hakurei/internal/pkg/pkg.go

// Package pkg provides utilities for packaging software.
package pkg

import (
	"bytes"
	"crypto/sha512"
	"encoding/base64"
	"encoding/binary"
	"errors"
	"io"
	"io/fs"
	"os"
	"path"
	"path/filepath"
	"slices"
	"sync"

	"hakurei.app/container/check"
)

type (
	// A Checksum is a SHA-384 checksum computed for a cured [Artifact].
	Checksum = [sha512.Size384]byte

	// An ID is a unique identifier returned by [Artifact.ID]. This value must
	// be deterministically determined ahead of time.
	ID Checksum
)

// Encode is abbreviation for base64.URLEncoding.EncodeToString(checksum[:]).
func Encode(checksum Checksum) string {
	return base64.URLEncoding.EncodeToString(checksum[:])
}

// Decode is abbreviation for base64.URLEncoding.Decode(checksum[:], []byte(s)).
func Decode(s string) (checksum Checksum, err error) {
	var n int
	n, err = base64.URLEncoding.Decode(checksum[:], []byte(s))
	if err == nil && n != len(Checksum{}) {
		err = io.ErrUnexpectedEOF
	}
	return
}

// MustDecode decodes a string representation of [Checksum] and panics if there
// is a decoding error or the resulting data is too short.
func MustDecode(s string) Checksum {
	if checksum, err := Decode(s); err != nil {
		panic(err)
	} else {
		return checksum
	}
}

// CacheDataFunc tries to load [File] from [Cache], and if that fails, obtains
// it via [File.Data] instead.
type CacheDataFunc func(f File) (data []byte, err error)

// An Artifact is a read-only reference to a piece of data that may be created
// deterministically but might not currently be available in memory or on the
// filesystem.
type Artifact interface {
	// Kind returns the [Kind] of artifact. This is usually unique to the
	// concrete type but two functionally identical implementations of
	// [Artifact] is allowed to return the same [Kind] value.
	Kind() Kind

	// Params returns opaque bytes that describes [Artifact]. Implementations
	// must guarantee that these values are unique among differing instances
	// of the same implementation with the same dependencies.
	//
	// Callers must not modify the retuned byte slice.
	//
	// Result must remain identical across multiple invocations.
	Params() []byte

	// Dependencies returns a slice of [Artifact] that the current instance
	// depends on to produce its contents.
	//
	// Callers must not modify the retuned slice.
	//
	// Result must remain identical across multiple invocations.
	Dependencies() []Artifact

	// Cure cures the current [Artifact] to the caller-specified temporary
	// pathname. This is not the final resting place of the [Artifact] and this
	// pathname should not be directly referred to in the final contents.
	//
	// If the implementation produces a single file, it must implement [File]
	// as well. In that case, Cure must produce a single regular file with
	// contents identical to that returned by [File.Data].
	Cure(work *check.Absolute, loadData CacheDataFunc) (err error)
}

// KnownIdent is optionally implemented by [Artifact] and is used instead of
// [Kind.Ident] when it is available.
//
// This is very subtle to use correctly. The implementation must ensure that
// this value is globally unique, otherwise [Cache] can enter an inconsistent
// state. This should not be implemented outside of testing.
type KnownIdent interface {
	// ID returns a globally unique identifier referring to the current
	// [Artifact]. This value must be known ahead of time and guaranteed to be
	// unique without having obtained the full contents of the [Artifact].
	ID() ID
}

// KnownChecksum is optionally implemented by [Artifact] for an artifact with
// output known ahead of time.
type KnownChecksum interface {
	// Checksum returns the address of a known checksum.
	//
	// Callers must not modify the [Checksum].
	//
	// Result must remain identical across multiple invocations.
	Checksum() Checksum
}

// A File refers to an [Artifact] backed by a single file.
type File interface {
	// Data returns the full contents of [Artifact]. If [Artifact.Checksum]
	// returns a non-nil address, Data is responsible for validating any data
	// it produces and must return [ChecksumMismatchError] if validation fails.
	//
	// Callers must not modify the returned byte slice.
	Data() ([]byte, error)

	Artifact
}

// Ident returns the identifier of an [Artifact].
func Ident(a Artifact) ID {
	if ki, ok := a.(KnownIdent); ok {
		return ki.ID()
	}
	return a.Kind().Ident(a.Params(), a.Dependencies()...)
}

// Kind corresponds to the concrete type of [Artifact] and is used to create
// identifier for an [Artifact] with dependencies.
type Kind uint64

const (
	// KindHTTPGet is the kind of [Artifact] returned by [NewHTTPGet].
	KindHTTPGet Kind = iota
	// KindTar is the kind of artifact returned by [NewTar].
	KindTar
)

// Ident returns a deterministic identifier for the supplied params and
// dependencies. The caller is responsible for ensuring params uniquely and
// deterministically describes the current [Artifact].
func (k Kind) Ident(params []byte, deps ...Artifact) ID {
	type extIdent [len(ID{}) + wordSize]byte
	identifiers := make([]extIdent, len(deps))
	for i, a := range deps {
		id := Ident(a)
		copy(identifiers[i][wordSize:], id[:])
		binary.LittleEndian.PutUint64(identifiers[i][:], uint64(a.Kind()))
	}
	slices.SortFunc(identifiers, func(a, b extIdent) int {
		return bytes.Compare(a[:], b[:])
	})
	slices.Compact(identifiers)

	h := sha512.New384()
	h.Write(binary.LittleEndian.AppendUint64(nil, uint64(k)))
	h.Write(params)
	for _, e := range identifiers {
		h.Write(e[:])
	}
	return ID(h.Sum(nil))
}

const (
	// dirIdentifier is the directory name appended to Cache.base for storing
	// artifacts named after their [ID].
	dirIdentifier = "identifier"
	// dirChecksum is the directory name appended to Cache.base for storing
	// artifacts named after their [Checksum].
	dirChecksum = "checksum"

	// dirWork is the directory name appended to Cache.base for working
	// pathnames set up during [Cache.Cure].
	dirWork = "work"

	// checksumLinknamePrefix is prepended to the encoded [Checksum] value
	// of an [Artifact] when creating a symbolic link to dirChecksum.
	checksumLinknamePrefix = "../" + dirChecksum + "/"
)

// Cache is a support layer that implementations of [Artifact] can use to store
// cured [Artifact] data in a content addressed fashion.
type Cache struct {
	// Directory where all [Cache] related files are placed.
	base *check.Absolute

	// Whether to validate [File.Data] for a [KnownChecksum] file. This
	// significantly reduces performance.
	strict bool

	// Synchronises access to dirChecksum.
	checksumMu sync.RWMutex

	// Identifier to content pair cache.
	ident map[ID]Checksum
	// Identifier to error pair for unrecoverably faulted [Artifact].
	identErr map[ID]error
	// Pending identifiers, accessed through Cure for entries not in ident.
	identPending map[ID]<-chan struct{}
	// Synchronises access to ident and corresponding filesystem entries.
	identMu sync.RWMutex
}

// IsStrict returns whether the [Cache] strictly verifies checksums.
func (c *Cache) IsStrict() bool { return c.strict }

// SetStrict sets whether the [Cache] strictly verifies checksums, even when
// the implementation promises to validate them internally. This significantly
// reduces performance and is not recommended outside of testing.
//
// This method is not safe for concurrent use with any other method.
func (c *Cache) SetStrict(strict bool) { c.strict = strict }

// A ChecksumMismatchError describes an [Artifact] with unexpected content.
type ChecksumMismatchError struct {
	// Actual and expected checksums.
	Got, Want Checksum
}

func (e *ChecksumMismatchError) Error() string {
	return "got " + Encode(e.Got) +
		" instead of " + Encode(e.Want)
}

// loadOrStoreIdent attempts to load a cached [Artifact] by its identifier or
// wait for a pending [Artifact] to cure. If neither is possible, the current
// identifier is stored in identPending and a non-nil channel is returned.
func (c *Cache) loadOrStoreIdent(id *ID) (
	done chan<- struct{},
	checksum Checksum,
	err error,
) {
	var ok bool

	c.identMu.Lock()
	if checksum, ok = c.ident[*id]; ok {
		c.identMu.Unlock()
		return
	}
	if err, ok = c.identErr[*id]; ok {
		c.identMu.Unlock()
		return
	}

	var notify <-chan struct{}
	if notify, ok = c.identPending[*id]; ok {
		c.identMu.Unlock()
		<-notify
		c.identMu.RLock()
		if checksum, ok = c.ident[*id]; !ok {
			err = c.identErr[*id]
		}
		c.identMu.RUnlock()
		return
	}

	d := make(chan struct{})
	c.identPending[*id] = d
	c.identMu.Unlock()
	done = d
	return
}

// finaliseIdent commits a checksum or error to ident for an identifier
// previously submitted to identPending.
func (c *Cache) finaliseIdent(
	done chan<- struct{},
	id *ID,
	checksum *Checksum,
	err error,
) {
	c.identMu.Lock()
	if err != nil {
		c.identErr[*id] = err
	} else {
		c.ident[*id] = *checksum
	}
	c.identMu.Unlock()

	close(done)
}

// loadData provides [CacheDataFunc] for [Artifact.Cure].
func (c *Cache) loadData(f File) (data []byte, err error) {
	var r *os.File
	if kc, ok := f.(KnownChecksum); ok {
		c.checksumMu.RLock()
		r, err = os.Open(c.base.Append(
			dirChecksum,
			Encode(kc.Checksum()),
		).String())
		c.checksumMu.RUnlock()
	} else {
		c.identMu.RLock()
		r, err = os.Open(c.base.Append(
			dirIdentifier,
			Encode(Ident(f)),
		).String())
		c.identMu.RUnlock()
	}

	if err != nil {
		if !errors.Is(err, os.ErrNotExist) {
			return
		}
		return f.Data()
	}

	data, err = io.ReadAll(r)
	closeErr := r.Close()
	if err == nil {
		err = closeErr
	}
	return
}

// InvalidFileModeError describes an [Artifact.Cure] that did not result in
// a regular file or directory located at the work pathname.
type InvalidFileModeError fs.FileMode

// Error returns a constant string.
func (e InvalidFileModeError) Error() string {
	return "artifact did not produce a regular file or directory"
}

// NoOutputError describes an [Artifact.Cure] that did not populate its
// work pathname despite completing successfully.
type NoOutputError struct{}

// Unwrap returns [os.ErrNotExist].
func (NoOutputError) Unwrap() error { return os.ErrNotExist }

// Error returns a constant string.
func (NoOutputError) Error() string {
	return "artifact cured successfully but did not produce any output"
}

// Cure cures the [Artifact] and returns its pathname and [Checksum].
func (c *Cache) Cure(a Artifact) (
	pathname *check.Absolute,
	checksum Checksum,
	err error,
) {
	id := Ident(a)
	ids := Encode(id)
	pathname = c.base.Append(
		dirIdentifier,
		ids,
	)
	defer func() {
		if err != nil {
			pathname = nil
			checksum = Checksum{}
		}
	}()

	var done chan<- struct{}
	done, checksum, err = c.loadOrStoreIdent(&id)
	if done == nil {
		return
	} else {
		defer func() { c.finaliseIdent(done, &id, &checksum, err) }()
	}

	_, err = os.Lstat(pathname.String())
	if err == nil {
		var name string
		if name, err = os.Readlink(pathname.String()); err != nil {
			return
		}
		checksum, err = Decode(path.Base(name))
		return
	}
	if !errors.Is(err, os.ErrNotExist) {
		return
	}

	var checksums string
	defer func() {
		if err == nil && checksums != "" {
			err = os.Symlink(
				checksumLinknamePrefix+checksums,
				pathname.String(),
			)
		}
	}()

	var checksumPathname *check.Absolute
	var checksumFi os.FileInfo
	if kc, ok := a.(KnownChecksum); ok {
		checksum = kc.Checksum()
		checksums = Encode(checksum)
		checksumPathname = c.base.Append(
			dirChecksum,
			checksums,
		)

		c.checksumMu.RLock()
		checksumFi, err = os.Stat(checksumPathname.String())
		c.checksumMu.RUnlock()

		if err != nil {
			if !errors.Is(err, os.ErrNotExist) {
				return
			}

			checksumFi, err = nil, nil
		}
	}

	if f, ok := a.(File); ok {
		if checksumFi != nil {
			if !checksumFi.Mode().IsRegular() {
				// unreachable
				err = InvalidFileModeError(checksumFi.Mode())
			}
			return
		}

		var data []byte
		data, err = f.Data()
		if err != nil {
			return
		}

		if checksumPathname == nil {
			h := sha512.New384()
			h.Write(data)
			h.Sum(checksum[:0])
			checksums = Encode(checksum)
			checksumPathname = c.base.Append(
				dirChecksum,
				checksums,
			)
		} else if c.IsStrict() {
			h := sha512.New384()
			h.Write(data)
			if got := Checksum(h.Sum(nil)); got != checksum {
				err = &ChecksumMismatchError{
					Got:  got,
					Want: checksum,
				}
				return
			}
		}

		c.checksumMu.Lock()
		var w *os.File
		w, err = os.OpenFile(
			checksumPathname.String(),
			os.O_CREATE|os.O_EXCL|os.O_WRONLY,
			0400,
		)
		if err != nil {
			c.checksumMu.Unlock()

			if errors.Is(err, os.ErrExist) {
				err = nil
			}
			return
		}
		_, err = w.Write(data)
		closeErr := w.Close()
		if err == nil {
			err = closeErr
		}
		c.checksumMu.Unlock()

		return
	} else {
		if checksumFi != nil {
			if !checksumFi.Mode().IsDir() {
				// unreachable
				err = InvalidFileModeError(checksumFi.Mode())
			}
			return
		}

		workPathname := c.base.Append(dirWork, ids)
		defer func() {
			// must not use the value of checksum string as it might be zeroed
			// to cancel the deferred symlink operation

			if err != nil {
				chmodErr := filepath.WalkDir(workPathname.String(), func(
					path string,
					d fs.DirEntry,
					err error,
				) error {
					if err != nil {
						return err
					}
					if d.IsDir() {
						return os.Chmod(path, 0700)
					}
					return nil
				})
				if errors.Is(chmodErr, os.ErrNotExist) {
					chmodErr = nil
				}
				removeErr := os.RemoveAll(workPathname.String())
				if chmodErr != nil || removeErr != nil {
					err = errors.Join(err, chmodErr, removeErr)
				} else if errors.Is(err, os.ErrExist) {
					// two artifacts may be backed by the same file
					err = nil
				}
			}
		}()

		if err = a.Cure(workPathname, c.loadData); err != nil {
			return
		}

		var fi os.FileInfo
		if fi, err = os.Lstat(workPathname.String()); err != nil {
			if errors.Is(err, os.ErrNotExist) {
				err = NoOutputError{}
			}
			return
		}

		if !fi.IsDir() {
			if !fi.Mode().IsRegular() {
				err = InvalidFileModeError(fi.Mode())
			} else {
				err = errors.New("non-file artifact produced regular file")
			}
			return
		}

		// override this before hashing since it will be made read-only after
		// the rename anyway so do not let perm bits affect the checksum
		if err = os.Chmod(workPathname.String(), 0700); err != nil {
			return
		}
		var gotChecksum Checksum
		if gotChecksum, err = HashDir(workPathname); err != nil {
			return
		}

		if checksumPathname == nil {
			checksum = gotChecksum
			checksums = Encode(checksum)
			checksumPathname = c.base.Append(
				dirChecksum,
				checksums,
			)
		} else {
			if gotChecksum != checksum {
				err = &ChecksumMismatchError{
					Got:  gotChecksum,
					Want: checksum,
				}
				return
			}
		}

		c.checksumMu.Lock()
		if err = os.Rename(
			workPathname.String(),
			checksumPathname.String(),
		); err != nil {
			if !errors.Is(err, os.ErrExist) {
				c.checksumMu.Unlock()
				return
			}
			// err is zeroed during deferred cleanup
		} else {
			err = os.Chmod(checksumPathname.String(), 0500)
		}
		c.checksumMu.Unlock()

		return
	}
}

// New returns the address to a new instance of [Cache].
func New(base *check.Absolute) (*Cache, error) {
	for _, name := range []string{
		dirIdentifier,
		dirChecksum,
		dirWork,
	} {
		if err := os.MkdirAll(base.Append(name).String(), 0700); err != nil &&
			!errors.Is(err, os.ErrExist) {
			return nil, err
		}
	}

	return &Cache{
		base: base,

		ident:        make(map[ID]Checksum),
		identErr:     make(map[ID]error),
		identPending: make(map[ID]<-chan struct{}),
	}, nil
}