Add BloomFilter skipping index SQL support (#283)

* Add bloom filter in grammar and IT Signed-off-by: Chen Dai <daichen@amazon.com> * Update user manual Signed-off-by: Chen Dai <daichen@amazon.com> * Add more IT Signed-off-by: Chen Dai <daichen@amazon.com> * Update doc Signed-off-by: Chen Dai <daichen@amazon.com> --------- Signed-off-by: Chen Dai <daichen@amazon.com>
opensearch-project · Mar 18, 2024 · 5a2df20 · 5a2df20
1 parent 7382c95
commit 5a2df20
Show file tree

Hide file tree

Showing 5 changed files with 80 additions and 5 deletions.
diff --git a/docs/index.md b/docs/index.md
@@ -153,7 +153,20 @@ High level API is dependent on query engine implementation. Please see Query Eng
 
 #### Skipping Index
 
-The default maximum size for the value set is 100. In cases where a file contains columns with high cardinality values, the value set will become null. This is the trade-off that prevents excessive memory consumption at the cost of not skipping the file.
+Provided below are the explanations for the parameters of the skipping algorithm. You can find the default values in the function signature below:
+
+- **VALUE_SET(limit=100):** If the column values of a file has higher cardinality than the limit (optional, default is 100), the value set will become null. This trade-off prevents excessive memory consumption at the expense of not skipping the file.
+
+- **BLOOM_FILTER**
+  - **BLOOM_FILTER(num_candidate=10, fpp=0.03):** By default, the adaptive BloomFilter algorithm is used. Users can configure:
+    1. The number of candidates (optional), starting with an expected number of distinct items at 1024 and doubling.
+    2. The false positive probability of each candidate (optional).
+    3. Examples: `BLOOM_FILTER`, `BLOOM_FILTER(20), BLOOM_FILTER(20, 0.01)`
+
+  - **BLOOM_FILTER(false, num_items=10000, fpp=0.03):** Setting the first parameter to `false` will revert to the non-adaptive algorithm. Users can configure:
+    1. The expected number of distinct values (optional).
+    2. The false positive probability (optional).
+    3. Examples: `BLOOM_FILTER(false)`, `BLOOM_FILTER(false, 1000000)`, `BLOOM_FILTER(false, 1000000, 0.01)`
 
 ```sql
 CREATE SKIPPING INDEX [IF NOT EXISTS]

diff --git a/flint-spark-integration/src/main/antlr4/FlintSparkSqlExtensions.g4 b/flint-spark-integration/src/main/antlr4/FlintSparkSqlExtensions.g4
@@ -183,7 +183,7 @@ indexColTypeList
     ;
 
 indexColType
-    : identifier skipType=(PARTITION | VALUE_SET | MIN_MAX)
+    : identifier skipType=(PARTITION | VALUE_SET | MIN_MAX | BLOOM_FILTER)
         (LEFT_PAREN skipParams RIGHT_PAREN)?
     ;
 

diff --git a/flint-spark-integration/src/main/antlr4/SparkSqlBase.g4 b/flint-spark-integration/src/main/antlr4/SparkSqlBase.g4
@@ -139,6 +139,7 @@ nonReserved
 
 // Flint lexical tokens
 
+BLOOM_FILTER: 'BLOOM_FILTER';
 MIN_MAX: 'MIN_MAX';
 SKIPPING: 'SKIPPING';
 VALUE_SET: 'VALUE_SET';

diff --git a/...ain/scala/org/opensearch/flint/spark/sql/skipping/FlintSparkSkippingIndexAstBuilder.scala b/...ain/scala/org/opensearch/flint/spark/sql/skipping/FlintSparkSkippingIndexAstBuilder.scala
@@ -8,10 +8,11 @@ package org.opensearch.flint.spark.sql.skipping
 import scala.collection.JavaConverters.collectionAsScalaIterableConverter
 
 import org.antlr.v4.runtime.tree.RuleNode
+import org.opensearch.flint.core.field.bloomfilter.BloomFilterFactory._
 import org.opensearch.flint.spark.FlintSpark
 import org.opensearch.flint.spark.skipping.FlintSparkSkippingIndex
 import org.opensearch.flint.spark.skipping.FlintSparkSkippingStrategy.SkippingKind
-import org.opensearch.flint.spark.skipping.FlintSparkSkippingStrategy.SkippingKind.{MIN_MAX, PARTITION, VALUE_SET}
+import org.opensearch.flint.spark.skipping.FlintSparkSkippingStrategy.SkippingKind.{BLOOM_FILTER, MIN_MAX, PARTITION, VALUE_SET}
 import org.opensearch.flint.spark.skipping.valueset.ValueSetSkippingStrategy.VALUE_SET_MAX_SIZE_KEY
 import org.opensearch.flint.spark.sql.{FlintSparkSqlCommand, FlintSparkSqlExtensionsVisitor, SparkSqlAstBuilder}
 import org.opensearch.flint.spark.sql.FlintSparkSqlAstBuilder.{getFullTableName, getSqlText}
@@ -52,6 +53,19 @@ trait FlintSparkSkippingIndexAstBuilder extends FlintSparkSqlExtensionsVisitor[A
             val valueSetParams = (Seq(VALUE_SET_MAX_SIZE_KEY) zip skipParams).toMap
             indexBuilder.addValueSet(colName, valueSetParams)
           case MIN_MAX => indexBuilder.addMinMax(colName)
+          case BLOOM_FILTER =>
+            // Determine if the given parameters are for adaptive algorithm by the first parameter
+            val bloomFilterParamKeys =
+              if (skipParams.headOption.exists(_.equalsIgnoreCase("false"))) {
+                Seq(
+                  BLOOM_FILTER_ADAPTIVE_KEY,
+                  CLASSIC_BLOOM_FILTER_NUM_ITEMS_KEY,
+                  CLASSIC_BLOOM_FILTER_FPP_KEY)
+              } else {
+                Seq(ADAPTIVE_NUMBER_CANDIDATE_KEY, CLASSIC_BLOOM_FILTER_FPP_KEY)
+              }
+            val bloomFilterParams = (bloomFilterParamKeys zip skipParams).toMap
+            indexBuilder.addBloomFilter(colName, bloomFilterParams)
         }
       }
 

diff --git a/integ-test/src/test/scala/org/opensearch/flint/spark/FlintSparkSkippingIndexSqlITSuite.scala b/integ-test/src/test/scala/org/opensearch/flint/spark/FlintSparkSkippingIndexSqlITSuite.scala
@@ -8,8 +8,9 @@ package org.opensearch.flint.spark
 import scala.Option.empty
 import scala.collection.JavaConverters.{mapAsJavaMapConverter, mapAsScalaMapConverter}
 
+import com.stephenn.scalatest.jsonassert.JsonMatchers.matchJson
 import org.json4s.{Formats, NoTypeHints}
-import org.json4s.native.JsonMethods.parse
+import org.json4s.native.JsonMethods.{compact, parse, render}
 import org.json4s.native.Serialization
 import org.opensearch.flint.core.FlintOptions
 import org.opensearch.flint.core.storage.FlintOpenSearchClient
@@ -46,7 +47,8 @@ class FlintSparkSkippingIndexSqlITSuite extends FlintSparkSuite {
            | (
            |   year PARTITION,
            |   name VALUE_SET,
-           |   age MIN_MAX
+           |   age MIN_MAX,
+           |   address BLOOM_FILTER
            | )
            | WITH (auto_refresh = true)
            | """.stripMargin)
@@ -80,6 +82,51 @@ class FlintSparkSkippingIndexSqlITSuite extends FlintSparkSuite {
       ))
   }
 
+  test("create skipping index with non-adaptive bloom filter") {
+    sql(s"""
+           | CREATE SKIPPING INDEX ON $testTable
+           | (
+           |   address BLOOM_FILTER(false, 1024, 0.01)
+           | )
+           | WITH (auto_refresh = true)
+           | """.stripMargin)
+
+    val job = spark.streams.active.find(_.name == testIndex)
+    awaitStreamingComplete(job.get.id.toString)
+
+    flint.queryIndex(testIndex).count() shouldBe 2
+
+    checkAnswer(sql(s"SELECT name FROM $testTable WHERE address = 'Vancouver'"), Row("Test"))
+    sql(s"SELECT name FROM $testTable WHERE address = 'San Francisco'").count() shouldBe 0
+  }
+
+  Seq(
+    (
+      s"CREATE SKIPPING INDEX ON $testTable (address BLOOM_FILTER(20, 0.01))",
+      """
+        |{
+        |   "adaptive": "true",
+        |   "num_candidates": "20",
+        |   "fpp": "0.01"
+        |}
+        |""".stripMargin),
+    (
+      s"CREATE SKIPPING INDEX ON $testTable (address BLOOM_FILTER(false, 100000, 0.001))",
+      """
+        |{
+        |   "adaptive": "false",
+        |   "num_items": "100000",
+        |   "fpp": "0.001"
+        |}
+        |""".stripMargin)).foreach { case (query, expectedParamJson) =>
+    test(s"create skipping index with bloom filter parameters $expectedParamJson") {
+      sql(query)
+      val metadata = flint.describeIndex(testIndex).get.metadata().getContent
+      val parameters = compact(render(parse(metadata) \\ "parameters"))
+      parameters should matchJson(expectedParamJson)
+    }
+  }
+
   test("create skipping index with streaming job options") {
     withTempDir { checkpointDir =>
       sql(s"""