node.go

package merkledag

import (
	"context"
	"encoding/json"
	"errors"
	"fmt"
	"math"
	"sort"

	blocks "github.com/ipfs/go-block-format"
	cid "github.com/ipfs/go-cid"
	format "github.com/ipfs/go-ipld-format"
	legacy "github.com/ipfs/go-ipld-legacy"
	logging "github.com/ipfs/go-log/v2"
	dagpb "github.com/ipld/go-codec-dagpb"
	ipld "github.com/ipld/go-ipld-prime"
	mh "github.com/multiformats/go-multihash"
	mhcore "github.com/multiformats/go-multihash/core"
)

// Common errors
var (
	ErrNotProtobuf  = fmt.Errorf("expected protobuf dag node")
	ErrNotRawNode   = fmt.Errorf("expected raw bytes node")
	ErrLinkNotFound = fmt.Errorf("no link by that name")
)

var log = logging.Logger("merkledag")

// for testing custom CidBuilders
var (
	zeros   [256]byte
	zeroCid = mustZeroCid()
)

type immutableProtoNode struct {
	encoded []byte
	dagpb.PBNode
}

// ProtoNode represents a node in the IPFS Merkle DAG.
// nodes have opaque data and a set of navigable links.
// ProtoNode is a go-ipld-legacy.UniversalNode, meaning it is both
// a go-ipld-prime node and a go-ipld-format node.
// ProtoNode maintains compatibility with it's original implementation
// as a go-ipld-format only node, which included some mutability, namely the
// the ability to add/remove links in place
//
// TODO: We should be able to eventually replace this implementation with
// * go-codec-dagpb for basic DagPB encode/decode to go-ipld-prime
// * go-unixfsnode ADLs for higher level DAGPB functionality
// For the time being however, go-unixfsnode is read only and
// this mutable protonode implementation is needed to support go-unixfs,
// the only library that implements both read and write for UnixFS v1.
type ProtoNode struct {
	links      []*format.Link
	linksDirty bool
	data       []byte

	// cache encoded/marshaled value, kept to make the go-ipld-prime Node interface
	// work (see prime.go), and to provide a cached []byte encoded form available
	encoded *immutableProtoNode
	cached  cid.Cid

	// builder specifies cid version and hashing function
	builder cid.Builder
}

var v0CidPrefix = cid.Prefix{
	Codec:    cid.DagProtobuf,
	MhLength: -1,
	MhType:   mh.SHA2_256,
	Version:  0,
}

var v1CidPrefix = cid.Prefix{
	Codec:    cid.DagProtobuf,
	MhLength: -1,
	MhType:   mh.SHA2_256,
	Version:  1,
}

// V0CidPrefix returns a prefix for CIDv0
func V0CidPrefix() cid.Prefix { return v0CidPrefix }

// V1CidPrefix returns a prefix for CIDv1 with the default settings
func V1CidPrefix() cid.Prefix { return v1CidPrefix }

// PrefixForCidVersion returns the Protobuf prefix for a given CID version
func PrefixForCidVersion(version int) (cid.Prefix, error) {
	switch version {
	case 0:
		return v0CidPrefix, nil
	case 1:
		return v1CidPrefix, nil
	default:
		return cid.Prefix{}, fmt.Errorf("unknown CID version: %d", version)
	}
}

// CidBuilder returns the CID Builder for this ProtoNode, it is never nil
func (n *ProtoNode) CidBuilder() cid.Builder {
	if n.builder == nil {
		n.builder = v0CidPrefix
	}
	return n.builder
}

// SetCidBuilder sets the CID builder if it is non nil, if nil then it
// is reset to the default value. An error will be returned if the builder
// is not usable.
func (n *ProtoNode) SetCidBuilder(builder cid.Builder) error {
	if builder == nil {
		n.builder = v0CidPrefix
		return nil
	}
	switch b := builder.(type) {
	case cid.Prefix:
		if err := checkHasher(b.MhType, b.MhLength); err != nil {
			return err
		}
	case *cid.Prefix:
		if err := checkHasher(b.MhType, b.MhLength); err != nil {
			return err
		}
	default:
		// We have to test it's a usable hasher by invoking it and checking it
		// doesn't error. This is only a basic check, there are still ways it may
		// break
		if _, err := builder.Sum(zeros[:]); err != nil {
			return err
		}
	}
	n.builder = builder.WithCodec(cid.DagProtobuf)
	n.cached = cid.Undef
	return nil
}

// check whether the hasher is likely to be a usable one
func checkHasher(indicator uint64, sizeHint int) error {
	mhLen := sizeHint
	if mhLen <= 0 {
		mhLen = -1
	}
	_, err := mhcore.GetVariableHasher(indicator, mhLen)
	return err
}

// LinkSlice is a slice of format.Links
type LinkSlice []*format.Link

func (ls LinkSlice) Len() int           { return len(ls) }
func (ls LinkSlice) Swap(a, b int)      { ls[a], ls[b] = ls[b], ls[a] }
func (ls LinkSlice) Less(a, b int) bool { return ls[a].Name < ls[b].Name }

// NodeWithData builds a new Protonode with the given data.
func NodeWithData(d []byte) *ProtoNode {
	return &ProtoNode{data: d}
}

// AddNodeLink adds a link to another node. The link will be added in
// sorted order.
//
// If sorting has not already been applied to this node (because
// it was deserialized from a form that did not have sorted links), the links
// list will be sorted. If a ProtoNode was deserialized from a badly encoded
// form that did not already have its links sorted, calling AddNodeLink and then
// RemoveNodeLink for the same link, will not result in an identically encoded
// form as the links will have been sorted.
func (n *ProtoNode) AddNodeLink(name string, that format.Node) error {
	lnk, err := format.MakeLink(that)
	if err != nil {
		return err
	}

	lnk.Name = name

	n.AddRawLink(name, lnk)

	return nil
}

// AddRawLink adds a copy of a link to this node. The link will be added in
// sorted order.
//
// If sorting has not already been applied to this node (because
// it was deserialized from a form that did not have sorted links), the links
// list will be sorted. If a ProtoNode was deserialized from a badly encoded
// form that did not already have its links sorted, calling AddRawLink and then
// RemoveNodeLink for the same link, will not result in an identically encoded
// form as the links will have been sorted.
func (n *ProtoNode) AddRawLink(name string, l *format.Link) error {
	lnk := &format.Link{
		Name: name,
		Size: l.Size,
		Cid:  l.Cid,
	}
	if err := checkLink(lnk); err != nil {
		return err
	}
	n.links = append(n.links, lnk)
	n.linksDirty = true // needs a sort
	n.encoded = nil
	return nil
}

// RemoveNodeLink removes a link on this node by the given name. If there are
// no links with this name, ErrLinkNotFound will be returned. If there are more
// than one link with this name, they will all be removed.
func (n *ProtoNode) RemoveNodeLink(name string) error {
	ref := n.links[:0]
	found := false

	for _, v := range n.links {
		if v.Name != name {
			ref = append(ref, v)
		} else {
			found = true
		}
	}

	if !found {
		return ErrLinkNotFound
	}

	n.links = ref
	// Even though a removal won't change sorting, this node may have come from
	// a deserialized state with badly sorted links. Now that we are mutating,
	// we need to ensure the resulting link list is sorted when it gets consumed.
	n.linksDirty = true
	n.encoded = nil

	return nil
}

// GetNodeLink returns a copy of the link with the given name.
func (n *ProtoNode) GetNodeLink(name string) (*format.Link, error) {
	for _, l := range n.links {
		if l.Name == name {
			return &format.Link{
				Name: l.Name,
				Size: l.Size,
				Cid:  l.Cid,
			}, nil
		}
	}
	return nil, ErrLinkNotFound
}

// GetLinkedProtoNode returns a copy of the ProtoNode with the given name.
func (n *ProtoNode) GetLinkedProtoNode(ctx context.Context, ds format.DAGService, name string) (*ProtoNode, error) {
	nd, err := n.GetLinkedNode(ctx, ds, name)
	if err != nil {
		return nil, err
	}

	pbnd, ok := nd.(*ProtoNode)
	if !ok {
		return nil, ErrNotProtobuf
	}

	return pbnd, nil
}

// GetLinkedNode returns a copy of the IPLD Node with the given name.
func (n *ProtoNode) GetLinkedNode(ctx context.Context, ds format.DAGService, name string) (format.Node, error) {
	lnk, err := n.GetNodeLink(name)
	if err != nil {
		return nil, err
	}

	return lnk.GetNode(ctx, ds)
}

// Copy returns a copy of the node. The resulting node will have a properly
// sorted Links list regardless of whether the original came from a badly
// serialized form that didn't have a sorted list.
// NOTE: This does not make copies of Node objects in the links.
func (n *ProtoNode) Copy() format.Node {
	nnode := new(ProtoNode)
	if len(n.data) > 0 {
		nnode.data = make([]byte, len(n.data))
		copy(nnode.data, n.data)
	}

	if len(n.links) > 0 {
		nnode.links = append([]*format.Link(nil), n.links...)
		// Sort links regardless of linksDirty state, this may have come from a
		// serialized form that had badly sorted links, in which case linksDirty
		// will not be true.
		sort.Stable(LinkSlice(nnode.links))
	}

	nnode.builder = n.builder

	return nnode
}

// RawData returns the encoded byte form of this node.
//
// Note that this method may return an empty byte slice if there is an error
// performing the encode. To check whether such an error may have occurred, use
// node.EncodeProtobuf(false), instead (or prior to calling RawData) and check
// for its returned error value; the result of EncodeProtobuf is cached so there
// is minimal overhead when invoking both methods.
func (n *ProtoNode) RawData() []byte {
	out, err := n.EncodeProtobuf(false)
	if err != nil {
		log.Errorf("failed to encode dag-pb block: %s", err.Error())
		return nil
	}
	return out
}

// Data returns the data stored by this node.
func (n *ProtoNode) Data() []byte {
	return n.data
}

// SetData stores data in this nodes.
func (n *ProtoNode) SetData(d []byte) {
	n.encoded = nil
	n.cached = cid.Undef
	n.data = d
}

// UpdateNodeLink return a copy of the node with the link name set to point to
// that. The link will be added in sorted order. If a link of the same name
// existed, it is removed.
//
// If sorting has not already been applied to this node (because
// it was deserialized from a form that did not have sorted links), the links
// list will be sorted in the returned copy.
func (n *ProtoNode) UpdateNodeLink(name string, that *ProtoNode) (*ProtoNode, error) {
	newnode := n.Copy().(*ProtoNode)
	_ = newnode.RemoveNodeLink(name) // ignore error
	err := newnode.AddNodeLink(name, that)
	return newnode, err
}

// Size returns the total size of the data addressed by node,
// including the total sizes of references.
func (n *ProtoNode) Size() (uint64, error) {
	b, err := n.EncodeProtobuf(false)
	if err != nil {
		return 0, err
	}

	s := uint64(len(b))
	for _, l := range n.links {
		s += l.Size
	}
	return s, nil
}

// Stat returns statistics on the node.
func (n *ProtoNode) Stat() (*format.NodeStat, error) {
	enc, err := n.EncodeProtobuf(false)
	if err != nil {
		return nil, err
	}

	cumSize, err := n.Size()
	if err != nil {
		return nil, err
	}

	return &format.NodeStat{
		Hash:           n.Cid().String(),
		NumLinks:       len(n.links),
		BlockSize:      len(enc),
		LinksSize:      len(enc) - len(n.data), // includes framing.
		DataSize:       len(n.data),
		CumulativeSize: int(cumSize),
	}, nil
}

// Loggable implements the ipfs/go-log.Loggable interface.
func (n *ProtoNode) Loggable() map[string]interface{} {
	return map[string]interface{}{
		"node": n.String(),
	}
}

// UnmarshalJSON reads the node fields from a JSON-encoded byte slice.
func (n *ProtoNode) UnmarshalJSON(b []byte) error {
	s := struct {
		Data  []byte         `json:"data"`
		Links []*format.Link `json:"links"`
	}{}

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

	n.data = s.Data
	// Links may not be sorted after deserialization, but we don't change
	// them until we mutate this node since we're representing the current,
	// as-serialized state. So n.linksDirty is not set here.
	n.links = s.Links
	for _, lnk := range s.Links {
		if err := checkLink(lnk); err != nil {
			return err
		}
	}

	n.encoded = nil
	return nil
}

func checkLink(lnk *format.Link) error {
	if lnk.Size > math.MaxInt64 {
		return fmt.Errorf("value of Tsize is too large: %d", lnk.Size)
	}
	if !lnk.Cid.Defined() {
		return errors.New("link must have a value Cid value")
	}
	return nil
}

// MarshalJSON returns a JSON representation of the node.
func (n *ProtoNode) MarshalJSON() ([]byte, error) {
	if n.linksDirty {
		// there was a mutation involving links, make sure we sort
		sort.Stable(LinkSlice(n.links))
		n.linksDirty = false
		n.encoded = nil
	}

	out := map[string]interface{}{
		"data":  n.data,
		"links": n.links,
	}

	return json.Marshal(out)
}

// Cid returns the node's Cid, calculated according to its prefix
// and raw data contents.
//
// Note that this method may return a CID representing a zero-length byte slice
// if there is an error performing the encode. To check whether such an error
// may have occurred, use node.EncodeProtobuf(false), instead (or prior to
// calling RawData) and check for its returned error value; the result of
// EncodeProtobuf is cached so there is minimal overhead when invoking both
// methods.
func (n *ProtoNode) Cid() cid.Cid {
	// re-encode if necessary and we'll get a new cached CID
	if _, err := n.EncodeProtobuf(false); err != nil {
		log.Errorf("failed to encode dag-pb block: %s", err.Error())
		// error, return a zero-CID
		c, err := n.CidBuilder().Sum([]byte{})
		if err != nil {
			// CidBuilder was a source of error, return _the_ dag-pb zero CIDv1
			return zeroCid
		}
		return c
	}
	return n.cached
}

// String prints the node's Cid.
//
// Note that this method may return a CID representing a zero-length byte slice
// if there is an error performing the encode. To check whether such an error
// may have occurred, use node.EncodeProtobuf(false), instead (or prior to
// calling RawData) and check for its returned error value; the result of
// EncodeProtobuf is cached so there is minimal overhead when invoking both
// methods.
func (n *ProtoNode) String() string {
	return n.Cid().String()
}

// Multihash hashes the encoded data of this node.
//
// Note that this method may return a multihash representing a zero-length byte
// slice if there is an error performing the encode. To check whether such an
// error may have occurred, use node.EncodeProtobuf(false), instead (or prior to
// calling RawData) and check for its returned error value; the result of
// EncodeProtobuf is cached so there is minimal overhead when invoking both
// methods.
func (n *ProtoNode) Multihash() mh.Multihash {
	return n.Cid().Hash()
}

// Links returns a copy of the node's links.
func (n *ProtoNode) Links() []*format.Link {
	if n.linksDirty {
		// there was a mutation involving links, make sure we sort
		sort.Stable(LinkSlice(n.links))
		n.linksDirty = false
		n.encoded = nil
	}
	return append([]*format.Link(nil), n.links...)
}

// SetLinks replaces the node links with a copy of the provided links. Sorting
// will be applied to the list.
func (n *ProtoNode) SetLinks(links []*format.Link) error {
	for _, lnk := range links {
		if err := checkLink(lnk); err != nil {
			return err
		}
	}
	n.links = append([]*format.Link(nil), links...)
	n.linksDirty = true // needs a sort
	n.encoded = nil
	return nil
}

// Resolve is an alias for ResolveLink.
func (n *ProtoNode) Resolve(path []string) (interface{}, []string, error) {
	return n.ResolveLink(path)
}

// ResolveLink consumes the first element of the path and obtains the link
// corresponding to it from the node. It returns the link
// and the path without the consumed element.
func (n *ProtoNode) ResolveLink(path []string) (*format.Link, []string, error) {
	if len(path) == 0 {
		return nil, nil, fmt.Errorf("end of path, no more links to resolve")
	}

	lnk, err := n.GetNodeLink(path[0])
	if err != nil {
		return nil, nil, err
	}

	return lnk, path[1:], nil
}

// Tree returns the link names of the ProtoNode.
// ProtoNodes are only ever one path deep, so anything different than an empty
// string for p results in nothing. The depth parameter is ignored.
func (n *ProtoNode) Tree(p string, depth int) []string {
	if p != "" {
		return nil
	}

	if n.linksDirty {
		// there was a mutation involving links, make sure we sort
		sort.Stable(LinkSlice(n.links))
		n.linksDirty = false
		n.encoded = nil
	}

	out := make([]string, 0, len(n.links))
	for _, lnk := range n.links {
		out = append(out, lnk.Name)
	}
	return out
}

func ProtoNodeConverter(b blocks.Block, nd ipld.Node) (legacy.UniversalNode, error) {
	pbNode, ok := nd.(dagpb.PBNode)
	if !ok {
		return nil, ErrNotProtobuf
	}
	encoded := &immutableProtoNode{b.RawData(), pbNode}
	pn := fromImmutableNode(encoded)
	pn.cached = b.Cid()
	pn.builder = b.Cid().Prefix()
	return pn, nil
}

// TODO: replace with cid.MustParse() when we bump go-cid
func mustZeroCid() cid.Cid {
	c, err := cid.Parse("bafybeihdwdcefgh4dqkjv67uzcmw7ojee6xedzdetojuzjevtenxquvyku")
	if err != nil {
		panic(err)
	}
	return c
}

var _ legacy.UniversalNode = &ProtoNode{}