zstd: Flush vs Close

[fhkong]

After the encoder has finished writing data, the data flushed to the cache using Flush and Close functions is different.
Moreover, the data flushed into the cache using Flush function cannot be decoded.

problem reproduce:

var buf bytes.Buffer
encoder, err := zstd.NewWriter(&buf)
require.NoError(t, err)
encoder.Write(rawData)
encoder.Flush()
fmt.Printf("use Flush compressed data len: %d bytes\n", len(buf.Bytes()))

buf.Reset()
encoder.Reset(&buf)
encoder.Write(rawData)
encoder.Close()
fmt.Printf("use Close compressed data len: %d bytes\n", len(buf.Bytes()))

[klauspost]

Flush does not indicate EOF, but it will allow the reader to read up until where the Flush was used. Therefore a stream ending with a flush is not complete.

Unless you have very special requirements you do not need to call Flush. You will know when you need it.

[fhkong]

Okay, thanks