Regex match

CMakeLists.txt

@@ -86,6 +86,7 @@ include_directories(third_party/spdlog)
 include_directories(third_party/nlohmann_json)
 include_directories(third_party/utf8proc/include)
 include_directories(third_party/pybind11/include)
+include_directories(third_party/re2/include)
 
 add_subdirectory(third_party)
 add_subdirectory(src)

src/CMakeLists.txt

@@ -11,7 +11,7 @@ add_subdirectory(storage)
 add_subdirectory(transaction)
 
 add_library(kuzu STATIC ${ALL_OBJECT_FILES})
-target_link_libraries(kuzu PUBLIC antlr4_cypher antlr4_runtime utf8proc
+target_link_libraries(kuzu PUBLIC antlr4_cypher antlr4_runtime utf8proc re2
         parquet_lib arrow_lib arrow_deps Threads::Threads)
 target_include_directories(
         kuzu PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>

src/antlr4/Cypher.g4

@@ -390,7 +390,10 @@ kU_ListSliceOperatorExpression
     : SP ? '[' oC_Expression? ':' oC_Expression? ']' ;
 
 oC_StringOperatorExpression
-    :  ( ( SP STARTS SP WITH ) | ( SP ENDS SP WITH ) | ( SP CONTAINS ) ) SP? oC_PropertyOrLabelsExpression ;
+    :  ( oC_RegularExpression | ( SP STARTS SP WITH ) | ( SP ENDS SP WITH ) | ( SP CONTAINS ) ) SP? oC_PropertyOrLabelsExpression ;
+
+oC_RegularExpression
+    :  SP? '=~' ;
 
 STARTS : ( 'S' | 's' ) ( 'T' | 't' ) ( 'A' | 'a' ) ( 'R' | 'r' ) ( 'T' | 't' ) ( 'S' | 's' ) ;
 

src/function/built_in_vector_operations.cpp

@@ -255,6 +255,7 @@ void BuiltInVectorOperations::registerStringOperations() {
     vectorOperations.insert({CONCAT_FUNC_NAME, ConcatVectorOperation::getDefinitions()});
     vectorOperations.insert({CONTAINS_FUNC_NAME, ContainsVectorOperation::getDefinitions()});
     vectorOperations.insert({ENDS_WITH_FUNC_NAME, EndsWithVectorOperation::getDefinitions()});
+    vectorOperations.insert({RE_MATCH_FUNC_NAME, REMatchVectorOperation::getDefinitions()});
     vectorOperations.insert({LCASE_FUNC_NAME, LowerVectorOperation::getDefinitions()});
     vectorOperations.insert({LEFT_FUNC_NAME, LeftVectorOperation::getDefinitions()});
     vectorOperations.insert({LENGTH_FUNC_NAME, LengthVectorOperation::getDefinitions()});

src/function/vector_string_operations.cpp

@@ -7,6 +7,7 @@
 #include "function/string/operations/left_operation.h"
 #include "function/string/operations/length_operation.h"
 #include "function/string/operations/lpad_operation.h"
+#include "function/string/operations/reg_expr_operation.h"
 #include "function/string/operations/repeat_operation.h"
 #include "function/string/operations/right_operation.h"
 #include "function/string/operations/rpad_operation.h"
@@ -54,6 +55,16 @@ vector<unique_ptr<VectorOperationDefinition>> EndsWithVectorOperation::getDefini
     return definitions;
 }
 
+vector<unique_ptr<VectorOperationDefinition>> REMatchVectorOperation::getDefinitions() {
+    vector<unique_ptr<VectorOperationDefinition>> definitions;
+    definitions.emplace_back(make_unique<VectorOperationDefinition>(RE_MATCH_FUNC_NAME,
+        vector<DataTypeID>{STRING, STRING}, BOOL,
+        BinaryExecFunction<ku_string_t, ku_string_t, uint8_t, operation::REMatch>,
+        BinarySelectFunction<ku_string_t, ku_string_t, operation::REMatch>,
+        false /* isVarLength */));
+    return definitions;
+}
+
 vector<unique_ptr<VectorOperationDefinition>> LeftVectorOperation::getDefinitions() {
     vector<unique_ptr<VectorOperationDefinition>> definitions;
     definitions.emplace_back(

src/include/common/expression_type.h

@@ -114,6 +114,7 @@ const string ARRAY_EXTRACT_FUNC_NAME = "ARRAY_EXTRACT";
 const string CONCAT_FUNC_NAME = "CONCAT";
 const string CONTAINS_FUNC_NAME = "CONTAINS";
 const string ENDS_WITH_FUNC_NAME = "ENDS_WITH";
+const string RE_MATCH_FUNC_NAME = "RE_MATCH";
 const string LCASE_FUNC_NAME = "LCASE";
 const string LEFT_FUNC_NAME = "LEFT";
 const string LENGTH_FUNC_NAME = "LENGTH";

src/include/function/string/operations/reg_expr_operation.h

@@ -0,0 +1,28 @@
+#pragma once
+
+#include <regex>
+
+#include "common/types/ku_string.h"
+#include "function/string/operations/find_operation.h"
+#include "re2.h"
+
+using namespace kuzu::common;
+using namespace kuzu::regex;
+
+namespace kuzu {
+namespace function {
+namespace operation {
+
+struct REMatch {
+    static inline void operation(ku_string_t& left, ku_string_t& right, uint8_t& result) {
+        // Cypher parses escape characters with 2 backslash eg. for expressing '.' requires '\\.'
+        // Since Regular Expression requires only 1 backslash '\.' we need to replace double slash
+        // with single
+        string pattern = std::regex_replace(right.getAsString(), std::regex(R"(\\\\)"), "\\");
+        result = RE2::FullMatch(left.getAsString(), pattern);
+    }
+};
+
+} // namespace operation
+} // namespace function
+} // namespace kuzu

src/include/function/string/vector_string_operations.h

@@ -63,6 +63,10 @@ struct EndsWithVectorOperation : public VectorStringOperations {
     static vector<unique_ptr<VectorOperationDefinition>> getDefinitions();
 };
 
+struct REMatchVectorOperation : public VectorStringOperations {
+    static vector<unique_ptr<VectorOperationDefinition>> getDefinitions();
+};
+
 struct LeftVectorOperation : public VectorStringOperations {
     static vector<unique_ptr<VectorOperationDefinition>> getDefinitions();
 };

src/parser/transformer.cpp

@@ -591,10 +591,13 @@ unique_ptr<ParsedExpression> Transformer::transformStringOperatorExpression(
     } else if (ctx.ENDS()) {
         return make_unique<ParsedFunctionExpression>(
             ENDS_WITH_FUNC_NAME, std::move(propertyExpression), std::move(right), rawExpression);
-    } else {
-        assert(ctx.CONTAINS());
+    } else if (ctx.CONTAINS()) {
         return make_unique<ParsedFunctionExpression>(
             CONTAINS_FUNC_NAME, std::move(propertyExpression), std::move(right), rawExpression);
+    } else {
+        assert(ctx.oC_RegularExpression());
+        return make_unique<ParsedFunctionExpression>(
+            RE_MATCH_FUNC_NAME, std::move(propertyExpression), std::move(right), rawExpression);
     }
 }
 

test/test_files/tinysnb/function/string.test

@@ -144,6 +144,37 @@ False
 ---- 1
 Carol
 
+-NAME RegExprMatchString
+-QUERY MATCH (a:person) WHERE a.fName =~ 'Ali.*' RETURN a.ID
+---- 1
+0
+
+-NAME RegExprAnchorMatchString
+-QUERY MATCH (a:person) UNWIND a.usedNames as x WITH x AS name WHERE name =~ '^A.*$' RETURN name
+---- 2
+Ad
+Aida
+
+-NAME RegExprUTF8MatchString
+-QUERY MATCH (m:movies) WHERE m.name =~ '.*â.*' RETURN m.name
+---- 1
+Sóló cón tu párejâ
+
+-NAME RegExprEscapeSeq1
+-QUERY Return RE_MATCH("peter_n@example.com", ".*\\.com");
+---- 1
+True
+
+-NAME RegExprEscapeSeq2
+-QUERY Return RE_MATCH("Alice.*^Alice", ".*\\.\\*.*");
+---- 1
+True
+
+-NAME RegExprEscapeSeq3
+-QUERY Return RE_MATCH("peter_n@examplecom", ".*\\.com");
+---- 1
+False
+
 -NAME LowerStructuredStr
 -QUERY MATCH (o:organisation) RETURN lower(o.name)
 ---- 3

third_party/CMakeLists.txt

@@ -2,3 +2,4 @@ add_subdirectory(antlr4_runtime)
 add_subdirectory(antlr4_cypher)
 add_subdirectory(utf8proc)
 add_subdirectory(pybind11)
+add_subdirectory(re2)