Update github.com/NYTimes/gziphandler

This commit is contained in:
Jordan Liggitt
2021-02-03 16:57:36 -05:00
parent 074a517720
commit 4e7a9e5bcc
20 changed files with 696 additions and 85 deletions

View File

@@ -1,10 +1,11 @@
package gziphandler
package gziphandler // import "github.com/NYTimes/gziphandler"
import (
"bufio"
"compress/gzip"
"fmt"
"io"
"mime"
"net"
"net/http"
"strconv"
@@ -28,9 +29,11 @@ const (
// The examples seem to indicate that it is.
DefaultQValue = 1.0
// DefaultMinSize defines the minimum size to reach to enable compression.
// It's 512 bytes.
DefaultMinSize = 512
// DefaultMinSize is the default minimum size until we enable gzip compression.
// 1500 bytes is the MTU size for the internet since that is the largest size allowed at the network layer.
// If you take a file that is 1300 bytes and compress it to 800 bytes, its still transmitted in that same 1500 byte packet regardless, so youve gained nothing.
// That being the case, you should restrict the gzip compression to files with a size greater than a single packet, 1400 bytes (1.4KB) is a safe value.
DefaultMinSize = 1400
)
// gzipWriterPools stores a sync.Pool for each compression level for reuse of
@@ -80,40 +83,71 @@ type GzipResponseWriter struct {
minSize int // Specifed the minimum response size to gzip. If the response length is bigger than this value, it is compressed.
buf []byte // Holds the first part of the write before reaching the minSize or the end of the write.
ignore bool // If true, then we immediately passthru writes to the underlying ResponseWriter.
contentTypes []parsedContentType // Only compress if the response is one of these content-types. All are accepted if empty.
}
type GzipResponseWriterWithCloseNotify struct {
*GzipResponseWriter
}
func (w GzipResponseWriterWithCloseNotify) CloseNotify() <-chan bool {
return w.ResponseWriter.(http.CloseNotifier).CloseNotify()
}
// Write appends data to the gzip writer.
func (w *GzipResponseWriter) Write(b []byte) (int, error) {
// If content type is not set.
if _, ok := w.Header()[contentType]; !ok {
// It infer it from the uncompressed body.
w.Header().Set(contentType, http.DetectContentType(b))
}
// GZIP responseWriter is initialized. Use the GZIP responseWriter.
if w.gw != nil {
n, err := w.gw.Write(b)
return n, err
return w.gw.Write(b)
}
// If we have already decided not to use GZIP, immediately passthrough.
if w.ignore {
return w.ResponseWriter.Write(b)
}
// Save the write into a buffer for later use in GZIP responseWriter (if content is long enough) or at close with regular responseWriter.
// On the first write, w.buf changes from nil to a valid slice
w.buf = append(w.buf, b...)
// If the global writes are bigger than the minSize, compression is enable.
if len(w.buf) >= w.minSize {
err := w.startGzip()
if err != nil {
return 0, err
var (
cl, _ = strconv.Atoi(w.Header().Get(contentLength))
ct = w.Header().Get(contentType)
ce = w.Header().Get(contentEncoding)
)
// Only continue if they didn't already choose an encoding or a known unhandled content length or type.
if ce == "" && (cl == 0 || cl >= w.minSize) && (ct == "" || handleContentType(w.contentTypes, ct)) {
// If the current buffer is less than minSize and a Content-Length isn't set, then wait until we have more data.
if len(w.buf) < w.minSize && cl == 0 {
return len(b), nil
}
// If the Content-Length is larger than minSize or the current buffer is larger than minSize, then continue.
if cl >= w.minSize || len(w.buf) >= w.minSize {
// If a Content-Type wasn't specified, infer it from the current buffer.
if ct == "" {
ct = http.DetectContentType(w.buf)
w.Header().Set(contentType, ct)
}
// If the Content-Type is acceptable to GZIP, initialize the GZIP writer.
if handleContentType(w.contentTypes, ct) {
if err := w.startGzip(); err != nil {
return 0, err
}
return len(b), nil
}
}
}
// If we got here, we should not GZIP this response.
if err := w.startPlain(); err != nil {
return 0, err
}
return len(b), nil
}
// startGzip initialize any GZIP specific informations.
// startGzip initializes a GZIP writer and writes the buffer.
func (w *GzipResponseWriter) startGzip() error {
// Set the GZIP header.
w.Header().Set(contentEncoding, "gzip")
@@ -125,28 +159,57 @@ func (w *GzipResponseWriter) startGzip() error {
// Write the header to gzip response.
if w.code != 0 {
w.ResponseWriter.WriteHeader(w.code)
// Ensure that no other WriteHeader's happen
w.code = 0
}
// Initialize the GZIP response.
w.init()
// Initialize and flush the buffer into the gzip response if there are any bytes.
// If there aren't any, we shouldn't initialize it yet because on Close it will
// write the gzip header even if nothing was ever written.
if len(w.buf) > 0 {
// Initialize the GZIP response.
w.init()
n, err := w.gw.Write(w.buf)
// Flush the buffer into the gzip reponse.
n, err := w.gw.Write(w.buf)
// This should never happen (per io.Writer docs), but if the write didn't
// accept the entire buffer but returned no specific error, we have no clue
// what's going on, so abort just to be safe.
if err == nil && n < len(w.buf) {
err = io.ErrShortWrite
}
return err
}
return nil
}
// startPlain writes to sent bytes and buffer the underlying ResponseWriter without gzip.
func (w *GzipResponseWriter) startPlain() error {
if w.code != 0 {
w.ResponseWriter.WriteHeader(w.code)
// Ensure that no other WriteHeader's happen
w.code = 0
}
w.ignore = true
// If Write was never called then don't call Write on the underlying ResponseWriter.
if w.buf == nil {
return nil
}
n, err := w.ResponseWriter.Write(w.buf)
w.buf = nil
// This should never happen (per io.Writer docs), but if the write didn't
// accept the entire buffer but returned no specific error, we have no clue
// what's going on, so abort just to be safe.
if err == nil && n < len(w.buf) {
return io.ErrShortWrite
err = io.ErrShortWrite
}
w.buf = nil
return err
}
// WriteHeader just saves the response code until close or GZIP effective writes.
func (w *GzipResponseWriter) WriteHeader(code int) {
w.code = code
if w.code == 0 {
w.code = code
}
}
// init graps a new gzip writer from the gzipWriterPool and writes the correct
@@ -161,21 +224,20 @@ func (w *GzipResponseWriter) init() {
// Close will close the gzip.Writer and will put it back in the gzipWriterPool.
func (w *GzipResponseWriter) Close() error {
if w.gw == nil {
// Gzip not trigged yet, write out regular response.
if w.code != 0 {
w.ResponseWriter.WriteHeader(w.code)
}
if w.buf != nil {
_, writeErr := w.ResponseWriter.Write(w.buf)
// Returns the error if any at write.
if writeErr != nil {
return fmt.Errorf("gziphandler: write to regular responseWriter at close gets error: %q", writeErr.Error())
}
}
if w.ignore {
return nil
}
if w.gw == nil {
// GZIP not triggered yet, write out regular response.
err := w.startPlain()
// Returns the error if any at write.
if err != nil {
err = fmt.Errorf("gziphandler: write to regular responseWriter at close gets error: %q", err.Error())
}
return err
}
err := w.gw.Close()
gzipWriterPools[w.index].Put(w.gw)
w.gw = nil
@@ -186,6 +248,14 @@ func (w *GzipResponseWriter) Close() error {
// http.ResponseWriter if it is an http.Flusher. This makes GzipResponseWriter
// an http.Flusher.
func (w *GzipResponseWriter) Flush() {
if w.gw == nil && !w.ignore {
// Only flush once startGzip or startPlain has been called.
//
// Flush is thus a no-op until we're certain whether a plain
// or gzipped response will be served.
return
}
if w.gw != nil {
w.gw.Flush()
}
@@ -230,27 +300,44 @@ func NewGzipLevelHandler(level int) (func(http.Handler) http.Handler, error) {
// NewGzipLevelAndMinSize behave as NewGzipLevelHandler except it let the caller
// specify the minimum size before compression.
func NewGzipLevelAndMinSize(level, minSize int) (func(http.Handler) http.Handler, error) {
if level != gzip.DefaultCompression && (level < gzip.BestSpeed || level > gzip.BestCompression) {
return nil, fmt.Errorf("invalid compression level requested: %d", level)
return GzipHandlerWithOpts(CompressionLevel(level), MinSize(minSize))
}
func GzipHandlerWithOpts(opts ...option) (func(http.Handler) http.Handler, error) {
c := &config{
level: gzip.DefaultCompression,
minSize: DefaultMinSize,
}
if minSize < 0 {
return nil, fmt.Errorf("minimum size must be more than zero")
for _, o := range opts {
o(c)
}
if err := c.validate(); err != nil {
return nil, err
}
return func(h http.Handler) http.Handler {
index := poolIndex(level)
index := poolIndex(c.level)
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.Header().Add(vary, acceptEncoding)
if acceptsGzip(r) {
gw := &GzipResponseWriter{
ResponseWriter: w,
index: index,
minSize: minSize,
minSize: c.minSize,
contentTypes: c.contentTypes,
}
defer gw.Close()
h.ServeHTTP(gw, r)
if _, ok := w.(http.CloseNotifier); ok {
gwcn := GzipResponseWriterWithCloseNotify{gw}
h.ServeHTTP(gwcn, r)
} else {
h.ServeHTTP(gw, r)
}
} else {
h.ServeHTTP(w, r)
}
@@ -258,6 +345,98 @@ func NewGzipLevelAndMinSize(level, minSize int) (func(http.Handler) http.Handler
}, nil
}
// Parsed representation of one of the inputs to ContentTypes.
// See https://golang.org/pkg/mime/#ParseMediaType
type parsedContentType struct {
mediaType string
params map[string]string
}
// equals returns whether this content type matches another content type.
func (pct parsedContentType) equals(mediaType string, params map[string]string) bool {
if pct.mediaType != mediaType {
return false
}
// if pct has no params, don't care about other's params
if len(pct.params) == 0 {
return true
}
// if pct has any params, they must be identical to other's.
if len(pct.params) != len(params) {
return false
}
for k, v := range pct.params {
if w, ok := params[k]; !ok || v != w {
return false
}
}
return true
}
// Used for functional configuration.
type config struct {
minSize int
level int
contentTypes []parsedContentType
}
func (c *config) validate() error {
if c.level != gzip.DefaultCompression && (c.level < gzip.BestSpeed || c.level > gzip.BestCompression) {
return fmt.Errorf("invalid compression level requested: %d", c.level)
}
if c.minSize < 0 {
return fmt.Errorf("minimum size must be more than zero")
}
return nil
}
type option func(c *config)
func MinSize(size int) option {
return func(c *config) {
c.minSize = size
}
}
func CompressionLevel(level int) option {
return func(c *config) {
c.level = level
}
}
// ContentTypes specifies a list of content types to compare
// the Content-Type header to before compressing. If none
// match, the response will be returned as-is.
//
// Content types are compared in a case-insensitive, whitespace-ignored
// manner.
//
// A MIME type without any other directive will match a content type
// that has the same MIME type, regardless of that content type's other
// directives. I.e., "text/html" will match both "text/html" and
// "text/html; charset=utf-8".
//
// A MIME type with any other directive will only match a content type
// that has the same MIME type and other directives. I.e.,
// "text/html; charset=utf-8" will only match "text/html; charset=utf-8".
//
// By default, responses are gzipped regardless of
// Content-Type.
func ContentTypes(types []string) option {
return func(c *config) {
c.contentTypes = []parsedContentType{}
for _, v := range types {
mediaType, params, err := mime.ParseMediaType(v)
if err == nil {
c.contentTypes = append(c.contentTypes, parsedContentType{mediaType, params})
}
}
}
}
// GzipHandler wraps an HTTP handler, to transparently gzip the response body if
// the client supports it (via the Accept-Encoding header). This will compress at
// the default compression level.
@@ -273,6 +452,27 @@ func acceptsGzip(r *http.Request) bool {
return acceptedEncodings["gzip"] > 0.0
}
// returns true if we've been configured to compress the specific content type.
func handleContentType(contentTypes []parsedContentType, ct string) bool {
// If contentTypes is empty we handle all content types.
if len(contentTypes) == 0 {
return true
}
mediaType, params, err := mime.ParseMediaType(ct)
if err != nil {
return false
}
for _, c := range contentTypes {
if c.equals(mediaType, params) {
return true
}
}
return false
}
// parseEncodings attempts to parse a list of codings, per RFC 2616, as might
// appear in an Accept-Encoding header. It returns a map of content-codings to
// quality values, and an error containing the errors encountered. It's probably