| /* |
| * 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} |