package splitstore import ( "crypto/rand" "golang.org/x/xerrors" bbloom "github.com/ipfs/bbloom" cid "github.com/ipfs/go-cid" blake2b "github.com/minio/blake2b-simd" ) const ( BloomFilterMinSize = 10_000_000 BloomFilterProbability = 0.01 ) type BloomMarkSetEnv struct{} var _ MarkSetEnv = (*BloomMarkSetEnv)(nil) type BloomMarkSet struct { salt []byte bf *bbloom.Bloom } var _ MarkSet = (*BloomMarkSet)(nil) func NewBloomMarkSetEnv() (*BloomMarkSetEnv, error) { return &BloomMarkSetEnv{}, nil } func (e *BloomMarkSetEnv) Create(name string, sizeHint int64) (MarkSet, error) { size := int64(BloomFilterMinSize) for size < sizeHint { size += BloomFilterMinSize } salt := make([]byte, 4) _, err := rand.Read(salt) if err != nil { return nil, xerrors.Errorf("error reading salt: %w", err) } bf, err := bbloom.New(float64(size), BloomFilterProbability) if err != nil { return nil, xerrors.Errorf("error creating bloom filter: %w", err) } return &BloomMarkSet{salt: salt, bf: bf}, nil } func (e *BloomMarkSetEnv) Close() error { return nil } func (s *BloomMarkSet) saltedKey(cid cid.Cid) []byte { hash := cid.Hash() key := make([]byte, len(s.salt)+len(hash)) n := copy(key, s.salt) copy(key[n:], hash) rehash := blake2b.Sum256(key) return rehash[:] } func (s *BloomMarkSet) Mark(cid cid.Cid) error { s.bf.Add(s.saltedKey(cid)) return nil } func (s *BloomMarkSet) Has(cid cid.Cid) (bool, error) { return s.bf.Has(s.saltedKey(cid)), nil } func (s *BloomMarkSet) Close() error { return nil }