Go - faster gzip reader

fastgzreader.go

package input

import (
	"bytes"
	"errors"
	"io"
	"os/exec"
	"strings"
)

// fastGzReader is an API-compatible drop-in replacement
// for gzip.Reader, that achieves a higher decoding speed
// by spawning an external zcat instance and pipeing data
// through it.
// Go's native gzip implementation is about 2x slower at
// decompressing data compared to zlib (mostly due to Go compiler
// inefficiencies). So for tasks where the gzip decoding
// speed is important, this is a quick workaround that doesn't
// require cgo.
// zcat is part of the gzip package and comes preinstalled on
// most Linux distributions and on OSX.
type fastGzReader struct {
	io.Reader
	stderr bytes.Buffer
	close  func() error
}

func newFastGzReader(r io.Reader) (fastGzReader, error) {
	var gz fastGzReader
	if err := gz.Reset(r); err != nil {
		return fastGzReader{}, err
	}
	return gz, nil
}

func (gz *fastGzReader) Reset(r io.Reader) error {
	if gz.close != nil {
		gz.close()
	}

	cmd := exec.Command("zcat")
	cmd.Stdin = r
	cmd.Stderr = &gz.stderr

	rpipe, err := cmd.StdoutPipe()
	if err != nil {
		return err
	}

	err = cmd.Start()
	if err != nil {
		rpipe.Close()
		return err
	}

	gz.Reader = rpipe
	gz.close = func() error {
		rpipe.Close()
		if err := cmd.Wait(); err != nil {
			if _, ok := err.(*exec.ExitError); ok {
				return errors.New(strings.TrimSpace(gz.stderr.String()))
			}
		}
		return err
	}
	return nil
}

func (gz fastGzReader) Close() error {
	return gz.close()
}

@rasky do you mind if I use this in an MIT-licensed project? Seems like the new gzip stuff for 1.7 is getting close, but this is still faster.

@brentp sorry for the delay; no problem with that. Consider this MIT licensed.

@brentp I've just pushed an updated version with error checking