feat(query): inverted index support json type (#15267)

Author: b41sh

Cargo.lock

@@ -108,7 +108,7 @@ members = [
 
 [workspace.dependencies]
 # databend maintains
-jsonb = { git = "https://github.com/datafuselabs/jsonb", rev = "a7325f4" }
+jsonb = { git = "https://github.com/datafuselabs/jsonb", rev = "3fe3acd" }
 
 opendal = { version = "0.45.1", features = [
     "layers-minitrace",

Cargo.toml

@@ -60,6 +60,7 @@ log = { workspace = true }
 tempfile = "3.4.0"
 
 [dev-dependencies]
+jsonb = { workspace = true }
 
 [build-dependencies]
 databend-common-building = { path = "../../common/building" }

src/query/ee/Cargo.toml

@@ -496,7 +496,6 @@ pub fn register(registry: &mut FunctionRegistry) {
                         builder.len(),
                         ErrorCode::GeometryError(e.to_string()).to_string(),
                     );
-                    return;
                 }
             };
             builder.commit_row();
@@ -531,7 +530,6 @@ pub fn register(registry: &mut FunctionRegistry) {
                             builder.len(),
                             ErrorCode::GeometryError(e.to_string()).to_string(),
                         );
-                        return;
                     }
                 };
                 builder.commit_row();

src/query/ee/tests/it/inverted_index/pruning.rs

@@ -384,9 +384,11 @@ impl Binder {
         for column in columns {
             match table_schema.field_with_name(&column.name) {
                 Ok(field) => {
-                    if field.data_type.remove_nullable() != TableDataType::String {
+                    if field.data_type.remove_nullable() != TableDataType::String
+                        && field.data_type.remove_nullable() != TableDataType::Variant
+                    {
                         return Err(ErrorCode::UnsupportedIndex(format!(
-                            "Inverted index currently only support String type, but the type of column {} is {}",
+                            "Inverted index currently only support String and variant type, but the type of column {} is {}",
                             column, field.data_type
                         )));
                     }

src/query/functions/src/scalars/geometry.rs

@@ -2234,14 +2234,20 @@ impl<'a> TypeChecker<'a> {
                 continue;
             }
             let field_names: Vec<&str> = field_str.split(':').collect();
+            // if the field is JSON type, must specify the key path in the object
+            // for example:
+            // the field `info` has the value: `{"tags":{"id":10,"env":"prod","name":"test"}}`
+            // a query can be written like this `info.tags.env:prod`
+            let field_name = field_names[0].trim();
+            let sub_field_names: Vec<&str> = field_name.split('.').collect();
             let column_expr = Expr::ColumnRef {
                 span: query_scalar.span(),
                 column: ColumnRef {
                     database: None,
                     table: None,
                     column: ColumnID::Name(Identifier::from_name(
                         query_scalar.span(),
-                        field_names[0].trim(),
+                        sub_field_names[0].trim(),
                     )),
                 },
             };

src/query/sql/src/planner/binder/ddl/index.rs

@@ -49,9 +49,11 @@ use databend_common_io::constants::DEFAULT_BLOCK_BUFFER_SIZE;
 use opendal::Operator;
 use serde::Deserialize;
 use serde::Serialize;
+use serde_json::Map;
 use tantivy::schema::Document;
 use tantivy::schema::Field;
 use tantivy::schema::IndexRecordOption;
+use tantivy::schema::JsonObjectOptions;
 use tantivy::schema::Schema;
 use tantivy::schema::TextFieldIndexing;
 use tantivy::schema::TextOptions;
@@ -63,6 +65,7 @@ use tantivy::tokenizer::StopWordFilter;
 use tantivy::tokenizer::TextAnalyzer;
 use tantivy::tokenizer::TokenizerManager;
 use tantivy::Directory;
+// use tantivy::schema::document::OwnedValue;
 use tantivy::Index;
 use tantivy::IndexBuilder;
 use tantivy::IndexSettings;
@@ -163,18 +166,26 @@ impl InvertedIndexWriter {
         let text_field_indexing = TextFieldIndexing::default()
             .set_tokenizer(&tokenizer_name)
             .set_index_option(index_record);
-        let text_options = TextOptions::default().set_indexing_options(text_field_indexing);
+        let text_options = TextOptions::default().set_indexing_options(text_field_indexing.clone());
+        let json_options = JsonObjectOptions::default().set_indexing_options(text_field_indexing);
 
         let mut schema_builder = Schema::builder();
         let mut index_fields = Vec::with_capacity(schema.fields.len());
         for field in &schema.fields {
-            if field.data_type().remove_nullable() != DataType::String {
-                return Err(ErrorCode::IllegalDataType(format!(
-                    "inverted index only support String type, but got {}",
-                    field.data_type()
-                )));
-            }
-            let index_field = schema_builder.add_text_field(field.name(), text_options.clone());
+            let index_field = match field.data_type().remove_nullable() {
+                DataType::String => {
+                    schema_builder.add_text_field(field.name(), text_options.clone())
+                }
+                DataType::Variant => {
+                    schema_builder.add_json_field(field.name(), json_options.clone())
+                }
+                _ => {
+                    return Err(ErrorCode::IllegalDataType(format!(
+                        "inverted index only support String and Variant type, but got {}",
+                        field.data_type()
+                    )));
+                }
+            };
             index_fields.push(index_field);
         }
         let index_schema = schema_builder.build();
@@ -221,15 +232,33 @@ impl InvertedIndexWriter {
             }
         }
 
+        let mut types = Vec::with_capacity(self.schema.num_fields());
+        for field in self.schema.fields() {
+            let ty = field.data_type().remove_nullable();
+            types.push(ty);
+        }
         for i in 0..block.num_rows() {
             let mut doc = Document::new();
-            for j in 0..block.num_columns() {
+            for (j, typ) in types.iter().enumerate() {
                 let field = Field::from_field_id(j as u32);
                 let column = block.get_by_offset(j);
-                if let ScalarRef::String(text) = unsafe { column.value.index_unchecked(i) } {
-                    doc.add_text(field, text);
-                } else {
-                    doc.add_text(field, "");
+                match unsafe { column.value.index_unchecked(i) } {
+                    ScalarRef::String(text) => doc.add_text(field, text),
+                    ScalarRef::Variant(jsonb_val) => {
+                        // only support object JSON, other JSON type will not add index.
+                        if let Ok(Some(obj_val)) = jsonb::to_serde_json_object(jsonb_val) {
+                            doc.add_json_object(field, obj_val);
+                        } else {
+                            doc.add_json_object(field, Map::new());
+                        }
+                    }
+                    _ => {
+                        if typ == &DataType::Variant {
+                            doc.add_json_object(field, Map::new());
+                        } else {
+                            doc.add_text(field, "");
+                        }
+                    }
                 }
             }
             self.operations.push(UserOperation::Add(doc));

src/query/sql/src/planner/semantic/type_check.rs

@@ -427,6 +427,41 @@ onlyif mysql
 statement error 1105
 SELECT id, score(), title FROM books WHERE query('title:(设计 实现)^5 description:(学习 +神经网络)^1.1') ORDER BY score() DESC
 
+statement ok
+CREATE TABLE t1 (id int, body json)
+
+statement ok
+CREATE INVERTED INDEX IF NOT EXISTS idx ON t1(body) tokenizer = 'chinese'
+
+statement ok
+INSERT INTO t1 VALUES
+(1, '{"title":"The Psychology of Persuasion","metadata":{"author":"Oliver","publishedDate":"2021-06-15","tags":["psychology","persuasion","behavior"]}}'),
+(2, '{"title":"Sustainable Energy Solutions","metadata":{"author":"Pamela","publishedDate":"2023-12-01","tags":["sustainable energy","solutions","environment"]}}'),
+(3, '{"title":"The Future of Autonomous Vehicles","metadata":{"author":"Quincy","publishedDate":"2022-05-05","tags":["autonomous vehicles","future","technology"]}}'),
+(4, '{"title":"The Role of AI in Customer Service","metadata":{"author":"Rachel","publishedDate":"2021-09-20","tags":["AI","customer service","automation"]}}'),
+(5, '{"title":"Internet of Things Applications","metadata":{"author":"Samuel","publishedDate":"2023-12-15","tags":["IoT","applications","technology"]}}'),
+(6, '{"title":"人工智能与机器学习","metadata":{"author":"张三","publishedDate":"2023-10-23","tags":["人工智能","机器学习","技术"]}}'),
+(7, '{"title":"区块链在金融行业的应用","metadata":{"author":"李四","publishedDate":"2023-09-18","tags":["区块链","金融行业","金融科技"]}}'),
+(8, '{"title":"物联网与智能家居","metadata":{"author":"王五","publishedDate":"2023-08-15","tags":["物联网","智能家居","生活"]}}'),
+(9, '{"title":"量子计算的未来","metadata":{"author":"赵六","publishedDate":"2023-07-20","tags":["量子计算","未来科技","物理学"]}}'),
+(10, '{"title":"网络安全与隐私保护","metadata":{"author":"刘七","publishedDate":"2023-06-25","tags":["网络安全","隐私保护","信息技术"]}}')
+
+query IFT
+SELECT id, score(), body FROM t1 WHERE query('body.title:energy')
+----
+2 3.2352333 {"metadata":{"author":"Pamela","publishedDate":"2023-12-01","tags":["sustainable energy","solutions","environment"]},"title":"Sustainable Energy Solutions"}
+
+query IFT
+SELECT id, score(), body FROM t1 WHERE query('body.metadata.tags:technology')
+----
+3 2.4057739 {"metadata":{"author":"Quincy","publishedDate":"2022-05-05","tags":["autonomous vehicles","future","technology"]},"title":"The Future of Autonomous Vehicles"}
+5 2.4057739 {"metadata":{"author":"Samuel","publishedDate":"2023-12-15","tags":["IoT","applications","technology"]},"title":"Internet of Things Applications"}
+
+query IFT
+SELECT id, score(), body FROM t1 WHERE query('body.metadata.tags:技术')
+----
+6 2.4057739 {"metadata":{"author":"张三","publishedDate":"2023-10-23","tags":["人工智能","机器学习","技术"]},"title":"人工智能与机器学习"}
+10 2.4057739 {"metadata":{"author":"刘七","publishedDate":"2023-06-25","tags":["网络安全","隐私保护","信息技术"]},"title":"网络安全与隐私保护"}
 
 statement ok
 use default

src/query/storages/fuse/src/io/write/inverted_index_writer.rs

@@ -8,7 +8,7 @@
 8	18446744073709551615
 9	12.34
 10	-56.78
-11	140000
+11	140000.0
 12	0.0099
 13	"abcd"
 14	"test"
@@ -28,7 +28,7 @@
 28	18446744073709551615
 29	12.34
 30	-56.78
-31	140000
+31	140000.0
 32	0.0099
 33	"abcd"
 34	"test"