Files
milvus/internal/core/src/query/Plan.cpp
T
Spade AandGitHub f6f716bcfd feat: impl StructArray -- support embedding searches embeddings in embedding list with element level filter expression (#45830)
issue: https://github.com/milvus-io/milvus/issues/42148

For a vector field inside a STRUCT, since a STRUCT can only appear as
the element type of an ARRAY field, the vector field in STRUCT is
effectively an array of vectors, i.e. an embedding list.
Milvus already supports searching embedding lists with metrics whose
names start with the prefix MAX_SIM_.

This PR allows Milvus to search embeddings inside an embedding list
using the same metrics as normal embedding fields. Each embedding in the
list is treated as an independent vector and participates in ANN search.

Further, since STRUCT may contain scalar fields that are highly related
to the embedding field, this PR introduces an element-level filter
expression to refine search results.
The grammar of the element-level filter is:

element_filter(structFieldName, $[subFieldName] == 3)

where $[subFieldName] refers to the value of subFieldName in each
element of the STRUCT array structFieldName.

It can be combined with existing filter expressions, for example:

"varcharField == 'aaa' && element_filter(struct_field, $[struct_int] ==
3)"

A full example:
```
struct_schema = milvus_client.create_struct_field_schema()
struct_schema.add_field("struct_str", DataType.VARCHAR, max_length=65535)
struct_schema.add_field("struct_int", DataType.INT32)
struct_schema.add_field("struct_float_vec", DataType.FLOAT_VECTOR, dim=EMBEDDING_DIM)

schema.add_field(
    "struct_field",
    datatype=DataType.ARRAY,
    element_type=DataType.STRUCT,
    struct_schema=struct_schema,
    max_capacity=1000,
)
...

filter = "varcharField == 'aaa' && element_filter(struct_field, $[struct_int] == 3 && $[struct_str] == 'abc')"
res = milvus_client.search(
    COLLECTION_NAME,
    data=query_embeddings,
    limit=10,
    anns_field="struct_field[struct_float_vec]",
    filter=filter,
    output_fields=["struct_field[struct_int]", "varcharField"],
)

```
TODO:
1. When an `element_filter` expression is used, a regular filter
expression must also be present. Remove this restriction.
2. Implement `element_filter` expressions in the `query`.

---------

Signed-off-by: SpadeA <tangchenjie1210@gmail.com>
2025-12-15 12:01:15 +08:00

218 lines
8.8 KiB
C++

// Licensed to the LF AI & Data foundation under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "Plan.h"
#include "common/Utils.h"
#include "PlanProto.h"
namespace milvus::query {
// deprecated
std::unique_ptr<PlaceholderGroup>
ParsePlaceholderGroup(const Plan* plan,
const std::string_view placeholder_group_blob) {
return ParsePlaceholderGroup(
plan,
reinterpret_cast<const uint8_t*>(placeholder_group_blob.data()),
placeholder_group_blob.size());
}
bool
check_data_type(
const FieldMeta& field_meta,
const milvus::proto::common::PlaceholderValue& placeholder_value) {
if (field_meta.get_data_type() == DataType::VECTOR_ARRAY) {
if (field_meta.get_element_type() == DataType::VECTOR_FLOAT) {
if (placeholder_value.element_level()) {
return placeholder_value.type() ==
milvus::proto::common::PlaceholderType::FloatVector;
} else {
return placeholder_value.type() ==
milvus::proto::common::PlaceholderType::
EmbListFloatVector;
}
} else if (field_meta.get_element_type() == DataType::VECTOR_FLOAT16) {
return placeholder_value.type() ==
milvus::proto::common::PlaceholderType::EmbListFloat16Vector;
} else if (field_meta.get_element_type() == DataType::VECTOR_BFLOAT16) {
return placeholder_value.type() ==
milvus::proto::common::PlaceholderType::
EmbListBFloat16Vector;
} else if (field_meta.get_element_type() == DataType::VECTOR_BINARY) {
return placeholder_value.type() ==
milvus::proto::common::PlaceholderType::EmbListBinaryVector;
} else if (field_meta.get_element_type() == DataType::VECTOR_INT8) {
return placeholder_value.type() ==
milvus::proto::common::PlaceholderType::EmbListInt8Vector;
}
return false;
}
return static_cast<int>(field_meta.get_data_type()) ==
static_cast<int>(placeholder_value.type());
}
std::unique_ptr<PlaceholderGroup>
ParsePlaceholderGroup(const Plan* plan,
const uint8_t* blob,
const int64_t blob_len) {
auto result = std::make_unique<PlaceholderGroup>();
milvus::proto::common::PlaceholderGroup ph_group;
auto ok = ph_group.ParseFromArray(blob, blob_len);
Assert(ok);
for (auto& ph : ph_group.placeholders()) {
Placeholder element;
element.tag_ = ph.tag();
element.element_level_ = ph.element_level();
Assert(plan->tag2field_.count(element.tag_));
auto field_id = plan->tag2field_.at(element.tag_);
auto& field_meta = plan->schema_->operator[](field_id);
AssertInfo(check_data_type(field_meta, ph),
"vector type must be the same, field {} - type {}, search "
"ph type {}",
field_meta.get_name().get(),
field_meta.get_data_type(),
static_cast<DataType>(ph.type()));
element.num_of_queries_ = ph.values_size();
AssertInfo(element.num_of_queries_ > 0, "must have queries");
if (ph.type() ==
milvus::proto::common::PlaceholderType::SparseFloatVector) {
element.sparse_matrix_ =
SparseBytesToRows(ph.values(), /*validate=*/true);
} else {
auto line_size = ph.values().Get(0).size();
auto& target = element.blob_;
if (field_meta.get_data_type() != DataType::VECTOR_ARRAY ||
ph.element_level()) {
if (field_meta.get_sizeof() != line_size &&
!ph.element_level()) {
ThrowInfo(DimNotMatch,
fmt::format(
"vector dimension mismatch, expected vector "
"size(byte) {}, actual {}.",
field_meta.get_sizeof(),
line_size));
}
target.reserve(line_size * element.num_of_queries_);
for (auto& line : ph.values()) {
AssertInfo(line_size == line.size(),
"vector dimension mismatch, expected vector "
"size(byte) {}, actual {}.",
line_size,
line.size());
target.insert(target.end(), line.begin(), line.end());
}
} else {
target.reserve(line_size * element.num_of_queries_);
auto dim = field_meta.get_dim();
// If the vector is embedding list, line contains multiple vectors.
// And we should record the offsets so that we can identify each
// embedding list in a flattened vectors.
auto& offsets = element.offsets_;
offsets.reserve(element.num_of_queries_ + 1);
size_t offset = 0;
offsets.push_back(offset);
auto bytes_per_vec = milvus::vector_bytes_per_element(
field_meta.get_element_type(), dim);
for (auto& line : ph.values()) {
target.insert(target.end(), line.begin(), line.end());
AssertInfo(
line.size() % bytes_per_vec == 0,
"line.size() % bytes_per_vec == 0 assert failed, "
"line.size() = {}, dim = {}, bytes_per_vec = {}",
line.size(),
dim,
bytes_per_vec);
offset += line.size() / bytes_per_vec;
offsets.push_back(offset);
}
}
}
result->emplace_back(std::move(element));
}
return result;
}
void
ParsePlanNodeProto(proto::plan::PlanNode& plan_node,
const void* serialized_expr_plan,
int64_t size) {
google::protobuf::io::ArrayInputStream array_stream(serialized_expr_plan,
size);
google::protobuf::io::CodedInputStream input_stream(&array_stream);
input_stream.SetRecursionLimit(std::numeric_limits<int32_t>::max());
auto res = plan_node.ParsePartialFromCodedStream(&input_stream);
if (!res) {
ThrowInfo(UnexpectedError, "parse plan node proto failed");
}
}
std::unique_ptr<Plan>
CreateSearchPlanByExpr(SchemaPtr schema,
const void* serialized_expr_plan,
const int64_t size) {
// Note: serialized_expr_plan is of binary format
proto::plan::PlanNode plan_node;
ParsePlanNodeProto(plan_node, serialized_expr_plan, size);
return ProtoParser(std::move(schema)).CreatePlan(plan_node);
}
std::unique_ptr<Plan>
CreateSearchPlanFromPlanNode(SchemaPtr schema,
const proto::plan::PlanNode& plan_node) {
return ProtoParser(std::move(schema)).CreatePlan(plan_node);
}
std::unique_ptr<RetrievePlan>
CreateRetrievePlanByExpr(SchemaPtr schema,
const void* serialized_expr_plan,
const int64_t size) {
proto::plan::PlanNode plan_node;
ParsePlanNodeProto(plan_node, serialized_expr_plan, size);
return ProtoParser(std::move(schema)).CreateRetrievePlan(plan_node);
}
int64_t
GetTopK(const Plan* plan) {
return plan->plan_node_->search_info_.topk_;
}
int64_t
GetFieldID(const Plan* plan) {
return plan->plan_node_->search_info_.field_id_.get();
}
int64_t
GetNumOfQueries(const PlaceholderGroup* group) {
return group->at(0).num_of_queries_;
}
// std::unique_ptr<RetrievePlan>
// CreateRetrievePlan(const Schema& schema, proto::segcore::RetrieveRequest&& request) {
// auto plan = std::make_unique<RetrievePlan>();
// plan->seg_offsets_ = std::unique_ptr<proto::schema::IDs>(request.release_ids());
// for (auto& field_id : request.output_fields_id()) {
// plan->field_ids_.push_back(schema.get_offset(FieldId(field_id)));
// }
// return plan;
//}
} // namespace milvus::query