feat: introduce UIntNode interface, used within DAG-CBOR codec

Author: rvagg

codec/dagcbor/marshal.go

@@ -99,13 +99,22 @@ func marshal(n datamodel.Node, tk *tok.Token, sink shared.TokenSink, options Enc
 		_, err = sink.Step(tk)
 		return err
 	case datamodel.Kind_Int:
-		v, err := n.AsInt()
-		if err != nil {
-			return err
+		if uin, ok := n.(datamodel.UintNode); ok {
+			v, err := uin.AsUint()
+			if err != nil {
+				return err
+			}
+			tk.Type = tok.TUint
+			tk.Uint = v
+		} else {
+			v, err := n.AsInt()
+			if err != nil {
+				return err
+			}
+			tk.Type = tok.TInt
+			tk.Int = v
 		}
-		tk.Type = tok.TInt
-		tk.Int = int64(v)
-		_, err = sink.Step(tk)
+		_, err := sink.Step(tk)
 		return err
 	case datamodel.Kind_Float:
 		v, err := n.AsFloat()

codec/dagcbor/roundtrip_test.go

@@ -3,12 +3,15 @@ package dagcbor
 import (
 	"bytes"
 	"crypto/rand"
+	"encoding/hex"
+	"math"
 	"strings"
 	"testing"
 
 	qt "github.com/frankban/quicktest"
 	cid "github.com/ipfs/go-cid"
 
+	"github.com/ipld/go-ipld-prime/datamodel"
 	"github.com/ipld/go-ipld-prime/fluent"
 	cidlink "github.com/ipld/go-ipld-prime/linking/cid"
 	"github.com/ipld/go-ipld-prime/node/basicnode"
@@ -115,3 +118,65 @@ func TestRoundtripLinksAndBytes(t *testing.T) {
 	reconstructed := nb.Build()
 	qt.Check(t, reconstructed, nodetests.NodeContentEquals, linkByteNode)
 }
+
+func TestInts(t *testing.T) {
+	data := []struct {
+		name      string
+		hex       string
+		value     uint64
+		intValue  int64
+		intErr    string
+		decodeErr string
+	}{
+		{"max uint64", "1bffffffffffffffff", math.MaxUint64, 0, "unsigned integer out of range of int64 type", ""},
+		{"max int64", "1b7fffffffffffffff", math.MaxInt64, math.MaxInt64, "", ""},
+		{"1", "01", 1, 1, "", ""},
+		{"0", "00", 0, 0, "", ""},
+		{"-1", "20", 0, -1, "", ""},
+		{"min int64", "3b7fffffffffffffff", 0, math.MinInt64, "", ""},
+		{"~min uint64", "3bfffffffffffffffe", 0, 0, "", "cbor: negative integer out of rage of int64 type"},
+		// TODO: 3bffffffffffffffff isn't properly handled by refmt, it's coerced to zero
+		// MaxUint64 gets overflowed here: https://github.com/polydawn/refmt/blob/30ac6d18308e584ca6a2e74ba81475559db94c5f/cbor/cborDecoderTerminals.go#L75
+	}
+
+	for _, td := range data {
+		t.Run(td.name, func(t *testing.T) {
+			buf, err := hex.DecodeString(td.hex) // max uint64
+			qt.Assert(t, err, qt.IsNil)
+			nb := basicnode.Prototype.Any.NewBuilder()
+			err = Decode(nb, bytes.NewReader(buf))
+			if td.decodeErr != "" {
+				qt.Assert(t, err, qt.IsNotNil)
+				qt.Assert(t, err.Error(), qt.Equals, td.decodeErr)
+				return
+			}
+			qt.Assert(t, err, qt.IsNil)
+			n := nb.Build()
+
+			ii, err := n.AsInt()
+			if td.intErr != "" {
+				qt.Assert(t, err.Error(), qt.Equals, td.intErr)
+			} else {
+				qt.Assert(t, err, qt.IsNil)
+				qt.Assert(t, ii, qt.Equals, int64(td.intValue))
+			}
+
+			// if the number is outside of the positive int64 range, we should be able
+			// to access it as a UintNode and be able to access the full int64 range
+			uin, ok := n.(datamodel.UintNode)
+			if td.value <= math.MaxInt64 {
+				qt.Assert(t, ok, qt.IsFalse)
+			} else {
+				qt.Assert(t, ok, qt.IsTrue)
+				val, err := uin.AsUint()
+				qt.Assert(t, err, qt.IsNil)
+				qt.Assert(t, val, qt.Equals, uint64(td.value))
+			}
+
+			var byts bytes.Buffer
+			err = Encode(n, &byts)
+			qt.Assert(t, err, qt.IsNil)
+			qt.Assert(t, hex.EncodeToString(byts.Bytes()), qt.Equals, td.hex)
+		})
+	}
+}

codec/dagcbor/unmarshal.go

@@ -13,6 +13,7 @@ import (
 
 	"github.com/ipld/go-ipld-prime/datamodel"
 	cidlink "github.com/ipld/go-ipld-prime/linking/cid"
+	"github.com/ipld/go-ipld-prime/node/basicnode"
 )
 
 var (
@@ -275,7 +276,12 @@ func unmarshal2(na datamodel.NodeAssembler, tokSrc shared.TokenSource, tk *tok.T
 		if *gas < 0 {
 			return ErrAllocationBudgetExceeded
 		}
-		return na.AssignInt(int64(tk.Uint)) // FIXME overflow check
+		// note that this pushes any overflow errors up the stack when AsInt() may
+		// be called on a UintNode that is too large to cast to an int64
+		if tk.Uint > math.MaxInt64 {
+			return na.AssignNode(basicnode.NewUint(tk.Uint))
+		}
+		return na.AssignInt(int64(tk.Uint))
 	case tok.TFloat64:
 		*gas -= 1
 		if *gas < 0 {

datamodel/node.go

@@ -167,6 +167,21 @@ type Node interface {
 	Prototype() NodePrototype
 }
 
+// UintNode is an optional interface that can be used to represent an Int node
+// that provides access to the full uint64 range.
+//
+// EXPERIMENTAL: this API is experimental and may be changed or removed in a
+// future use. A future iteration may replace this with a BigInt interface to
+// access a larger range of integers that may be enabled by alternative codecs.
+type UintNode interface {
+	Node
+
+	// AsUint returns a uint64 representing the underlying integer if possible.
+	// This may return an error if the Node represents a negative integer that
+	// cannot be represented as a uint64.
+	AsUint() (uint64, error)
+}
+
 // LargeBytesNode is an optional interface extending a Bytes node that allows its
 // contents to be accessed through an io.ReadSeeker instead of a []byte slice. Use of
 // an io.Reader is encouraged, as it allows for streaming large byte slices

node/basicnode/int.go

@@ -1,26 +1,40 @@
 package basicnode
 
 import (
+	"fmt"
+	"math"
+
 	"github.com/ipld/go-ipld-prime/datamodel"
 	"github.com/ipld/go-ipld-prime/node/mixins"
 )
 
 var (
 	_ datamodel.Node          = plainInt(0)
+	_ datamodel.Node          = plainUint(0)
+	_ datamodel.UintNode      = plainUint(0)
 	_ datamodel.NodePrototype = Prototype__Int{}
 	_ datamodel.NodeBuilder   = &plainInt__Builder{}
 	_ datamodel.NodeAssembler = &plainInt__Assembler{}
 )
 
 func NewInt(value int64) datamodel.Node {
-	v := plainInt(value)
-	return &v
+	return plainInt(value)
+}
+
+// NewUint creates a new uint64-backed Node which will behave as a plain Int
+// node but also conforms to the datamodel.UintNode interface which can access
+// the full uint64 range.
+//
+// EXPERIMENTAL: this API is experimental and may be changed or removed in a
+// future release.
+func NewUint(value uint64) datamodel.Node {
+	return plainUint(value)
 }
 
 // plainInt is a simple boxed int that complies with datamodel.Node.
 type plainInt int64
 
-// -- Node interface methods -->
+// -- Node interface methods for plainInt -->
 
 func (plainInt) Kind() datamodel.Kind {
 	return datamodel.Kind_Int
@@ -74,6 +88,74 @@ func (plainInt) Prototype() datamodel.NodePrototype {
 	return Prototype__Int{}
 }
 
+// plainUint is a simple boxed uint64 that complies with datamodel.Node,
+// allowing representation of the uint64 range above the int64 maximum via the
+// UintNode interface
+type plainUint uint64
+
+// -- Node interface methods for plainUint -->
+
+func (plainUint) Kind() datamodel.Kind {
+	return datamodel.Kind_Int
+}
+func (plainUint) LookupByString(string) (datamodel.Node, error) {
+	return mixins.Int{TypeName: "int"}.LookupByString("")
+}
+func (plainUint) LookupByNode(key datamodel.Node) (datamodel.Node, error) {
+	return mixins.Int{TypeName: "int"}.LookupByNode(nil)
+}
+func (plainUint) LookupByIndex(idx int64) (datamodel.Node, error) {
+	return mixins.Int{TypeName: "int"}.LookupByIndex(0)
+}
+func (plainUint) LookupBySegment(seg datamodel.PathSegment) (datamodel.Node, error) {
+	return mixins.Int{TypeName: "int"}.LookupBySegment(seg)
+}
+func (plainUint) MapIterator() datamodel.MapIterator {
+	return nil
+}
+func (plainUint) ListIterator() datamodel.ListIterator {
+	return nil
+}
+func (plainUint) Length() int64 {
+	return -1
+}
+func (plainUint) IsAbsent() bool {
+	return false
+}
+func (plainUint) IsNull() bool {
+	return false
+}
+func (plainUint) AsBool() (bool, error) {
+	return mixins.Int{TypeName: "int"}.AsBool()
+}
+func (n plainUint) AsInt() (int64, error) {
+	if uint64(n) > uint64(math.MaxInt64) {
+		return -1, fmt.Errorf("unsigned integer out of range of int64 type")
+	}
+	return int64(n), nil
+}
+func (plainUint) AsFloat() (float64, error) {
+	return mixins.Int{TypeName: "int"}.AsFloat()
+}
+func (plainUint) AsString() (string, error) {
+	return mixins.Int{TypeName: "int"}.AsString()
+}
+func (plainUint) AsBytes() ([]byte, error) {
+	return mixins.Int{TypeName: "int"}.AsBytes()
+}
+func (plainUint) AsLink() (datamodel.Link, error) {
+	return mixins.Int{TypeName: "int"}.AsLink()
+}
+func (plainUint) Prototype() datamodel.NodePrototype {
+	return Prototype__Int{}
+}
+
+// allows plainUint to conform to the plainUint interface
+
+func (n plainUint) AsUint() (uint64, error) {
+	return uint64(n), nil
+}
+
 // -- NodePrototype -->
 
 type Prototype__Int struct{}