Use bytes instead of string internally.

calculation/result.go

@@ -8,6 +8,7 @@ import (
 
 	"github.com/xichen2020/eventdb/document/field"
 	"github.com/xichen2020/eventdb/generated/proto/servicepb"
+	"github.com/xichen2020/eventdb/x/bytes"
 )
 
 // Result represents a merge-able calculation result. In simple cases this can be simply
@@ -34,7 +35,7 @@ type Result interface {
 
 var (
 	nan         = math.NaN()
-	emptyString string
+	emptyString []byte
 
 	errMergingDifferentResultTypes = errors.New("merging calculation results with different result types")
 )
@@ -173,7 +174,7 @@ func (r *minNumberResult) MarshalJSON() ([]byte, error) { return json.Marshal(r.
 
 type minStringResult struct {
 	hasValues bool
-	v         string
+	v         []byte
 }
 
 // NewMinStringResult creates a new minimum string result.
@@ -187,7 +188,7 @@ func (r *minStringResult) Add(v ValueUnion) {
 		r.v = v.StringVal
 		return
 	}
-	if r.v > v.StringVal {
+	if bytes.GreaterThan(r.v, v.StringVal) {
 		r.v = v.StringVal
 	}
 }
@@ -204,7 +205,7 @@ func (r *minStringResult) MergeInPlace(other Result) error {
 		*r = *or
 		return nil
 	}
-	if r.v > or.v {
+	if bytes.GreaterThan(r.v, or.v) {
 		r.v = or.v
 	}
 	return nil
@@ -269,7 +270,7 @@ func (r *maxNumberResult) MarshalJSON() ([]byte, error) { return json.Marshal(r.
 
 type maxStringResult struct {
 	hasValues bool
-	v         string
+	v         []byte
 }
 
 // NewMaxStringResult creates a new maximum string result.
@@ -283,7 +284,7 @@ func (r *maxStringResult) Add(v ValueUnion) {
 		r.v = v.StringVal
 		return
 	}
-	if r.v < v.StringVal {
+	if bytes.LessThan(r.v, v.StringVal) {
 		r.v = v.StringVal
 	}
 }
@@ -300,7 +301,7 @@ func (r *maxStringResult) MergeInPlace(other Result) error {
 		*r = *or
 		return nil
 	}
-	if r.v < or.v {
+	if bytes.LessThan(r.v, or.v) {
 		r.v = or.v
 	}
 	return nil

calculation/value.go

@@ -7,7 +7,9 @@ import (
 
 	"github.com/xichen2020/eventdb/document/field"
 	"github.com/xichen2020/eventdb/generated/proto/servicepb"
+	"github.com/xichen2020/eventdb/x/bytes"
 	"github.com/xichen2020/eventdb/x/compare"
+	"github.com/xichen2020/eventdb/x/safe"
 )
 
 // ValueType is the type of a calculation value.
@@ -27,7 +29,7 @@ var (
 type ValueUnion struct {
 	Type      ValueType
 	NumberVal float64
-	StringVal string
+	StringVal []byte
 }
 
 // NewValueFromProto creates a value from protobuf message.
@@ -39,7 +41,7 @@ func NewValueFromProto(pbValue servicepb.CalculationValue) (ValueUnion, error) {
 		v.NumberVal = pbValue.NumberVal
 	case servicepb.CalculationValue_STRING:
 		v.Type = StringType
-		v.StringVal = pbValue.StringVal
+		v.StringVal = []byte(pbValue.StringVal)
 	default:
 		return v, fmt.Errorf("invalid protobuf calculation value type %v", pbValue.Type)
 	}
@@ -56,7 +58,7 @@ func (u ValueUnion) MarshalJSON() ([]byte, error) {
 		}
 		return json.Marshal(u.NumberVal)
 	case StringType:
-		return json.Marshal(u.StringVal)
+		return json.Marshal(safe.ToString(u.StringVal))
 	default:
 		return nil, fmt.Errorf("unexpected value type %v", u.Type)
 	}
@@ -73,7 +75,7 @@ func (u ValueUnion) ToProto() servicepb.CalculationValue {
 	case StringType:
 		return servicepb.CalculationValue{
 			Type:      servicepb.CalculationValue_STRING,
-			StringVal: u.StringVal,
+			StringVal: safe.ToString(u.StringVal),
 		}
 	default:
 		panic(fmt.Errorf("unexpected calculation value type %v", u.Type))
@@ -86,7 +88,7 @@ func NewNumberUnion(v float64) ValueUnion {
 }
 
 // NewStringUnion creates a new string union.
-func NewStringUnion(v string) ValueUnion {
+func NewStringUnion(v []byte) ValueUnion {
 	return ValueUnion{Type: StringType, StringVal: v}
 }
 
@@ -107,7 +109,7 @@ func CompareValue(v1, v2 ValueUnion) (int, error) {
 	case NumberType:
 		return compare.DoubleCompare(v1.NumberVal, v2.NumberVal), nil
 	case StringType:
-		return compare.StringCompare(v1.StringVal, v2.StringVal), nil
+		return bytes.Compare(v1.StringVal, v2.StringVal), nil
 	}
 	panic("should never reach here")
 }

document/field/field_test.go

@@ -10,7 +10,7 @@ func TestNewValuesLessThan(t *testing.T) {
 	v1 := Values{
 		{
 			Type:      StringType,
-			StringVal: "foo",
+			StringVal: []byte("foo"),
 		},
 		{
 			Type:   IntType,
@@ -25,7 +25,7 @@ func TestNewValuesLessThan(t *testing.T) {
 	v2 := Values{
 		{
 			Type:      StringType,
-			StringVal: "foo",
+			StringVal: []byte("foo"),
 		},
 		{
 			Type:   IntType,
@@ -50,7 +50,7 @@ func TestFilterValues(t *testing.T) {
 	v := Values{
 		{
 			Type:      StringType,
-			StringVal: "foo",
+			StringVal: []byte("foo"),
 		},
 		{
 			Type:   IntType,
@@ -82,7 +82,7 @@ func TestArrayBasedIterator(t *testing.T) {
 		},
 		{
 			Path:  []string{"baz"},
-			Value: NewStringUnion("blah"),
+			Value: NewStringUnion([]byte("blah")),
 		},
 	}
 	it := NewArrayBasedIterator(fields, nil)

document/field/value.go

@@ -6,6 +6,7 @@ import (
 	"math"
 
 	"github.com/xichen2020/eventdb/generated/proto/servicepb"
+	"github.com/xichen2020/eventdb/x/bytes"
 	"github.com/xichen2020/eventdb/x/compare"
 	"github.com/xichen2020/eventdb/x/safe"
 
@@ -137,7 +138,7 @@ type ValueUnion struct {
 	BoolVal      bool
 	IntVal       int
 	DoubleVal    float64
-	StringVal    string
+	StringVal    []byte
 	TimeNanosVal int64
 }
 
@@ -156,7 +157,7 @@ type IntAsUnionFn func(v int) ValueUnion
 type DoubleAsUnionFn func(v float64) ValueUnion
 
 // StringAsUnionFn converts a string to a value union.
-type StringAsUnionFn func(v string) ValueUnion
+type StringAsUnionFn func(v []byte) ValueUnion
 
 // TimeAsUnionFn converts a time to a value union.
 type TimeAsUnionFn func(v int64) ValueUnion
@@ -178,7 +179,7 @@ func NewValueFromProto(pbValue servicepb.FieldValue) (ValueUnion, error) {
 		v.DoubleVal = pbValue.DoubleVal
 	case servicepb.FieldValue_STRING:
 		v.Type = StringType
-		v.StringVal = safe.ToString(pbValue.StringVal)
+		v.StringVal = pbValue.StringVal
 	case servicepb.FieldValue_TIME:
 		v.Type = TimeType
 		v.TimeNanosVal = pbValue.TimeNanosVal
@@ -213,7 +214,7 @@ func NewDoubleUnion(v float64) ValueUnion {
 }
 
 // NewStringUnion creates a new string union.
-func NewStringUnion(v string) ValueUnion {
+func NewStringUnion(v []byte) ValueUnion {
 	return ValueUnion{
 		Type:      StringType,
 		StringVal: v,
@@ -240,7 +241,7 @@ func (v ValueUnion) MarshalJSON() ([]byte, error) {
 	case DoubleType:
 		return json.Marshal(v.DoubleVal)
 	case StringType:
-		return json.Marshal(v.StringVal)
+		return json.Marshal(safe.ToString(v.StringVal))
 	case TimeType:
 		return json.Marshal(v.TimeNanosVal)
 	default:
@@ -265,7 +266,7 @@ func (v *ValueUnion) ToProto() (servicepb.FieldValue, error) {
 		fb.DoubleVal = v.DoubleVal
 	case StringType:
 		fb.Type = servicepb.FieldValue_STRING
-		fb.StringVal = safe.ToBytes(v.StringVal)
+		fb.StringVal = v.StringVal
 	case TimeType:
 		fb.Type = servicepb.FieldValue_TIME
 		fb.TimeNanosVal = v.TimeNanosVal
@@ -296,7 +297,7 @@ func (v *ValueUnion) Equal(other *ValueUnion) bool {
 	case DoubleType:
 		return v.DoubleVal == other.DoubleVal
 	case StringType:
-		return v.StringVal == other.StringVal
+		return bytes.Equal(v.StringVal, other.StringVal)
 	case TimeType:
 		return v.TimeNanosVal == other.TimeNanosVal
 	}
@@ -324,7 +325,7 @@ func (v *ValueUnion) Hash() uint64 {
 		// NB(xichen): Hashing on bit patterns for doubles might be problematic.
 		return 31*hash + math.Float64bits(v.DoubleVal)
 	case StringType:
-		return 31*hash + xxhash.Sum64(safe.ToBytes(v.StringVal))
+		return 31*hash + xxhash.Sum64(v.StringVal)
 	case TimeType:
 		return 31*hash + uint64(v.TimeNanosVal)
 	}
@@ -354,7 +355,7 @@ func CompareValue(v1, v2 ValueUnion) (int, error) {
 	case DoubleType:
 		return compare.DoubleCompare(v1.DoubleVal, v2.DoubleVal), nil
 	case StringType:
-		return compare.StringCompare(v1.StringVal, v2.StringVal), nil
+		return bytes.Compare(v1.StringVal, v2.StringVal), nil
 	case TimeType:
 		return compare.TimeCompare(v1.TimeNanosVal, v2.TimeNanosVal), nil
 	default:

document/field/value_extract_bench_test.go

@@ -8,7 +8,7 @@ import "testing"
 var (
 	benchStringValue = ValueUnion{
 		Type:      StringType,
-		StringVal: "foo",
+		StringVal: []byte("foo"),
 	}
 	benchIntValue = ValueUnion{
 		Type:   IntType,

filter/filter_op.go

@@ -7,6 +7,7 @@ import (
 	"github.com/xichen2020/eventdb/document/field"
 	"github.com/xichen2020/eventdb/generated/proto/servicepb"
 	"github.com/xichen2020/eventdb/index"
+	"github.com/xichen2020/eventdb/x/bytes"
 )
 
 // Op represents a filter operator.
@@ -346,26 +347,26 @@ func (f Op) IntMaybeInRange(min, max, filterVal int) bool {
 // If this returns false, it means filterVal is definitely not within the value range [min, max].
 // If this returns true, it doesn't necessarily mean the filterVal exists in the values that this
 // filter is acting on.
-func (f Op) StringMaybeInRange(min, max, filterVal string) bool {
+func (f Op) StringMaybeInRange(min, max, filterVal []byte) bool {
 	switch f {
 	case Equals, StartsWith:
-		if filterVal > max || filterVal < min {
+		if bytes.GreaterThan(filterVal, max) || bytes.LessThan(filterVal, min) {
 			return false
 		}
 	case LargerThan:
-		if filterVal >= max {
+		if bytes.GreaterThanOrEqual(filterVal, max) {
 			return false
 		}
 	case LargerThanOrEqual:
-		if filterVal > max {
+		if bytes.GreaterThan(filterVal, max) {
 			return false
 		}
 	case SmallerThan:
-		if filterVal <= min {
+		if bytes.LessThanOrEqual(filterVal, min) {
 			return false
 		}
 	case SmallerThanOrEqual:
-		if filterVal < min {
+		if bytes.LessThan(filterVal, min) {
 			return false
 		}
 	}