Skip to content

fix(udf): replace state during JSON decoding #1626

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
fallintoplace wants to merge 3 commits into
base: main
Choose a base branch
from
Open

fix(udf): replace state during JSON decoding #1626

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d899051
fix(udf): replace metadata state during decoding
fallintoplace Aug 1, 2026
4a31904
test(udf): pin receiver replacement semantics
fallintoplace Aug 1, 2026
efc40bd
fix(udf): replace nested decoder state
fallintoplace Aug 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
71 changes: 39 additions & 32 deletions udf/metadata.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -204,24 +204,26 @@ type Parameter struct {

func (p *Parameter) UnmarshalJSON(b []byte) error {
type Alias Parameter
next := Parameter{}
aux := struct {
Type json.RawMessage `json:"type"`
*Alias
}{Alias: (*Alias)(p)}
}{Alias: (*Alias)(&next)}

if err := json.Unmarshal(b, &aux); err != nil {
return err
}

if len(aux.Type) == 0 {
return fmt.Errorf("%w: parameter %q requires a type", ErrInvalidUDFMetadata, p.Name)
return fmt.Errorf("%w: parameter %q requires a type", ErrInvalidUDFMetadata, next.Name)
}

typ, err := unmarshalType(aux.Type)
if err != nil {
return err
}
p.Type = typ
next.Type = typ
*p = next

return nil
}
Expand Down Expand Up @@ -252,26 +254,25 @@ type DefinitionVersion struct {

func (v *DefinitionVersion) UnmarshalJSON(b []byte) error {
type Alias DefinitionVersion
next := DefinitionVersion{VersionID: -1, TimestampMS: -1}
aux := struct {
Representations []json.RawMessage `json:"representations"`
*Alias
}{Alias: (*Alias)(v)}

v.VersionID = -1
v.TimestampMS = -1
}{Alias: (*Alias)(&next)}

if err := json.Unmarshal(b, &aux); err != nil {
return err
}

v.Representations = make([]Representation, len(aux.Representations))
next.Representations = make([]Representation, len(aux.Representations))
for i, raw := range aux.Representations {
repr, err := unmarshalRepresentation(raw)
if err != nil {
return err
}
v.Representations[i] = repr
next.Representations[i] = repr
}
*v = next

return nil
}
Expand Down Expand Up @@ -404,6 +405,7 @@ type Definition struct {

func (d *Definition) UnmarshalJSON(b []byte) error {
type Alias Definition
next := Definition{CurrentVersionID: -1}
// definition-id and parameters are shadowed to distinguish a required
// field that is absent from one that is legitimately empty: a
// zero-parameter definition has "" as its definition-id and [] as its
Expand All @@ -413,9 +415,7 @@ func (d *Definition) UnmarshalJSON(b []byte) error {
Parameters json.RawMessage `json:"parameters"`
ReturnType json.RawMessage `json:"return-type"`
*Alias
}{Alias: (*Alias)(d)}

d.CurrentVersionID = -1
}{Alias: (*Alias)(&next)}

if err := json.Unmarshal(b, &aux); err != nil {
return err
Expand All @@ -424,27 +424,28 @@ func (d *Definition) UnmarshalJSON(b []byte) error {
if aux.DefinitionID == nil {
return fmt.Errorf("%w: definition is missing definition-id", ErrInvalidUDFMetadata)
}
d.DefinitionID = *aux.DefinitionID
next.DefinitionID = *aux.DefinitionID

if len(aux.Parameters) == 0 {
return fmt.Errorf("%w: definition %q is missing parameters", ErrInvalidUDFMetadata, d.DefinitionID)
return fmt.Errorf("%w: definition %q is missing parameters", ErrInvalidUDFMetadata, next.DefinitionID)
}
if err := json.Unmarshal(aux.Parameters, &d.Parameters); err != nil {
if err := json.Unmarshal(aux.Parameters, &next.Parameters); err != nil {
return err
}
if d.Parameters == nil {
return fmt.Errorf("%w: definition %q is missing parameters", ErrInvalidUDFMetadata, d.DefinitionID)
if next.Parameters == nil {
return fmt.Errorf("%w: definition %q is missing parameters", ErrInvalidUDFMetadata, next.DefinitionID)
}

if len(aux.ReturnType) == 0 {
return fmt.Errorf("%w: definition %q requires a return-type", ErrInvalidUDFMetadata, d.DefinitionID)
return fmt.Errorf("%w: definition %q requires a return-type", ErrInvalidUDFMetadata, next.DefinitionID)
}

typ, err := unmarshalType(aux.ReturnType)
if err != nil {
return err
}
d.ReturnType = typ
next.ReturnType = typ
*d = next

return nil
}
Expand Down Expand Up @@ -605,14 +606,13 @@ type DefinitionVersionRef struct {

func (r *DefinitionVersionRef) UnmarshalJSON(b []byte) error {
type Alias DefinitionVersionRef
next := DefinitionVersionRef{VersionID: -1}
// definition-id is shadowed to distinguish the required field being
// absent from the legitimately empty id of a zero-parameter definition.
aux := struct {
DefinitionID *string `json:"definition-id"`
*Alias
}{Alias: (*Alias)(r)}

r.VersionID = -1
}{Alias: (*Alias)(&next)}

if err := json.Unmarshal(b, &aux); err != nil {
return err
Expand All @@ -621,7 +621,8 @@ func (r *DefinitionVersionRef) UnmarshalJSON(b []byte) error {
if aux.DefinitionID == nil {
return fmt.Errorf("%w: definition-log reference is missing definition-id", ErrInvalidUDFMetadata)
}
r.DefinitionID = *aux.DefinitionID
next.DefinitionID = *aux.DefinitionID
*r = next

return nil
}
Expand All @@ -639,11 +640,14 @@ type DefinitionLogEntry struct {

func (e *DefinitionLogEntry) UnmarshalJSON(b []byte) error {
type Alias DefinitionLogEntry
aux := (*Alias)(e)
next := DefinitionLogEntry{TimestampMS: -1}

e.TimestampMS = -1
if err := json.Unmarshal(b, (*Alias)(&next)); err != nil {
return err
}
*e = next

return json.Unmarshal(b, aux)
return nil
}

func (e DefinitionLogEntry) validate() error {
Expand Down Expand Up @@ -870,14 +874,13 @@ func (m *metadata) init() {

func (m *metadata) UnmarshalJSON(b []byte) error {
type Alias metadata
next := metadata{FormatVersionValue: -1}
// definition-log is shadowed to reject the required field being absent
// or null, which the default decoding cannot distinguish from empty.
aux := struct {
DefinitionLog json.RawMessage `json:"definition-log"`
*Alias
}{Alias: (*Alias)(m)}

m.FormatVersionValue = -1
}{Alias: (*Alias)(&next)}

if err := json.Unmarshal(b, &aux); err != nil {
return err
Expand All @@ -886,16 +889,20 @@ func (m *metadata) UnmarshalJSON(b []byte) error {
if len(aux.DefinitionLog) == 0 {
return fmt.Errorf("%w: definition-log is required", ErrInvalidUDFMetadata)
}
if err := json.Unmarshal(aux.DefinitionLog, &m.DefinitionLogList); err != nil {
if err := json.Unmarshal(aux.DefinitionLog, &next.DefinitionLogList); err != nil {
return err
}
if m.DefinitionLogList == nil {
if next.DefinitionLogList == nil {
return fmt.Errorf("%w: definition-log is required", ErrInvalidUDFMetadata)
}
if err := next.validate(); err != nil {
return err
}

*m = next
m.init()

return m.validate()
return nil
}

// indexBy indexes a slice into a map, using a provided extractKey function.
Expand Down
82 changes: 82 additions & 0 deletions udf/metadata_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -749,6 +749,88 @@ func TestParseMetadataWithEmptyDefinitionLog(t *testing.T) {
assert.Empty(t, meta.DefinitionLog())
}

func TestMetadataUnmarshalReplacesReceiverState(t *testing.T) {
firstDoc := minimalMetadata(t)
firstDoc["doc"] = "old documentation"
firstDoc["properties"] = map[string]any{"old": "value"}
first, err := json.Marshal(firstDoc)
require.NoError(t, err)

var meta metadata
require.NoError(t, json.Unmarshal(first, &meta))
_, ok := meta.DefinitionByID("int")
require.True(t, ok)

secondDoc := minimalMetadata(t)
secondDefinition := definition(t, secondDoc, 0)
secondDefinition["definition-id"] = "long"
secondDefinition["parameters"].([]any)[0].(map[string]any)["type"] = "long"
refs := secondDoc["definition-log"].([]any)[0].(map[string]any)["definition-versions"].([]any)
refs[0].(map[string]any)["definition-id"] = "long"
second, err := json.Marshal(secondDoc)
require.NoError(t, err)
require.NoError(t, json.Unmarshal(second, &meta))

assert.Empty(t, meta.Doc())
assert.Empty(t, meta.Properties())
_, ok = meta.DefinitionByID("int")
assert.False(t, ok)
_, ok = meta.DefinitionByID("long")
assert.True(t, ok)

delete(secondDoc, "definitions")
missingDefinitions, err := json.Marshal(secondDoc)
require.NoError(t, err)
err = json.Unmarshal(missingDefinitions, &meta)
assert.ErrorContains(t, err, "at least one definition is required")
_, ok = meta.DefinitionByID("long")
assert.True(t, ok, "a failed decode must not partially replace valid metadata")
}

func TestUDFValuesUnmarshalReplaceReceiverState(t *testing.T) {
t.Run("parameter", func(t *testing.T) {
var parameter Parameter
require.NoError(t, json.Unmarshal([]byte(`{"name":"x","type":"int","doc":"old"}`), &parameter))
require.NoError(t, json.Unmarshal([]byte(`{"name":"y","type":"long"}`), &parameter))
assert.Equal(t, "y", parameter.Name)
assert.Equal(t, "long", parameter.Type.String())
assert.Empty(t, parameter.Doc)
})

t.Run("definition version", func(t *testing.T) {
var version DefinitionVersion
first := `{"version-id":1,"representations":[{"type":"sql","dialect":"trino","sql":"x"}],"deterministic":true,"on-null-input":"return-null","timestamp-ms":1}`
second := `{"version-id":2,"representations":[{"type":"sql","dialect":"spark","sql":"y"}],"timestamp-ms":2}`
require.NoError(t, json.Unmarshal([]byte(first), &version))
require.NoError(t, json.Unmarshal([]byte(second), &version))
assert.Equal(t, 2, version.VersionID)
assert.False(t, version.Deterministic)
assert.Empty(t, version.OnNullInput)
})

t.Run("definition", func(t *testing.T) {
firstDoc := definition(t, minimalMetadata(t), 0)
firstDoc["specific-name"] = "old_name"
firstDoc["doc"] = "old documentation"
first, err := json.Marshal(firstDoc)
require.NoError(t, err)
second, err := json.Marshal(definition(t, minimalMetadata(t), 0))
require.NoError(t, err)

var def Definition
require.NoError(t, json.Unmarshal(first, &def))
require.NoError(t, json.Unmarshal(second, &def))
assert.Empty(t, def.SpecificName)
assert.Empty(t, def.Doc)
})

t.Run("definition log entry", func(t *testing.T) {
entry := DefinitionLogEntry{DefinitionVersions: []DefinitionVersionRef{{DefinitionID: "int", VersionID: 1}}}
require.NoError(t, json.Unmarshal([]byte(`{"timestamp-ms":2}`), &entry))
assert.Empty(t, entry.DefinitionVersions)
})
}

func TestOnNullInputValues(t *testing.T) {
for _, behavior := range []OnNullInput{OnNullInputCall, OnNullInputReturnNull} {
m := minimalMetadata(t)
Expand Down
Loading