Introduced sha256 support for git-sizer

git-sizer.go

@@ -134,7 +134,7 @@ func mainImplementation(ctx context.Context, stdout, stderr io.Writer, args []st
 
 	// Try to open the repository, but it's not an error yet if this
 	// fails, because the user might only be asking for `--help`.
-	repo, repoErr := git.NewRepository(".")
+	repo, repoErr := git.NewRepositoryFromPath(".")
 
 	flags := pflag.NewFlagSet("git-sizer", pflag.ContinueOnError)
 	flags.Usage = func() {

git/git.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"errors"
 	"fmt"
+	"io/fs"
 	"os"
 	"os/exec"
 	"path/filepath"
@@ -15,26 +16,32 @@ type ObjectType string
 
 // Repository represents a Git repository on disk.
 type Repository struct {
-	path string
+	// gitDir is the path to the `GIT_DIR` for this repository. It
+	// might be absolute or it might be relative to the current
+	// directory.
+	gitDir string
 
 	// gitBin is the path of the `git` executable that should be used
 	// when running commands in this repository.
-	gitBin string
+	gitBin   string
+	hashAlgo HashAlgo
 }
 
-// smartJoin returns the path that can be described as `relPath`
-// relative to `path`, given that `path` is either absolute or is
-// relative to the current directory.
+// smartJoin returns `relPath` if it is an absolute path. If not, it
+// assumes that `relPath` is relative to `path`, so it joins them
+// together and returns the result. In that case, if `path` itself is
+// relative, then the return value is also relative.
 func smartJoin(path, relPath string) string {
 	if filepath.IsAbs(relPath) {
 		return relPath
 	}
 	return filepath.Join(path, relPath)
 }
 
-// NewRepository creates a new repository object that can be used for
-// running `git` commands within that repository.
-func NewRepository(path string) (*Repository, error) {
+// NewRepositoryFromGitDir creates a new `Repository` object that can
+// be used for running `git` commands, given the value of `GIT_DIR`
+// for the repository.
+func NewRepositoryFromGitDir(gitDir string) (*Repository, error) {
 	// Find the `git` executable to be used:
 	gitBin, err := findGitBin()
 	if err != nil {
@@ -43,6 +50,44 @@ func NewRepository(path string) (*Repository, error) {
 		)
 	}
 
+	hashAlgo := HashSHA1
+	//nolint:gosec // `gitDir` is the path we need Git to access.
+	cmd := exec.Command(gitBin, "--git-dir", gitDir, "rev-parse", "--show-object-format")
+	if out, err := cmd.Output(); err == nil {
+		if string(bytes.TrimSpace(out)) == "sha256" {
+			hashAlgo = HashSHA256
+		}
+	}
+
+	repo := Repository{
+		gitDir:   gitDir,
+		gitBin:   gitBin,
+		hashAlgo: hashAlgo,
+	}
+
+	full, err := repo.IsFull()
+	if err != nil {
+		return nil, fmt.Errorf("determining whether the repository is a full clone: %w", err)
+	}
+	if !full {
+		return nil, errors.New("this appears to be a shallow clone; full clone required")
+	}
+
+	return &repo, nil
+}
+
+// NewRepositoryFromPath creates a new `Repository` object that can be
+// used for running `git` commands within `path`. It does so by asking
+// `git` what `GIT_DIR` to use. Git, in turn, bases its decision on
+// the path and the environment.
+func NewRepositoryFromPath(path string) (*Repository, error) {
+	gitBin, err := findGitBin()
+	if err != nil {
+		return nil, fmt.Errorf(
+			"could not find 'git' executable (is it in your PATH?): %w", err,
+		)
+	}
+
 	//nolint:gosec // `gitBin` is chosen carefully, and `path` is the
 	// path to the repository.
 	cmd := exec.Command(gitBin, "-C", path, "rev-parse", "--git-dir")
@@ -63,25 +108,28 @@ func NewRepository(path string) (*Repository, error) {
 	}
 	gitDir := smartJoin(path, string(bytes.TrimSpace(out)))
 
-	//nolint:gosec // `gitBin` is chosen carefully.
-	cmd = exec.Command(gitBin, "rev-parse", "--git-path", "shallow")
-	cmd.Dir = gitDir
-	out, err = cmd.Output()
+	return NewRepositoryFromGitDir(gitDir)
+}
+
+// IsFull returns `true` iff `repo` appears to be a full clone.
+func (repo *Repository) IsFull() (bool, error) {
+	shallow, err := repo.GitPath("shallow")
 	if err != nil {
-		return nil, fmt.Errorf(
-			"could not run 'git rev-parse --git-path shallow': %w", err,
-		)
+		return false, err
 	}
-	shallow := smartJoin(gitDir, string(bytes.TrimSpace(out)))
+
 	_, err = os.Lstat(shallow)
 	if err == nil {
-		return nil, errors.New("this appears to be a shallow clone; full clone required")
+		return false, nil
+	}
+
+	if !errors.Is(err, fs.ErrNotExist) {
+		return false, err
 	}
 
-	return &Repository{
-		path:   gitDir,
-		gitBin: gitBin,
-	}, nil
+	// The `shallow` file is absent, which is what we expect
+	// for a full clone.
+	return true, nil
 }
 
 func (repo *Repository) GitCommand(callerArgs ...string) *exec.Cmd {
@@ -103,15 +151,41 @@ func (repo *Repository) GitCommand(callerArgs ...string) *exec.Cmd {
 
 	cmd.Env = append(
 		os.Environ(),
-		"GIT_DIR="+repo.path,
+		"GIT_DIR="+repo.gitDir,
 		// Disable grafts when running our commands:
 		"GIT_GRAFT_FILE="+os.DevNull,
 	)
 
 	return cmd
 }
 
-// Path returns the path to `repo`.
-func (repo *Repository) Path() string {
-	return repo.path
+// GitDir returns the path to `repo`'s `GIT_DIR`. It might be absolute
+// or it might be relative to the current directory.
+func (repo *Repository) GitDir() string {
+	return repo.gitDir
+}
+
+// GitPath returns that path of a file within the git repository, by
+// calling `git rev-parse --git-path $relPath`. The returned path is
+// relative to the current directory.
+func (repo *Repository) GitPath(relPath string) (string, error) {
+	cmd := repo.GitCommand("rev-parse", "--git-path", relPath)
+	out, err := cmd.Output()
+	if err != nil {
+		return "", fmt.Errorf(
+			"running 'git rev-parse --git-path %s': %w", relPath, err,
+		)
+	}
+	// `git rev-parse --git-path` is documented to return the path
+	// relative to the current directory. Since we haven't changed the
+	// current directory, we can use it as-is:
+	return string(bytes.TrimSpace(out)), nil
+}
+
+func (repo *Repository) HashAlgo() HashAlgo {
+	return repo.hashAlgo
+}
+
+func (repo *Repository) HashSize() int {
+	return repo.hashAlgo.HashSize()
 }

git/git_bin.go

@@ -2,26 +2,41 @@ package git
 
 import (
 	"path/filepath"
+	"sync"
 
 	"github.com/cli/safeexec"
 )
 
+// This variable will be used to memoize the result of `findGitBin()`,
+// since its return value only depends on the environment.
+var gitBinMemo struct {
+	once sync.Once
+
+	gitBin string
+	err    error
+}
+
 // findGitBin finds the `git` binary in PATH that should be used by
 // the rest of `git-sizer`. It uses `safeexec` to find the executable,
 // because on Windows, `exec.Cmd` looks not only in PATH, but also in
 // the current directory. This is a potential risk if the repository
 // being scanned is hostile and non-bare because it might possibly
 // contain an executable file named `git`.
 func findGitBin() (string, error) {
-	gitBin, err := safeexec.LookPath("git")
-	if err != nil {
-		return "", err
-	}
+	gitBinMemo.once.Do(func() {
+		p, err := safeexec.LookPath("git")
+		if err != nil {
+			gitBinMemo.err = err
+			return
+		}
 
-	gitBin, err = filepath.Abs(gitBin)
-	if err != nil {
-		return "", err
-	}
+		p, err = filepath.Abs(p)
+		if err != nil {
+			gitBinMemo.err = err
+			return
+		}
 
-	return gitBin, nil
+		gitBinMemo.gitBin = p
+	})
+	return gitBinMemo.gitBin, gitBinMemo.err
 }

git/gitconfig.go

@@ -90,10 +90,10 @@ func (config *Config) FullKey(key string) string {
 // a component boundary (i.e., at a '.'). If yes, it returns `true`
 // and the part of the key after the prefix; e.g.:
 //
-//     configKeyMatchesPrefix("foo.bar", "foo") → true, "bar"
-//     configKeyMatchesPrefix("foo.bar", "foo.") → true, "bar"
-//     configKeyMatchesPrefix("foo.bar", "foo.bar") → true, ""
-//     configKeyMatchesPrefix("foo.bar", "foo.bar.") → false, ""
+//	configKeyMatchesPrefix("foo.bar", "foo") → true, "bar"
+//	configKeyMatchesPrefix("foo.bar", "foo.") → true, "bar"
+//	configKeyMatchesPrefix("foo.bar", "foo.bar") → true, ""
+//	configKeyMatchesPrefix("foo.bar", "foo.bar.") → false, ""
 func configKeyMatchesPrefix(key, prefix string) (bool, string) {
 	if prefix == "" {
 		return true, key

git/obj_iter.go

@@ -30,7 +30,7 @@ func (repo *Repository) NewObjectIter(ctx context.Context) (*ObjectIter, error)
 		errCh:    make(chan error),
 		headerCh: make(chan BatchHeader),
 	}
-
+	hashHexSize := repo.HashSize() * 2
 	iter.p.Add(
 		// Read OIDs from `iter.oidCh` and write them to `git
 		// rev-list`:
@@ -68,10 +68,10 @@ func (repo *Repository) NewObjectIter(ctx context.Context) (*ObjectIter, error)
 		pipe.LinewiseFunction(
 			"copy-oids",
 			func(_ context.Context, _ pipe.Env, line []byte, stdout *bufio.Writer) error {
-				if len(line) < 40 {
+				if len(line) < hashHexSize {
 					return fmt.Errorf("line too short: '%s'", line)
 				}
-				if _, err := stdout.Write(line[:40]); err != nil {
+				if _, err := stdout.Write(line[:hashHexSize]); err != nil {
 					return fmt.Errorf("writing OID to 'git cat-file': %w", err)
 				}
 				if err := stdout.WriteByte('\n'); err != nil {

git/obj_resolver.go

@@ -9,12 +9,12 @@ func (repo *Repository) ResolveObject(name string) (OID, error) {
 	cmd := repo.GitCommand("rev-parse", "--verify", "--end-of-options", name)
 	output, err := cmd.Output()
 	if err != nil {
-		return NullOID, fmt.Errorf("resolving object %q: %w", name, err)
+		return repo.HashAlgo().NullOID(), fmt.Errorf("resolving object %q: %w", name, err)
 	}
 	oidString := string(bytes.TrimSpace(output))
 	oid, err := NewOID(oidString)
 	if err != nil {
-		return NullOID, fmt.Errorf("parsing output %q from 'rev-parse': %w", oidString, err)
+		return repo.HashAlgo().NullOID(), fmt.Errorf("parsing output %q from 'rev-parse': %w", oidString, err)
 	}
 	return oid, nil
 }

git/oid.go

@@ -1,31 +1,76 @@
 package git
 
 import (
+	"bytes"
+	//nolint:gosec // Git indeed does use SHA-1, still
+	"crypto/sha1"
+	"crypto/sha256"
 	"encoding/hex"
 	"errors"
 )
 
+const (
+	HashSizeSHA256 = sha256.Size
+	HashSizeSHA1   = sha1.Size
+	HashSizeMax    = HashSizeSHA256
+)
+
+type HashAlgo int
+
+const (
+	HashUnknown HashAlgo = iota
+	HashSHA1
+	HashSHA256
+)
+
 // OID represents the SHA-1 object ID of a Git object, in binary
 // format.
 type OID struct {
-	v [20]byte
+	v        [HashSizeMax]byte
+	hashSize int
 }
 
-// NullOID is the null object ID; i.e., all zeros.
-var NullOID OID
+func (h HashAlgo) NullOID() OID {
+	switch h {
+	case HashSHA1:
+		return OID{hashSize: HashSizeSHA1}
+	case HashSHA256:
+		return OID{hashSize: HashSizeSHA256}
+	}
+	return OID{}
+}
+
+func (h HashAlgo) HashSize() int {
+	switch h {
+	case HashSHA1:
+		return HashSizeSHA1
+	case HashSHA256:
+		return HashSizeSHA256
+	}
+	return 0
+}
+
+// defaultNullOID is the null object ID; i.e., all zeros.
+var defaultNullOID OID
+
+func IsNullOID(o OID) bool {
+	return bytes.Equal(o.v[:], defaultNullOID.v[:])
+}
 
 // OIDFromBytes converts a byte slice containing an object ID in
 // binary format into an `OID`.
 func OIDFromBytes(oidBytes []byte) (OID, error) {
 	var oid OID
-	if len(oidBytes) != len(oid.v) {
+	oidSize := len(oidBytes)
+	if oidSize != HashSizeSHA1 && oidSize != HashSizeSHA256 {
 		return OID{}, errors.New("bytes oid has the wrong length")
 	}
-	copy(oid.v[0:20], oidBytes)
+	oid.hashSize = oidSize
+	copy(oid.v[0:oidSize], oidBytes)
 	return oid, nil
 }
 
-// NewOID converts an object ID in hex format (i.e., `[0-9a-f]{40}`)
+// NewOID converts an object ID in hex format (i.e., `[0-9a-f]{40,64}`)
 // into an `OID`.
 func NewOID(s string) (OID, error) {
 	oidBytes, err := hex.DecodeString(s)
@@ -37,18 +82,18 @@ func NewOID(s string) (OID, error) {
 
 // String formats `oid` as a string in hex format.
 func (oid OID) String() string {
-	return hex.EncodeToString(oid.v[:])
+	return hex.EncodeToString(oid.v[:oid.hashSize])
 }
 
 // Bytes returns a byte slice view of `oid`, in binary format.
 func (oid OID) Bytes() []byte {
-	return oid.v[:]
+	return oid.v[:oid.hashSize]
 }
 
 // MarshalJSON expresses `oid` as a JSON string with its enclosing
 // quotation marks.
 func (oid OID) MarshalJSON() ([]byte, error) {
-	src := oid.v[:]
+	src := oid.v[:oid.hashSize]
 	dst := make([]byte, hex.EncodedLen(len(src))+2)
 	dst[0] = '"'
 	dst[len(dst)-1] = '"'