoin/word-jaccard.aql
@@ -0,0 +1,41 @@
+/*
+ * Description : Fuzzy joins two datasets, DBLP and CSX, based on the similarity-jaccard function of their titles' word tokens.
+ * DBLP has a keyword index on title, and we expect the join to be transformed into an indexed nested-loop join.
+ * 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 type CSXType as closed {
+ id: int32,
+ csxid: string,
+ title: string,
+ authors: string,
+ misc: string
+}
+
+create dataset DBLP(DBLPType) partitioned by key id;
+
+create dataset CSX(CSXType) partitioned by key id;
+
+create index keyword_index on DBLP(title) type keyword;
+
+write output to nc1:"rttest/inverted-index-join-noeqjoin_word-jaccard.adm";
+
+for $a in dataset('DBLP')
+for $b in dataset('CSX')
+where similarity-jaccard(word-tokens($a.title), word-tokens($b.title)) >= 0.5f
+ and $a.id < $b.id
+return {"atitle": $a.title, "btitle": $b.title}
+