Allow un-quoted skip index parameters

rschu1ze · rschu1ze · commit fb76cb90b1ba · 2024-08-12T15:32:25.000Z
Previously, only this syntax to create a skip index worked:

   INDEX index_name column_name TYPE vector_similarity('hnsw', 'L2Distance')

Now, this syntax will work as well:

  INDEX index_name column_name TYPE vector_similarity(hnsw, L2Distance)
diff --git a/docs/en/engines/table-engines/mergetree-family/annindexes.md b/docs/en/engines/table-engines/mergetree-family/annindexes.md
@@ -43,7 +43,7 @@ CREATE TABLE table
 (
   id Int64,
   vectors Array(Float32),
-  INDEX index_name vec TYPE vector_similarity(method, distance_function[, quantization, connectivity, expansion_add, expansion_search]) [GRANULARITY N]
+  INDEX index_name vectors TYPE vector_similarity(method, distance_function[, quantization, connectivity, expansion_add, expansion_search]) [GRANULARITY N]
 )
 ENGINE = MergeTree
 ORDER BY id;
@@ -59,6 +59,19 @@ Parameters:
 - `ef_construction`: (optional, default: 128)
 - `ef_search`: (optional, default: 64)
 
+Example:
+
+```sql
+CREATE TABLE table
+(
+  id Int64,
+  vectors Array(Float32),
+  INDEX idx vectors TYPE vector_similarity('hnsw', 'L2Distance') -- Alternative syntax: TYPE vector_similarity(hnsw, L2Distance)
+)
+ENGINE = MergeTree
+ORDER BY id;
+```
+
 Vector similarity indexes are based on the [USearch library](https://github.com/unum-cloud/usearch), which implements the [HNSW
 algorithm](https://arxiv.org/abs/1603.09320), i.e., a hierarchical graph where each point represents a vector and the edges represent
 similarity. Such hierarchical structures can be very efficient on large collections. They may often fetch 0.05% or less data from the
diff --git a/src/Storages/IndicesDescription.cpp b/src/Storages/IndicesDescription.cpp
@@ -3,6 +3,7 @@
 #include <Storages/IndicesDescription.h>
 
 #include <Parsers/ASTFunction.h>
+#include <Parsers/ASTIdentifier.h>
 #include <Parsers/ASTIndexDeclaration.h>
 #include <Parsers/ASTLiteral.h>
 #include <Parsers/ParserCreateQuery.h>
@@ -130,10 +131,15 @@ IndexDescription IndexDescription::getIndexFromAST(const ASTPtr & definition_ast
     {
         for (size_t i = 0; i < index_type->arguments->children.size(); ++i)
         {
-            const auto * argument = index_type->arguments->children[i]->as<ASTLiteral>();
-            if (!argument)
+            const auto & child = index_type->arguments->children[i];
+            if (const auto * ast_literal = child->as<ASTLiteral>(); ast_literal != nullptr)
+                /// E.g. INDEX index_name column_name TYPE vector_similarity('hnsw', 'f32')
+                result.arguments.emplace_back(ast_literal->value);
+            else if (const auto * ast_identifier = child->as<ASTIdentifier>(); ast_identifier != nullptr)
+                /// E.g. INDEX index_name column_name TYPE vector_similarity(hnsw, f32)
+                result.arguments.emplace_back(ast_identifier->name());
+            else
                 throw Exception(ErrorCodes::INCORRECT_QUERY, "Only literals can be skip index arguments");
-            result.arguments.emplace_back(argument->value);
         }
     }
 
diff --git a/tests/queries/0_stateless/02354_vector_search_unquoted_index_parameters.reference b/tests/queries/0_stateless/02354_vector_search_unquoted_index_parameters.reference
diff --git a/tests/queries/0_stateless/02354_vector_search_unquoted_index_parameters.sql b/tests/queries/0_stateless/02354_vector_search_unquoted_index_parameters.sql
@@ -0,0 +1,23 @@
+-- Tags: no-fasttest, no-ordinary-database
+
+SET allow_experimental_vector_similarity_index = 1;
+
+-- Tests that quoted and unquoted parameters can be passed to vector search indexes.
+
+DROP TABLE IF EXISTS tab1;
+DROP TABLE IF EXISTS tab2;
+
+CREATE TABLE tab1 (id Int32, vec Array(Float32), PRIMARY KEY id, INDEX vec_idx(vec) TYPE vector_similarity('hnsw', 'L2Distance'));
+CREATE TABLE tab2 (id Int32, vec Array(Float32), PRIMARY KEY id, INDEX vec_idx(vec) TYPE vector_similarity(hnsw, L2Distance));
+
+DROP TABLE tab1;
+DROP TABLE tab2;
+
+CREATE TABLE tab1 (id Int32, vec Array(Float32), PRIMARY KEY id);
+CREATE TABLE tab2 (id Int32, vec Array(Float32), PRIMARY KEY id);
+
+ALTER TABLE tab1 ADD INDEX idx1(vec) TYPE vector_similarity('hnsw', 'L2Distance');
+ALTER TABLE tab2 ADD INDEX idx2(vec) TYPE vector_similarity(hnsw, L2Distance);
+
+DROP TABLE tab1;
+DROP TABLE tab2;

Original file line number	Diff line number	Diff line change
`@@ -3,6 +3,7 @@`
`3`	`3`	`#include <Storages/IndicesDescription.h>`
`4`	`4`
`5`	`5`	`#include <Parsers/ASTFunction.h>`
	`6`	`+#include <Parsers/ASTIdentifier.h>`
`6`	`7`	`#include <Parsers/ASTIndexDeclaration.h>`
`7`	`8`	`#include <Parsers/ASTLiteral.h>`
`8`	`9`	`#include <Parsers/ParserCreateQuery.h>`
`@@ -130,10 +131,15 @@ IndexDescription IndexDescription::getIndexFromAST(const ASTPtr & definition_ast`
`130`	`131`	`{`
`131`	`132`	`for (size_t i = 0; i < index_type->arguments->children.size(); ++i)`
`132`	`133`	`{`
`133`		`- const auto * argument = index_type->arguments->children[i]->as<ASTLiteral>();`
`134`		`- if (!argument)`
	`134`	`+ const auto & child = index_type->arguments->children[i];`
	`135`	`+ if (const auto * ast_literal = child->as<ASTLiteral>(); ast_literal != nullptr)`
	`136`	`+ /// E.g. INDEX index_name column_name TYPE vector_similarity('hnsw', 'f32')`
	`137`	`+ result.arguments.emplace_back(ast_literal->value);`
	`138`	`+ else if (const auto * ast_identifier = child->as<ASTIdentifier>(); ast_identifier != nullptr)`
	`139`	`+ /// E.g. INDEX index_name column_name TYPE vector_similarity(hnsw, f32)`
	`140`	`+ result.arguments.emplace_back(ast_identifier->name());`
	`141`	`+ else`
`135`	`142`	`throw Exception(ErrorCodes::INCORRECT_QUERY, "Only literals can be skip index arguments");`
`136`		`- result.arguments.emplace_back(argument->value);`
`137`	`143`	`}`
`138`	`144`	`}`
`139`	`145`