Reputation: 312
Debugging differences between Python's zlib and golang's zlib. Why don't the following have the same results?
compress.go
:
package main
import (
"compress/flate"
"bytes"
"fmt"
)
func compress(source string) []byte {
w, _ := flate.NewWriter(nil, 7)
buf := new(bytes.Buffer)
w.Reset(buf)
w.Write([]byte(source))
w.Close()
return buf.Bytes()
}
func main() {
example := "foo"
compressed := compress(example)
fmt.Println(compressed)
}
compress.py
:
from __future__ import print_function
import zlib
def compress(source):
# golang zlib strips header + checksum
compressor = zlib.compressobj(7, zlib.DEFLATED, -15)
compressor.compress(source)
# python zlib defaults to Z_FLUSH, but
# https://golang.org/pkg/compress/flate/#Writer.Flush
# says "Flush is equivalent to Z_SYNC_FLUSH"
return compressor.flush(zlib.Z_SYNC_FLUSH)
def main():
example = u"foo"
compressed = compress(example)
print(list(bytearray(compressed)))
if __name__ == "__main__":
main()
Results
$ go version
go version go1.7.3 darwin/amd64
$ go build compress.go
$ ./compress
[74 203 207 7 4 0 0 255 255]
$ python --version
$ python 2.7.12
$ python compress.py
[74, 203, 207, 7, 0, 0, 0, 255, 255]
The Python version has 0
for the fifth byte, but the golang version has 4
-- what's causing the different output?
Upvotes: 8
Views: 2288
Reputation: 109442
The output from the python example isn't a "complete" stream, its just flushing the buffer after compressing the first string. You can get the same output from the Go code by replacing Close()
with Flush()
:
https://play.golang.org/p/BMcjTln-ej
func compress(source string) []byte {
buf := new(bytes.Buffer)
w, _ := flate.NewWriter(buf, 7)
w.Write([]byte(source))
w.Flush()
return buf.Bytes()
}
However, you are comparing output from zlib in python, which uses DEFLATE internally to produce a zlib format output, and flate
in Go, which is a DEFLATE implementation. I don't know if you can get the python zlib library to output the raw, complete DEFLATE stream, but trying to get different libraries to output byte-for-byte matches of compressed data doesn't seem useful or maintainable. The output of the compression libraries is only guaranteed to be compatible, not identical.
Upvotes: 4