mirror of
https://github.com/milvus-io/milvus.git
synced 2026-07-21 02:05:41 +00:00
issue: #47420 ## What this PR does Project-wide migration of raw `fmt.Errorf` / `errors.New` in function bodies onto the `merr` framework, plus the Sys-vs-Input error classification and the machinery it drives (retriability, fine-grained metrics, segcore unification), plus the convention docs and a linter that keeps it from regressing. Scope: storage, proxy, coordinators (root/data/query), query node, data node, `pkg/util` & `internal/util`, expression parser, message queue, streaming, and misc packages. Bare raw-error usages went from ~3000 to a ~340 allowlist (package-level sentinels / build-tag / test sites). --- ## How to review this PR It is large but the vast majority is mechanical. Changes fall into three tiers; spend review budget on Part 2 and Part 3. ### Part 1 — Mechanical standardization (low risk, verify by rule) Each converted call follows one of a small fixed set of rules. To review, check that each site obeys the matching rule rather than reading every line: | Pattern | Rule | |---|---| | `fmt.Errorf("...")` originating a new error | → `merr.WrapErrXxxMsg("...")` with a code matching the failure's meaning | | Adding context to an existing typed error | → `merr.Wrap(err, "...")` / `merr.Wrapf(...)` — **preserves** the inner code (never `WrapErr*Err`, which overwrites it) | | Errors inside the streaming subsystem | → `status.New*` factories (StreamingError), **not** merr — this is the component-internal dialect (see `docs/dev/error_handling_guide.md`) | | Low-level / control-flow signal caught by `errors.Is` | → kept as a package-level `errors.New` sentinel (lowercase, same-package) | Conventions are documented in `docs/dev/error_handling_guide.md` (how-to) and `docs/dev/error_sentinel_convention.md` (rules + audit). A `gocritic`/`ruleguard` rule (`rawmerrerror`, in `rules.go`) enforces "no raw `return errors.New/fmt.Errorf`" under `make verifiers`. ### Part 2 — Behavior changes (review these closely) These are the sites where the wire contract or runtime behavior changes, not just the source text. Listed by category; representative locations given, full set in the diff. **A. gRPC wire-code shifts: `UnexpectedError(1)/Code 65535` → typed code.** Where a handler previously returned a raw error (collapsed to `Code=65535` on the wire), it now returns a typed merr, so the client sees a real code. The most common shift is to `IllegalArgument(5)/Code 1100` (ParameterInvalid). Touch points include datanode task handlers (CreateTask/Query/Drop), proxy Upsert, querynode GetMetrics, datacoord CreateIndex, httpserver query-response builder, and typeutil schema validation. One code refinement: an index-param validation moved `1100` → `1101` (ParameterMissing). **Client/SDK assertions and any code that switched on `Code=65535` for these paths must be re-checked** (the go_client e2e assertions were already aligned in this PR). **B. Prometheus `status` label contract change (externally visible).** The proxy metric's coarse `fail` / `rejected` values are split into `fail_input` / `fail_system` and `rejected_user` / `rejected_system` (in `requestutil.ParseMetricLabel`; auth/privilege rejections count as `rejected_user`), so dashboards can attribute a failure to caller vs operator. **Dashboards/alerts querying `status="fail"` must migrate to `status=~"fail_.*"`, and `status="rejected"` to `status=~"rejected_.*"`.** The in-repo Grafana dashboard is already migrated; external dashboards built on the old values silently go empty after upgrade. This is the one change that requires an ops-side migration. **C. Retriability semantics.** - C1: `merr.Status(err)` now forces `Retriable=false` when the error is an `InputError` — a malformed request can never succeed on blind retry, so clients never get the self-contradictory "your input is wrong but you may retry". - C2: `retry.Do` short-circuits an `InputError` (non-retriable) — **but only when the caller did not pass a `RetryErr` predicate**. The check is an `if c.isRetryErr != nil { ... } else if InputError { ... }` *mutually exclusive* branch (`pkg/util/retry/retry.go`): an explicit `RetryErr` takes precedence and bypasses the InputError abort. `retry.Handle` deliberately does **not** apply the InputError abort (its callers signal abort via `shouldRetry=false`). Four flusher startup callsites that must retry through transient "not ready" errors were given explicit `RetryErr` escape hatches. **D. segcore (C++→Go) error classification.** A single shared Go-side table (`pkg/util/merr/segcore.go`) maps each segcore code to a merr sentinel + InputError/signal category, replacing scattered hand-written `if errorCode == ...` switches in the cgo wrappers. **Wire `Code` values change for every segcore pass-through error, not just the remapped ones.** Named sentinels remap (C++ `2003` → merr `2001`, `2033` → `2002`, Folly/Knowhere codes likewise); **all remaining pass-through codes (`2004`–`2043`, previously surfaced to clients as raw C++ enum values) now serialize as `2000`** (`ErrSegcore`), with the original C++ code preserved in the `Reason` text (`segcoreCode=...`); unknown/future codes collapse to `2000` as well (pinned by the `wire_code_projection` test). Transient segcore classes (object storage / file IO / OOM / mmap / FieldNotLoaded — 11 codes) now report `Retriable=true`. **Any client switching on raw segcore codes in the `2004`–`2043` range must be re-checked**; the in-Reason code remains available for diagnostics. Signal codes (PretendFinished / FollyCancel) are recognized centrally. `errors.Is`-based control flow on these (e.g. scheduler skip/retry) is preserved. **E. InputError classification (25 sentinels + dynamic marks).** 25 sentinels in `errors.go` carry `WithErrorType(InputError)` (the Collection / ResourceGroup / Database families, `ErrIndexDuplicate`, `ErrParameterInvalid`, `ErrPrivilegeNotAuthenticated`, `ErrImportFailed`, `ErrQueryPlan`, ...), plus dynamic marks for the 8 segcore input codes (ExprInvalid, DimNotMatch, MetricTypeInvalid, FieldIDInvalid, ...) and `WrapErrAsInputError`. The widest blast radius is `ErrParameterInvalid` (1100): ~2335 `WrapErrParameterInvalid*` callsites now classify as input / non-retriable. Because of C1/C2 this changes retriability for any path that returns these. **The audit to confirm no transient path was mis-marked is the single most important review item** (see Part 3). One reverse correction: storage field-stats parsing moved from `ErrParameterInvalid` (input) to `ErrDataIntegrity` — a corrupted stored stat is data corruption, not user input. ### Part 3 — Known risks & traps (called out proactively) 1. **`merr.Wrap` vs `WrapErr*Err` (code-masking).** `WrapErr*Err` builds a `wrappedMilvusError{sentinel: ErrServiceInternal}` whose `code()` returns the *outer* sentinel — it overwrites the inner typed code and hides the `errors.Is` chain. This is intentional (use it to *deliberately* downgrade), but it was a recurring conversion defect; the rule "add context with `merr.Wrap`, downgrade with `WrapErr*Err`" is enforced by convention and reviewed across the diff. 2. **InputError × `retry.Do` blast radius.** Marking a sentinel `InputError` makes any `retry.Do(...)` without a `RetryErr` predicate stop retrying it. Reviewers should sanity-check that no transient use of the 19 newly-marked sentinels (especially `ErrParameterInvalid`) sits inside a retry loop that needed to keep spinning. The known flusher cases were handled (see C2). 3. **The ~340 raw-error allowlist.** What remains as bare `errors.New` is, by design: package-level sentinels (caught by `errors.Is`), `//go:build test` sites, and out-of-band trees (`cmd/`, `tests/`, codegen, walimpls). The linter only bans the *direct-return* form; assignment-then-return escapes and the full no-exceptions ban are deferred to an AST-based linter (Tier 2, documented). 4. **segcore C++ second step deferred.** This PR unifies classification on the Go side; splitting the dual-semantic C++ codes at the source is a follow-up. --- ## Validation - `make verifiers`: Go side clean (gofmt + static-check across modules, including the new `rawmerrerror` rule with a 0-hit baseline repo-wide). - `make test-go`: passing; the one real regression introduced (a datanode `invalid_task_type` assertion shifting `1` → `5` from a ParameterInvalid conversion) was fixed in-tree. - go_client e2e CreateIndex assertions aligned to the new merr messages. --------- Signed-off-by: zhenshan.cao <zhenshan.cao@zilliz.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
650 lines
20 KiB
Go
650 lines
20 KiB
Go
// 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.
|
|
|
|
package compactor
|
|
|
|
import (
|
|
"github.com/apache/arrow/go/v17/arrow"
|
|
"github.com/apache/arrow/go/v17/arrow/array"
|
|
"github.com/apache/arrow/go/v17/arrow/memory"
|
|
|
|
"github.com/milvus-io/milvus-proto/go-api/v3/schemapb"
|
|
"github.com/milvus-io/milvus/internal/storage"
|
|
"github.com/milvus-io/milvus/internal/util/function"
|
|
"github.com/milvus-io/milvus/pkg/v3/common"
|
|
"github.com/milvus-io/milvus/pkg/v3/util/merr"
|
|
"github.com/milvus-io/milvus/pkg/v3/util/typeutil"
|
|
)
|
|
|
|
type FunctionMaterializer interface {
|
|
Materialize(rec storage.Record) (map[int64]arrow.Array, error)
|
|
Close()
|
|
}
|
|
|
|
type rowRange struct {
|
|
start int
|
|
end int
|
|
}
|
|
|
|
type recordSelection struct {
|
|
ranges []rowRange
|
|
length int
|
|
}
|
|
|
|
func (s *recordSelection) Len() int {
|
|
if s == nil {
|
|
return 0
|
|
}
|
|
return s.length
|
|
}
|
|
|
|
type RecordMaterializer struct {
|
|
materializers []FunctionMaterializer
|
|
missingFields []*schemapb.FieldSchema
|
|
schema *schemapb.CollectionSchema
|
|
}
|
|
|
|
func NewRecordMaterializer(schema *schemapb.CollectionSchema, functions []*schemapb.FunctionSchema, existingFields map[int64]struct{}) (*RecordMaterializer, error) {
|
|
materializer := &RecordMaterializer{schema: schema}
|
|
materializedFields := make(map[int64]struct{})
|
|
for _, functionSchema := range functions {
|
|
outputIndexes := functionOutputIndexesToMaterialize(functionSchema, existingFields)
|
|
if len(outputIndexes) == 0 {
|
|
continue
|
|
}
|
|
for _, outputIndex := range outputIndexes {
|
|
materializedFields[functionSchema.GetOutputFieldIds()[outputIndex]] = struct{}{}
|
|
}
|
|
|
|
runner, err := function.NewFunctionRunner(schema, functionSchema)
|
|
if err != nil {
|
|
materializer.Close()
|
|
return nil, err
|
|
}
|
|
if runner == nil {
|
|
materializer.Close()
|
|
return nil, merr.WrapErrFunctionFailedMsg("failed to set up function runner for %s", functionSchema.GetName())
|
|
}
|
|
functionMaterializer, err := newFunctionMaterializer(schema, runner, outputIndexes, true)
|
|
if err != nil {
|
|
runner.Close()
|
|
materializer.Close()
|
|
return nil, err
|
|
}
|
|
materializer.materializers = append(materializer.materializers, functionMaterializer)
|
|
}
|
|
materializer.missingFields = missingNonMaterializedSchemaFields(schema, existingFields, materializedFields)
|
|
return materializer, nil
|
|
}
|
|
|
|
func (m *RecordMaterializer) Wrap(rec storage.Record) (storage.Record, error) {
|
|
return m.WrapWithSelection(rec, nil)
|
|
}
|
|
|
|
func (m *RecordMaterializer) WrapWithSelection(rec storage.Record, selection *recordSelection) (storage.Record, error) {
|
|
base := rec
|
|
var selected *selectedRecord
|
|
if selection != nil {
|
|
selected = newSelectedRecord(rec, m.schema, selection)
|
|
base = selected
|
|
}
|
|
if !m.hasMaterialization() {
|
|
return base, nil
|
|
}
|
|
|
|
computed := make(map[int64]arrow.Array)
|
|
for _, materializer := range m.materializers {
|
|
arrays, err := materializer.Materialize(base)
|
|
if err != nil {
|
|
releaseArrowArrays(computed)
|
|
if base != rec {
|
|
base.Release()
|
|
}
|
|
if selected != nil && selected.err != nil {
|
|
return nil, selected.err
|
|
}
|
|
return nil, err
|
|
}
|
|
if selected != nil && selected.err != nil {
|
|
releaseArrowArrays(computed)
|
|
base.Release()
|
|
return nil, selected.err
|
|
}
|
|
for fieldID, arr := range arrays {
|
|
computed[fieldID] = arr
|
|
}
|
|
}
|
|
for _, field := range m.missingFields {
|
|
fieldID := field.GetFieldID()
|
|
if _, ok := computed[fieldID]; ok {
|
|
continue
|
|
}
|
|
arr, err := storage.GenerateEmptyArrayFromSchema(field, base.Len())
|
|
if err != nil {
|
|
releaseArrowArrays(computed)
|
|
if base != rec {
|
|
base.Release()
|
|
}
|
|
return nil, err
|
|
}
|
|
computed[fieldID] = arr
|
|
}
|
|
if len(computed) == 0 {
|
|
return base, nil
|
|
}
|
|
return &materializedRecord{base: base, computed: computed}, nil
|
|
}
|
|
|
|
func (m *RecordMaterializer) Close() {
|
|
if m == nil {
|
|
return
|
|
}
|
|
for _, materializer := range m.materializers {
|
|
materializer.Close()
|
|
}
|
|
}
|
|
|
|
func (m *RecordMaterializer) hasMaterialization() bool {
|
|
return m != nil && (len(m.materializers) > 0 || len(m.missingFields) > 0)
|
|
}
|
|
|
|
type materializedRecord struct {
|
|
base storage.Record
|
|
computed map[int64]arrow.Array
|
|
}
|
|
|
|
var _ storage.Record = (*materializedRecord)(nil)
|
|
|
|
func (r *materializedRecord) Column(fieldID storage.FieldID) arrow.Array {
|
|
if col, ok := r.computed[fieldID]; ok {
|
|
return col
|
|
}
|
|
return r.base.Column(fieldID)
|
|
}
|
|
|
|
func (r *materializedRecord) Len() int {
|
|
return r.base.Len()
|
|
}
|
|
|
|
func (r *materializedRecord) Retain() {
|
|
r.base.Retain()
|
|
for _, col := range r.computed {
|
|
col.Retain()
|
|
}
|
|
}
|
|
|
|
func (r *materializedRecord) retainBase() {
|
|
r.base.Retain()
|
|
}
|
|
|
|
func (r *materializedRecord) Release() {
|
|
r.base.Release()
|
|
for _, col := range r.computed {
|
|
col.Release()
|
|
}
|
|
}
|
|
|
|
type selectedRecord struct {
|
|
base storage.Record
|
|
fields map[int64]*schemapb.FieldSchema
|
|
selection *recordSelection
|
|
columns map[int64]arrow.Array
|
|
err error
|
|
}
|
|
|
|
var _ storage.Record = (*selectedRecord)(nil)
|
|
|
|
func newSelectedRecord(base storage.Record, schema *schemapb.CollectionSchema, selection *recordSelection) *selectedRecord {
|
|
fields := make(map[int64]*schemapb.FieldSchema)
|
|
for _, field := range typeutil.GetAllFieldSchemas(schema) {
|
|
fields[field.GetFieldID()] = field
|
|
}
|
|
return &selectedRecord{
|
|
base: base,
|
|
fields: fields,
|
|
selection: selection,
|
|
columns: make(map[int64]arrow.Array),
|
|
}
|
|
}
|
|
|
|
func (r *selectedRecord) Column(fieldID storage.FieldID) arrow.Array {
|
|
if col, ok := r.columns[fieldID]; ok {
|
|
return col
|
|
}
|
|
field := r.fields[fieldID]
|
|
if field == nil {
|
|
return nil
|
|
}
|
|
builder := storage.NewRecordBuilder(&schemapb.CollectionSchema{Fields: []*schemapb.FieldSchema{field}})
|
|
defer builder.Release()
|
|
for _, rowRange := range r.selection.ranges {
|
|
if err := builder.Append(r.base, rowRange.start, rowRange.end); err != nil {
|
|
r.err = err
|
|
return nil
|
|
}
|
|
}
|
|
selected := builder.Build()
|
|
defer selected.Release()
|
|
col := selected.Column(fieldID)
|
|
if col == nil {
|
|
r.err = merr.WrapErrServiceInternalMsg("selected record field %d not found", fieldID)
|
|
return nil
|
|
}
|
|
col.Retain()
|
|
r.columns[fieldID] = col
|
|
return col
|
|
}
|
|
|
|
func (r *selectedRecord) Len() int {
|
|
return r.selection.Len()
|
|
}
|
|
|
|
func (r *selectedRecord) Retain() {
|
|
r.base.Retain()
|
|
for _, col := range r.columns {
|
|
col.Retain()
|
|
}
|
|
}
|
|
|
|
func (r *selectedRecord) Release() {
|
|
r.base.Release()
|
|
for _, col := range r.columns {
|
|
col.Release()
|
|
}
|
|
}
|
|
|
|
type materializedRecordReader struct {
|
|
base storage.RecordReader
|
|
materializer *RecordMaterializer
|
|
current storage.Record
|
|
}
|
|
|
|
var _ storage.RecordReader = (*materializedRecordReader)(nil)
|
|
|
|
func newMaterializedRecordReader(base storage.RecordReader, materializer *RecordMaterializer) storage.RecordReader {
|
|
if !materializer.hasMaterialization() {
|
|
return base
|
|
}
|
|
return &materializedRecordReader{base: base, materializer: materializer}
|
|
}
|
|
|
|
func (r *materializedRecordReader) Next() (storage.Record, error) {
|
|
if r.current != nil {
|
|
r.current.Release()
|
|
r.current = nil
|
|
}
|
|
rec, err := r.base.Next()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
wrapped, err := r.materializer.Wrap(rec)
|
|
if err != nil {
|
|
rec.Release()
|
|
return nil, err
|
|
}
|
|
if materialized, ok := wrapped.(*materializedRecord); ok {
|
|
materialized.retainBase()
|
|
} else {
|
|
wrapped.Retain()
|
|
}
|
|
r.current = wrapped
|
|
return wrapped, nil
|
|
}
|
|
|
|
func (r *materializedRecordReader) Close() error {
|
|
if r.current != nil {
|
|
r.current.Release()
|
|
r.current = nil
|
|
}
|
|
r.materializer.Close()
|
|
return r.base.Close()
|
|
}
|
|
|
|
type bm25FunctionMaterializer struct {
|
|
runner function.FunctionRunner
|
|
inputFieldIDs []int64
|
|
outputFieldIDs []int64
|
|
missingOutputIndexes []int
|
|
outputFields map[int64]*schemapb.FieldSchema
|
|
ownRunner bool
|
|
}
|
|
|
|
type minHashFunctionMaterializer struct {
|
|
runner function.FunctionRunner
|
|
inputFieldIDs []int64
|
|
outputFieldIDs []int64
|
|
missingOutputIndexes []int
|
|
outputFields map[int64]*schemapb.FieldSchema
|
|
ownRunner bool
|
|
}
|
|
|
|
var (
|
|
_ FunctionMaterializer = (*bm25FunctionMaterializer)(nil)
|
|
_ FunctionMaterializer = (*minHashFunctionMaterializer)(nil)
|
|
)
|
|
|
|
func newFunctionMaterializer(schema *schemapb.CollectionSchema, runner function.FunctionRunner, missingOutputIndexes []int, ownRunner bool) (FunctionMaterializer, error) {
|
|
functionSchema := runner.GetSchema()
|
|
switch functionSchema.GetType() {
|
|
case schemapb.FunctionType_BM25:
|
|
return newBM25FunctionMaterializer(schema, runner, missingOutputIndexes, ownRunner)
|
|
case schemapb.FunctionType_MinHash:
|
|
return newMinHashFunctionMaterializer(schema, runner, missingOutputIndexes, ownRunner)
|
|
default:
|
|
return nil, merr.WrapErrParameterInvalidMsg("unsupported function type %s", functionSchema.GetType().String())
|
|
}
|
|
}
|
|
|
|
func newMinHashFunctionMaterializer(schema *schemapb.CollectionSchema, runner function.FunctionRunner, missingOutputIndexes []int, ownRunner bool) (*minHashFunctionMaterializer, error) {
|
|
functionSchema := runner.GetSchema()
|
|
inputFields := runner.GetInputFields()
|
|
if len(inputFields) == 0 {
|
|
return nil, merr.WrapErrFunctionFailedMsg("minhash function should have input fields")
|
|
}
|
|
inputFieldIDs := make([]int64, 0, len(inputFields))
|
|
for _, inputField := range inputFields {
|
|
if inputField == nil || typeutil.GetField(schema, inputField.GetFieldID()) == nil {
|
|
return nil, merr.WrapErrFunctionFailedMsg("input field not found in schema")
|
|
}
|
|
if inputField.GetDataType() != schemapb.DataType_VarChar && inputField.GetDataType() != schemapb.DataType_Text {
|
|
return nil, merr.WrapErrFunctionFailedMsg("input field data type must be varchar or text for minhash function materialization")
|
|
}
|
|
inputFieldIDs = append(inputFieldIDs, inputField.GetFieldID())
|
|
}
|
|
|
|
outputFieldIDs := functionSchema.GetOutputFieldIds()
|
|
if len(outputFieldIDs) == 0 {
|
|
return nil, merr.WrapErrFunctionFailedMsg("minhash function should have output fields")
|
|
}
|
|
|
|
outputFields := make(map[int64]*schemapb.FieldSchema, len(outputFieldIDs))
|
|
for _, outputFieldID := range outputFieldIDs {
|
|
outputField := typeutil.GetField(schema, outputFieldID)
|
|
if outputField == nil {
|
|
return nil, merr.WrapErrFunctionFailedMsg("output field not found in schema")
|
|
}
|
|
if outputField.GetDataType() != schemapb.DataType_BinaryVector {
|
|
return nil, merr.WrapErrFunctionFailedMsg("output field data type must be binary vector for minhash function materialization")
|
|
}
|
|
if outputField.GetNullable() {
|
|
return nil, merr.WrapErrFunctionFailedMsg("function output field cannot be nullable: function %s, field %s", functionSchema.GetName(), outputField.GetName())
|
|
}
|
|
outputFields[outputFieldID] = outputField
|
|
}
|
|
|
|
return &minHashFunctionMaterializer{
|
|
runner: runner,
|
|
inputFieldIDs: inputFieldIDs,
|
|
outputFieldIDs: outputFieldIDs,
|
|
missingOutputIndexes: missingOutputIndexes,
|
|
outputFields: outputFields,
|
|
ownRunner: ownRunner,
|
|
}, nil
|
|
}
|
|
|
|
func newBM25FunctionMaterializer(schema *schemapb.CollectionSchema, runner function.FunctionRunner, missingOutputIndexes []int, ownRunner bool) (*bm25FunctionMaterializer, error) {
|
|
functionSchema := runner.GetSchema()
|
|
inputFields := runner.GetInputFields()
|
|
if len(inputFields) == 0 {
|
|
return nil, merr.WrapErrParameterInvalidMsg("bm25 function should have input fields")
|
|
}
|
|
inputFieldIDs := make([]int64, 0, len(inputFields))
|
|
for _, inputField := range inputFields {
|
|
if inputField == nil || typeutil.GetField(schema, inputField.GetFieldID()) == nil {
|
|
return nil, merr.WrapErrParameterInvalidMsg("input field not found in schema")
|
|
}
|
|
if inputField.GetDataType() != schemapb.DataType_VarChar && inputField.GetDataType() != schemapb.DataType_Text {
|
|
return nil, merr.WrapErrParameterInvalidMsg("input field data type must be varchar or text for bm25 function materialization")
|
|
}
|
|
inputFieldIDs = append(inputFieldIDs, inputField.GetFieldID())
|
|
}
|
|
|
|
outputFieldIDs := functionSchema.GetOutputFieldIds()
|
|
if len(outputFieldIDs) == 0 {
|
|
return nil, merr.WrapErrParameterInvalidMsg("bm25 function should have output fields")
|
|
}
|
|
|
|
outputFields := make(map[int64]*schemapb.FieldSchema, len(outputFieldIDs))
|
|
for _, outputFieldID := range outputFieldIDs {
|
|
outputField := typeutil.GetField(schema, outputFieldID)
|
|
if outputField == nil {
|
|
return nil, merr.WrapErrParameterInvalidMsg("output field not found in schema")
|
|
}
|
|
if outputField.GetDataType() != schemapb.DataType_SparseFloatVector {
|
|
return nil, merr.WrapErrParameterInvalidMsg("output field data type must be sparse float vector for bm25 function materialization")
|
|
}
|
|
if outputField.GetNullable() {
|
|
return nil, merr.WrapErrParameterInvalidMsg("function output field cannot be nullable: function %s, field %s", functionSchema.GetName(), outputField.GetName())
|
|
}
|
|
outputFields[outputFieldID] = outputField
|
|
}
|
|
|
|
return &bm25FunctionMaterializer{
|
|
runner: runner,
|
|
inputFieldIDs: inputFieldIDs,
|
|
outputFieldIDs: outputFieldIDs,
|
|
missingOutputIndexes: missingOutputIndexes,
|
|
outputFields: outputFields,
|
|
ownRunner: ownRunner,
|
|
}, nil
|
|
}
|
|
|
|
func (m *bm25FunctionMaterializer) Materialize(rec storage.Record) (map[int64]arrow.Array, error) {
|
|
inputs := make([]any, 0, len(m.inputFieldIDs))
|
|
for _, inputFieldID := range m.inputFieldIDs {
|
|
input, err := stringInputsFromRecord(rec, inputFieldID)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
inputs = append(inputs, input)
|
|
}
|
|
outputs, err := m.runner.BatchRun(inputs...)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
if len(outputs) != len(m.outputFieldIDs) {
|
|
return nil, merr.WrapErrFunctionFailedMsg("bm25 function materialization expects %d outputs, got %d", len(m.outputFieldIDs), len(outputs))
|
|
}
|
|
|
|
result := make(map[int64]arrow.Array, len(m.missingOutputIndexes))
|
|
for _, outputIndex := range m.missingOutputIndexes {
|
|
outputFieldID := m.outputFieldIDs[outputIndex]
|
|
outputSparseArray, ok := outputs[outputIndex].(*schemapb.SparseFloatArray)
|
|
if !ok {
|
|
releaseArrowArrays(result)
|
|
return nil, merr.WrapErrFunctionFailedMsg("unexpected output type from BM25 function runner, expected SparseFloatArray, got %T", outputs[outputIndex])
|
|
}
|
|
arr, err := buildSparseFloatVectorArrowArray(m.outputFields[outputFieldID], outputSparseArray, rec.Len())
|
|
if err != nil {
|
|
releaseArrowArrays(result)
|
|
return nil, err
|
|
}
|
|
result[outputFieldID] = arr
|
|
}
|
|
return result, nil
|
|
}
|
|
|
|
func (m *bm25FunctionMaterializer) Close() {
|
|
if m.ownRunner && m.runner != nil {
|
|
m.runner.Close()
|
|
}
|
|
}
|
|
|
|
func (m *minHashFunctionMaterializer) Materialize(rec storage.Record) (map[int64]arrow.Array, error) {
|
|
inputs := make([]any, 0, len(m.inputFieldIDs))
|
|
for _, inputFieldID := range m.inputFieldIDs {
|
|
input, err := stringInputsFromRecord(rec, inputFieldID)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
inputs = append(inputs, input)
|
|
}
|
|
outputs, err := m.runner.BatchRun(inputs...)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
if len(outputs) != len(m.outputFieldIDs) {
|
|
return nil, merr.WrapErrFunctionFailedMsg("minhash function materialization expects %d outputs, got %d", len(m.outputFieldIDs), len(outputs))
|
|
}
|
|
|
|
result := make(map[int64]arrow.Array, len(m.missingOutputIndexes))
|
|
for _, outputIndex := range m.missingOutputIndexes {
|
|
outputFieldID := m.outputFieldIDs[outputIndex]
|
|
outputFieldData, ok := outputs[outputIndex].(*schemapb.FieldData)
|
|
if !ok {
|
|
releaseArrowArrays(result)
|
|
return nil, merr.WrapErrFunctionFailedMsg("unexpected output type from MinHash function runner, expected FieldData, got %T", outputs[outputIndex])
|
|
}
|
|
vectorField := outputFieldData.GetVectors()
|
|
if vectorField == nil || vectorField.GetBinaryVector() == nil {
|
|
releaseArrowArrays(result)
|
|
return nil, merr.WrapErrFunctionFailedMsg("unexpected output from MinHash function runner, expected binary vector field data")
|
|
}
|
|
fieldData := &storage.BinaryVectorFieldData{
|
|
Data: vectorField.GetBinaryVector(),
|
|
Dim: int(vectorField.GetDim()),
|
|
}
|
|
if fieldData.RowNum() != rec.Len() {
|
|
releaseArrowArrays(result)
|
|
return nil, merr.WrapErrFunctionFailedMsg("minhash function output row count mismatch, expected %d, got %d", rec.Len(), fieldData.RowNum())
|
|
}
|
|
arr, err := buildArrowArrayFromFieldData(m.outputFields[outputFieldID], fieldData, rec.Len())
|
|
if err != nil {
|
|
releaseArrowArrays(result)
|
|
return nil, err
|
|
}
|
|
result[outputFieldID] = arr
|
|
}
|
|
return result, nil
|
|
}
|
|
|
|
func (m *minHashFunctionMaterializer) Close() {
|
|
if m.ownRunner && m.runner != nil {
|
|
m.runner.Close()
|
|
}
|
|
}
|
|
|
|
func functionOutputIndexesToMaterialize(functionSchema *schemapb.FunctionSchema, existingFields map[int64]struct{}) []int {
|
|
outputFieldIDs := functionSchema.GetOutputFieldIds()
|
|
indexes := make([]int, 0, len(outputFieldIDs))
|
|
hasMissingOutput := false
|
|
for idx, outputFieldID := range outputFieldIDs {
|
|
indexes = append(indexes, idx)
|
|
if _, ok := existingFields[outputFieldID]; !ok {
|
|
hasMissingOutput = true
|
|
}
|
|
}
|
|
if !hasMissingOutput {
|
|
return nil
|
|
}
|
|
return indexes
|
|
}
|
|
|
|
func missingNonMaterializedSchemaFields(schema *schemapb.CollectionSchema, existingFields map[int64]struct{}, materializedFields map[int64]struct{}) []*schemapb.FieldSchema {
|
|
missing := make([]*schemapb.FieldSchema, 0)
|
|
for _, field := range typeutil.GetAllFieldSchemas(schema) {
|
|
fieldID := field.GetFieldID()
|
|
if common.IsSystemField(fieldID) {
|
|
continue
|
|
}
|
|
if _, ok := existingFields[fieldID]; ok {
|
|
continue
|
|
}
|
|
if _, ok := materializedFields[fieldID]; ok {
|
|
continue
|
|
}
|
|
missing = append(missing, field)
|
|
}
|
|
return missing
|
|
}
|
|
|
|
func stringInputsFromRecord(rec storage.Record, fieldID int64) ([]string, error) {
|
|
col := rec.Column(fieldID)
|
|
if col == nil {
|
|
return nil, merr.WrapErrFunctionFailedMsg("input field %d not found in record", fieldID)
|
|
}
|
|
inputs := make([]string, rec.Len())
|
|
switch values := col.(type) {
|
|
case *array.String:
|
|
for i := 0; i < rec.Len(); i++ {
|
|
if values.IsValid(i) {
|
|
inputs[i] = values.Value(i)
|
|
}
|
|
}
|
|
case *array.Binary:
|
|
return nil, merr.WrapErrFunctionFailedMsg("cannot materialize bm25 from text binary values without lob decoding")
|
|
default:
|
|
return nil, merr.WrapErrFunctionFailedMsg("input field %d data type must be varchar or text for bm25 function materialization, got %T", fieldID, col)
|
|
}
|
|
return inputs, nil
|
|
}
|
|
|
|
func buildSparseFloatVectorArrowArray(field *schemapb.FieldSchema, outputSparseArray *schemapb.SparseFloatArray, rowCount int) (arrow.Array, error) {
|
|
if len(outputSparseArray.GetContents()) != rowCount {
|
|
return nil, merr.WrapErrFunctionFailedMsg("bm25 function output row count mismatch, expected %d, got %d", rowCount, len(outputSparseArray.GetContents()))
|
|
}
|
|
|
|
fieldData := &storage.SparseFloatVectorFieldData{
|
|
SparseFloatArray: schemapb.SparseFloatArray{
|
|
Contents: outputSparseArray.GetContents(),
|
|
Dim: outputSparseArray.GetDim(),
|
|
},
|
|
}
|
|
|
|
return buildArrowArrayFromFieldData(field, fieldData, rowCount)
|
|
}
|
|
|
|
func buildArrowArrayFromFieldData(field *schemapb.FieldSchema, fieldData storage.FieldData, rowCount int) (arrow.Array, error) {
|
|
if fieldData.RowNum() != rowCount {
|
|
return nil, merr.WrapErrFunctionFailedMsg("function output row count mismatch for field %d, expected %d, got %d", field.GetFieldID(), rowCount, fieldData.RowNum())
|
|
}
|
|
|
|
outputSchema := &schemapb.CollectionSchema{Fields: []*schemapb.FieldSchema{field}}
|
|
arrowSchema, err := storage.ConvertToArrowSchema(outputSchema, true)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
builder := array.NewRecordBuilder(memory.DefaultAllocator, arrowSchema)
|
|
defer builder.Release()
|
|
|
|
insertData := &storage.InsertData{Data: map[int64]storage.FieldData{
|
|
field.GetFieldID(): fieldData,
|
|
}}
|
|
if err := storage.BuildRecord(builder, insertData, outputSchema); err != nil {
|
|
return nil, err
|
|
}
|
|
record := builder.NewRecord()
|
|
defer record.Release()
|
|
|
|
col := record.Column(0)
|
|
col.Retain()
|
|
return col, nil
|
|
}
|
|
|
|
func releaseArrowArrays(arrays map[int64]arrow.Array) {
|
|
for _, arr := range arrays {
|
|
arr.Release()
|
|
}
|
|
}
|
|
|
|
func releaseWrappedRecord(wrapped storage.Record, base storage.Record) {
|
|
if wrapped != base {
|
|
wrapped.Release()
|
|
return
|
|
}
|
|
base.Release()
|
|
}
|