implement CidFromReader

And reuse a CidFromBytes test for it, which includes both CIDv0 and
CIDv1 cases as inputs.

Fixes #126.

.github/workflows/automerge.yml

@@ -33,9 +33,7 @@ jobs:
   automerge:
     needs: automerge-check
     runs-on: ubuntu-latest
-    # The check for the user is redundant here, as this job depends on the automerge-check job,
-    # but it prevents this job from spinning up, just to be skipped shortly after.
-    if: github.event.pull_request.user.login == 'web3-bot' && needs.automerge-check.outputs.status == 'true'
+    if: ${{ needs.automerge-check.outputs.status == 'true' }}
     steps:
     - name: Wait on tests
       uses: lewagon/wait-on-check-action@bafe56a6863672c681c3cf671f5e10b20abf2eaa # v0.2

.github/workflows/go-check.yml

@@ -8,28 +8,17 @@ jobs:
   unit:
     runs-on: ubuntu-latest
     name: All
-    env:
-      RUNGOGENERATE: false
     steps:
       - uses: actions/checkout@v2
         with:
           submodules: recursive
       - uses: actions/setup-go@v2
         with:
-          go-version: "1.17.x"
-      - name: Run repo-specific setup
-        uses: ./.github/actions/go-check-setup
-        if: hashFiles('./.github/actions/go-check-setup') != ''
-      - name: Read config
-        if: hashFiles('./.github/workflows/go-check-config.json') != ''
-        run: |
-          if jq -re .gogenerate ./.github/workflows/go-check-config.json; then
-            echo "RUNGOGENERATE=true" >> $GITHUB_ENV
-          fi
+          go-version: "1.16.x"
       - name: Install staticcheck
-        run: go install honnef.co/go/tools/cmd/staticcheck@df71e5d0e0ed317ebf43e6e59cf919430fa4b8f2 # 2021.1.1 (v0.2.1)
+        run: go install honnef.co/go/tools/cmd/staticcheck@434f5f3816b358fe468fa83dcba62d794e7fe04b # 2021.1 (v0.2.0)
       - name: Check that go.mod is tidy
-        uses: protocol/multiple-go-modules@v1.2
+        uses: protocol/multiple-go-modules@v1.0
         with:
           run: |
             go mod tidy
@@ -48,27 +37,14 @@ jobs:
           fi
       - name: go vet
         if: ${{ success() || failure() }} # run this step even if the previous one failed
-        uses: protocol/multiple-go-modules@v1.2
+        uses: protocol/multiple-go-modules@v1.0
         with:
           run: go vet ./...
       - name: staticcheck
         if: ${{ success() || failure() }} # run this step even if the previous one failed
-        uses: protocol/multiple-go-modules@v1.2
+        uses: protocol/multiple-go-modules@v1.0
         with:
           run: |
             set -o pipefail
             staticcheck ./... | sed -e 's@\(.*\)\.go@./\1.go@g'
-      - name: go generate
-        uses: protocol/multiple-go-modules@v1.2
-        if: (success() || failure()) && env.RUNGOGENERATE == 'true'
-        with:
-          run: |
-            git clean -fd # make sure there aren't untracked files / directories
-            go generate ./...
-            # check if go generate modified or added any files
-            if ! $(git add . && git diff-index HEAD --exit-code --quiet); then
-              echo "go generated caused changes to the repository:"
-              git status --short
-              exit 1
-            fi
 

.github/workflows/go-test.yml

@@ -10,9 +10,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ "ubuntu", "windows", "macos" ]
-        go: [ "1.16.x", "1.17.x" ]
-    env:
-      COVERAGES: ""
+        go: [ "1.15.x", "1.16.x" ]
     runs-on: ${{ matrix.os }}-latest
     name: ${{ matrix.os}} (go ${{ matrix.go }})
     steps:
@@ -26,30 +24,24 @@ jobs:
         run: |
           go version
           go env
-      - name: Run repo-specific setup
-        uses: ./.github/actions/go-test-setup
-        if: hashFiles('./.github/actions/go-test-setup') != ''
       - name: Run tests
-        uses: protocol/multiple-go-modules@v1.2
+        uses: protocol/multiple-go-modules@v1.0
         with:
-          run: go test -v -coverprofile module-coverage.txt ./...
+          run: go test -v -coverprofile coverage.txt ./...
       - name: Run tests (32 bit)
         if: ${{ matrix.os != 'macos' }} # can't run 32 bit tests on OSX.
-        uses: protocol/multiple-go-modules@v1.2
+        uses: protocol/multiple-go-modules@v1.0
         env:
           GOARCH: 386
         with:
           run: go test -v ./...
       - name: Run tests with race detector
         if: ${{ matrix.os == 'ubuntu' }} # speed things up. Windows and OSX VMs are slow
-        uses: protocol/multiple-go-modules@v1.2
+        uses: protocol/multiple-go-modules@v1.0
         with:
           run: go test -v -race ./...
-      - name: Collect coverage files
-        shell: bash
-        run: echo "COVERAGES=$(find . -type f -name 'module-coverage.txt' | tr -s '\n' ',' | sed 's/,$//')" >> $GITHUB_ENV
       - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@51d810878be5422784e86451c0e7c14e5860ec47 # v2.0.2
+        uses: codecov/codecov-action@a1ed4b322b4b38cb846afb5a0ebfa17086917d27 # v1.5.0
         with:
-          files: '${{ env.COVERAGES }}'
+          file: coverage.txt
           env_vars: OS=${{ matrix.os }}, GO=${{ matrix.go }}

cid.go

@@ -22,7 +22,6 @@ package cid
 import (
 	"bytes"
 	"encoding"
-	"encoding/binary"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -174,24 +173,16 @@ func NewCidV0(mhash mh.Multihash) Cid {
 // Panics if the multihash is invalid.
 func NewCidV1(codecType uint64, mhash mh.Multihash) Cid {
 	hashlen := len(mhash)
-
-	// Two 8 bytes (max) numbers plus hash.
-	// We use strings.Builder to only allocate once.
-	var b strings.Builder
-	b.Grow(1 + varint.UvarintSize(codecType) + hashlen)
-
-	b.WriteByte(1)
-
-	var buf [binary.MaxVarintLen64]byte
-	n := varint.PutUvarint(buf[:], codecType)
-	b.Write(buf[:n])
-
-	cn, _ := b.Write(mhash)
+	// two 8 bytes (max) numbers plus hash
+	buf := make([]byte, 1+varint.UvarintSize(codecType)+hashlen)
+	n := varint.PutUvarint(buf, 1)
+	n += varint.PutUvarint(buf[n:], codecType)
+	cn := copy(buf[n:], mhash)
 	if cn != hashlen {
 		panic("copy hash length is inconsistent")
 	}
 
-	return Cid{b.String()}
+	return Cid{string(buf[:n+hashlen])}
 }
 
 var (
@@ -703,10 +694,19 @@ type bufByteReader struct {
 	direct   io.ByteReader
 	fallback io.Reader
 
-	dst []byte
+	consumed int
+	dst      []byte
 }
 
 func (r *bufByteReader) ReadByte() (byte, error) {
+	// We still have some of the initial bytes to use.
+	if r.consumed < len(r.dst) {
+		b := r.dst[r.consumed]
+		r.consumed++
+		return b, nil
+	}
+	r.consumed++
+
 	// The underlying reader has ReadByte; use it.
 	if br := r.direct; br != nil {
 		b, err := br.ReadByte()
@@ -718,8 +718,6 @@ func (r *bufByteReader) ReadByte() (byte, error) {
 	}
 
 	// Fall back to a one-byte Read.
-	// TODO: consider reading straight into dst,
-	// once we have benchmarks and if they prove that to be faster.
 	var p [1]byte
 	if _, err := io.ReadFull(r.fallback, p[:]); err != nil {
 		return 0, err
@@ -739,30 +737,38 @@ func CidFromReader(r io.Reader) (int, Cid, error) {
 	// 64 bytes is enough for any CIDv0,
 	// and it's enough for most CIDv1s in practice.
 	// If the digest is too long, we'll allocate more.
-	br := toBufByteReader(r, make([]byte, 0, 64))
+	buf := make([]byte, 0, 64)
+
+	// We read two bytes, to tell if this is a CIDv0 or a CIDv1.
+	if n, err := io.ReadFull(r, buf[:2]); err != nil {
+		return n, Undef, err
+	}
+	buf = buf[:2]
+
+	// If we have a CIDv0, read the rest of the bytes and cast the buffer.
+	if buf[0] == mh.SHA2_256 && buf[1] == 32 {
+		if n, err := io.ReadFull(r, buf[2:34]); err != nil {
+			return len(buf) + n, Undef, err
+		}
+
+		buf = buf[:34]
+		h, err := mh.Cast(buf)
+		if err != nil {
+			return len(buf), Undef, err
+		}
+
+		return len(buf), Cid{string(h)}, nil
+	}
 
-	// We read the first varint, to tell if this is a CIDv0 or a CIDv1.
 	// The varint package wants a io.ByteReader, so we must wrap our io.Reader.
+	// Note that we already read two bytes, so bufByteReader uses those first.
+	// After those two bytes, bufByteReader appends the read bytes to br.dst.
+	br := toBufByteReader(r, buf[:2])
 	vers, err := varint.ReadUvarint(br)
 	if err != nil {
 		return len(br.dst), Undef, err
 	}
 
-	// If we have a CIDv0, read the rest of the bytes and cast the buffer.
-	if vers == mh.SHA2_256 {
-		if n, err := io.ReadFull(r, br.dst[1:34]); err != nil {
-			return len(br.dst) + n, Undef, err
-		}
-
-		br.dst = br.dst[:34]
-		h, err := mh.Cast(br.dst)
-		if err != nil {
-			return len(br.dst), Undef, err
-		}
-
-		return len(br.dst), Cid{string(h)}, nil
-	}
-
 	if vers != 1 {
 		return len(br.dst), Undef, fmt.Errorf("expected 1 as the cid version number, got: %d", vers)
 	}
@@ -790,38 +796,29 @@ func CidFromReader(r io.Reader) (int, Cid, error) {
 		return len(br.dst), Undef, err
 	}
 
+	// Update buf's length.
+	// We're not reading single bytes beyond this point.
+	buf = br.dst
+	br = nil
+
+	// Multihash digest; might be too long, so allocate.
 	// Refuse to make large allocations to prevent OOMs due to bugs.
+	// TODO: reuse buf if it has enough space
 	const maxDigestAlloc = 32 << 20 // 32MiB
 	if mhl > maxDigestAlloc {
-		return len(br.dst), Undef, fmt.Errorf("refusing to allocate %d bytes for a digest", mhl)
+		return len(buf), Undef, fmt.Errorf("refusing to allocate %d bytes for a digest", mhl)
 	}
-
-	// Fine to convert mhl to int, given maxDigestAlloc.
-	prefixLength := len(br.dst)
-	cidLength := prefixLength + int(mhl)
-	if cidLength > cap(br.dst) {
-		// If the multihash digest doesn't fit in our initial 64 bytes,
-		// efficiently extend the slice via append+make.
-		br.dst = append(br.dst, make([]byte, cidLength-len(br.dst))...)
-	} else {
-		// The multihash digest fits inside our buffer,
-		// so just extend its capacity.
-		br.dst = br.dst[:cidLength]
-	}
-
-	if n, err := io.ReadFull(r, br.dst[prefixLength:cidLength]); err != nil {
-		// We can't use len(br.dst) here,
-		// as we've only read n bytes past prefixLength.
-		return prefixLength + n, Undef, err
+	digest := make([]byte, int(mhl))
+	if n, err := io.ReadFull(r, digest); err != nil {
+		return len(buf) + n, Undef, err
 	}
+	buf = append(buf, digest...)
 
 	// This simply ensures the multihash is valid.
-	// TODO: consider removing this bit, as it's probably redundant;
-	// for now, it helps ensure consistency with CidFromBytes.
-	_, _, err = mh.MHFromBytes(br.dst[mhStart:])
+	_, _, err = mh.MHFromBytes(buf[mhStart:])
 	if err != nil {
-		return len(br.dst), Undef, err
+		return len(buf), Undef, err
 	}
 
-	return len(br.dst), Cid{string(br.dst)}, nil
+	return len(buf), Cid{string(buf)}, nil
 }

cid_fuzz.go

@@ -1,4 +1,3 @@
-//go:build gofuzz
 // +build gofuzz
 
 package cid

cid_test.go

@@ -667,7 +667,6 @@ func TestReadCidsFromBuffer(t *testing.T) {
 		"k2cwueckqkibutvhkr4p2ln2pjcaxaakpd9db0e7j7ax1lxhhxy3ekpv",
 		"Qmf5Qzp6nGBku7CEn2UQx4mgN8TW69YUok36DrGa6NN893",
 		"zb2rhZi1JR4eNc2jBGaRYJKYM8JEB4ovenym8L1CmFsRAytkz",
-		"bafkqarjpmzuwyzltorxxezjpkvcfgqkfjfbfcvslivje2vchkzdu6rckjjcfgtkolaze6mssjqzeyn2ekrcfatkjku2vowseky3fswkfkm2deqkrju3e2",
 	}
 
 	var cids []Cid
@@ -722,71 +721,49 @@ func TestReadCidsFromBuffer(t *testing.T) {
 	}
 }
 
-func TestBadCidInput(t *testing.T) {
-	for _, name := range []string{
-		"FromBytes",
-		"FromReader",
-	} {
-		t.Run(name, func(t *testing.T) {
-			usingReader := name == "FromReader"
+func TestBadCidFromBytes(t *testing.T) {
+	l, c, err := CidFromBytes([]byte{mh.SHA2_256, 32, 0x00})
+	if err == nil {
+		t.Fatal("expected not-enough-bytes for V0 CidFromBytes")
+	}
+	if l != 0 {
+		t.Fatal("expected length=0 from bad CidFromBytes")
+	}
+	if c != Undef {
+		t.Fatal("expected Undef CID from bad CidFromBytes")
+	}
 
-			fromBytes := CidFromBytes
-			if usingReader {
-				fromBytes = func(data []byte) (int, Cid, error) {
-					return CidFromReader(bytes.NewReader(data))
-				}
-			}
+	c, err = Decode("bafkreie5qrjvaw64n4tjm6hbnm7fnqvcssfed4whsjqxzslbd3jwhsk3mm")
+	if err != nil {
+		t.Fatal(err)
+	}
+	byts := make([]byte, c.ByteLen())
+	copy(byts, c.Bytes())
+	byts[1] = 0x80 // bad codec varint
+	byts[2] = 0x00
+	l, c, err = CidFromBytes(byts)
+	if err == nil {
+		t.Fatal("expected not-enough-bytes for V1 CidFromBytes")
+	}
+	if l != 0 {
+		t.Fatal("expected length=0 from bad CidFromBytes")
+	}
+	if c != Undef {
+		t.Fatal("expected Undef CID from bad CidFromBytes")
+	}
 
-			l, c, err := fromBytes([]byte{mh.SHA2_256, 32, 0x00})
-			if err == nil {
-				t.Fatal("expected not-enough-bytes for V0 CID")
-			}
-			if !usingReader && l != 0 {
-				t.Fatal("expected length==0 from bad CID")
-			} else if usingReader && l == 0 {
-				t.Fatal("expected length!=0 from bad CID")
-			}
-			if c != Undef {
-				t.Fatal("expected Undef CID from bad CID")
-			}
-
-			c, err = Decode("bafkreie5qrjvaw64n4tjm6hbnm7fnqvcssfed4whsjqxzslbd3jwhsk3mm")
-			if err != nil {
-				t.Fatal(err)
-			}
-			byts := make([]byte, c.ByteLen())
-			copy(byts, c.Bytes())
-			byts[1] = 0x80 // bad codec varint
-			byts[2] = 0x00
-			l, c, err = fromBytes(byts)
-			if err == nil {
-				t.Fatal("expected not-enough-bytes for V1 CID")
-			}
-			if !usingReader && l != 0 {
-				t.Fatal("expected length==0 from bad CID")
-			} else if usingReader && l == 0 {
-				t.Fatal("expected length!=0 from bad CID")
-			}
-			if c != Undef {
-				t.Fatal("expected Undef CID from bad CID")
-			}
-
-			copy(byts, c.Bytes())
-			byts[2] = 0x80 // bad multihash varint
-			byts[3] = 0x00
-			l, c, err = fromBytes(byts)
-			if err == nil {
-				t.Fatal("expected not-enough-bytes for V1 CID")
-			}
-			if !usingReader && l != 0 {
-				t.Fatal("expected length==0 from bad CID")
-			} else if usingReader && l == 0 {
-				t.Fatal("expected length!=0 from bad CID")
-			}
-			if c != Undef {
-				t.Fatal("expected Undef CID from bad CidFromBytes")
-			}
-		})
+	copy(byts, c.Bytes())
+	byts[2] = 0x80 // bad multihash varint
+	byts[3] = 0x00
+	l, c, err = CidFromBytes(byts)
+	if err == nil {
+		t.Fatal("expected not-enough-bytes for V1 CidFromBytes")
+	}
+	if l != 0 {
+		t.Fatal("expected length=0 from bad CidFromBytes")
+	}
+	if c != Undef {
+		t.Fatal("expected Undef CID from bad CidFromBytes")
 	}
 }
 

go.mod

@@ -7,4 +7,4 @@ require (
 	golang.org/x/crypto v0.0.0-20210506145944-38f3c27a63bf // indirect
 )
 
-go 1.16
+go 1.15