Documentation
¶
Index ¶
Constants ¶
const ( NameePE = "pe" // PENormalizer NameNTFS = "ntfs" // NTFSTimestampNormalizer )
Stable, persisted names for each normalizer. These are written into the repository config so that read paths (restore, verify, index rebuild, prune) can reconstruct exactly the normalizer the backup used — the chunk identity hash is computed over normalized bytes but the original bytes are stored, so verification must re-normalize with the same normalizer.
Variables ¶
This section is empty.
Functions ¶
func IdentityHashInput ¶
func IdentityHashInput(n Normalizer, data []byte) []byte
IdentityHashInput returns the bytes whose SHA-256 is the chunk identity: the normalized form when a normalizer is set, otherwise data unchanged. Read paths use this to verify stored (original) chunk bytes against the normalized identity hash recorded in the manifest.
func Names ¶
func Names(n Normalizer) []string
Names returns the stable persisted name(s) for a normalizer, flattening a CompositeNormalizer into its parts. Returns nil for a nil or no-op normalizer.
Types ¶
type CompositeNormalizer ¶
type CompositeNormalizer struct {
Normalizers []Normalizer
}
CompositeNormalizer chains multiple normalizers, piping the output of each into the next.
func (*CompositeNormalizer) Normalize ¶
func (c *CompositeNormalizer) Normalize(data []byte) []byte
type NTFSTimestampNormalizer ¶
type NTFSTimestampNormalizer struct{}
NTFSTimestampNormalizer zeros the four 8-byte timestamps in $STANDARD_INFORMATION (0x10) attributes of MFT records. These timestamps differ across machines even for identical files, hurting cross-machine dedup. Only the hash input is modified; original data is stored faithfully.
func (*NTFSTimestampNormalizer) Normalize ¶
func (n *NTFSTimestampNormalizer) Normalize(data []byte) []byte
Normalize scans for MFT "FILE" records at 1024-byte boundaries and zeros the 32 bytes of timestamps in each $STANDARD_INFORMATION attribute.
type NoopNormalizer ¶
type NoopNormalizer struct{}
NoopNormalizer passes data through unchanged.
func (*NoopNormalizer) Normalize ¶
func (n *NoopNormalizer) Normalize(data []byte) []byte
type Normalizer ¶
type Normalizer interface {
// Normalize returns a normalized copy of data for hashing purposes.
// The original data is never modified.
Normalize(data []byte) []byte
}
Normalizer transforms chunk data before hashing to improve dedup across semantically-identical-but-byte-different content.
func FromNames ¶
func FromNames(names []string) (Normalizer, error)
FromNames reconstructs a Normalizer from persisted names. It returns nil when names is empty (no normalization). The order of names is preserved so composite normalization is reproduced exactly.
type PENormalizer ¶
type PENormalizer struct{}
PENormalizer zeros PE header timestamps and checksums to improve dedup of Windows executables across builds.
func (*PENormalizer) Normalize ¶
func (n *PENormalizer) Normalize(data []byte) []byte
Normalize zeros TimeDateStamp and CheckSum fields in any PE headers found within the data. Returns a copy; the original is unmodified.
Source Files