Charset detection heuristics added with chardet lib.

client/client.go

@@ -1,10 +1,12 @@
 package client
 
 import (
+	"bytes"
 	"context"
 	"github.com/chromedp/cdproto/dom"
 	"github.com/chromedp/cdproto/network"
 	"github.com/chromedp/chromedp"
+	"github.com/musabgultekin/chardet"
 	"github.com/pkg/errors"
 	"golang.org/x/net/html/charset"
 	"io"
@@ -26,6 +28,9 @@ type Client struct {
 	*http.Client
 }
 
+const DefaultUserAgent = "Geziyor 1.0"
+const DefaultMaxBody int64 = 1024 * 1024 * 1024 // 1GB
+
 // NewClient creates http.Client with modified values for typical web scraper
 func NewClient() *Client {
 	client := &http.Client{
@@ -70,11 +75,11 @@ func (c *Client) DoRequestClient(req *Request, maxBodySize int64, charsetDetectD
 	// Limit response body reading
 	bodyReader := io.LimitReader(resp.Body, maxBodySize)
 
-	// Start reading body and determine encoding
-	if !charsetDetectDisabled && resp.Request.Method != "HEAD" {
-		bodyReader, err = charset.NewReader(bodyReader, resp.Header.Get("Content-Type"))
+	// Convert response if encoding provided
+	if req.Encoding != "" && resp.Request.Method != "HEAD" {
+		bodyReader, err = charset.NewReader(bodyReader, "text/html; charset="+req.Encoding)
 		if err != nil {
-			return nil, errors.Wrap(err, "Determine encoding error")
+			return nil, errors.Wrap(err, "Reading provided encoding error")
 		}
 	}
 
@@ -83,6 +88,27 @@ func (c *Client) DoRequestClient(req *Request, maxBodySize int64, charsetDetectD
 		return nil, errors.Wrap(err, "Reading body error")
 	}
 
+	// Decoding body
+	if req.Encoding == "" && resp.Request.Method != "HEAD" {
+		contentType := resp.Header.Get("Content-Type")
+		// Charset detection
+		// If enabled and charset not provided in content-type
+		if !charsetDetectDisabled && !strings.Contains(contentType, "charset") {
+			if res, err := chardet.NewHtmlDetector().DetectBest(body); err == nil {
+				contentType = "text/html; charset=" + res.Charset
+			}
+		}
+		convertedReader, err := charset.NewReader(bytes.NewReader(body), contentType)
+		if err != nil {
+			return nil, errors.Wrap(err, "Determine encoding error")
+		}
+		convertedBody, err := ioutil.ReadAll(convertedReader)
+		if err != nil {
+			return nil, errors.Wrap(err, "Determine encoding error")
+		}
+		body = convertedBody
+	}
+
 	response := Response{
 		Response: resp,
 		Body:     body,
@@ -137,7 +163,8 @@ func (c *Client) DoRequestChrome(req *Request) (*Response, error) {
 		return nil, errors.Wrap(err, "Request getting rendered error")
 	}
 
-	// Set new URL in case of redirection
+	// Update changed data
+	req.Header = ConvertMapToHeader(res.RequestHeaders)
 	req.URL, _ = url.Parse(res.URL)
 
 	response := Response{

client/client_test.go

@@ -1,7 +1,9 @@
 package client
 
 import (
+	"fmt"
 	"net/http"
+	"net/http/httptest"
 	"reflect"
 	"testing"
 )
@@ -89,3 +91,49 @@ func TestConvertMapToHeader(t *testing.T) {
 		})
 	}
 }
+
+func TestCharsetFromHeaders(t *testing.T) {
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "text/plain; charset=iso-8859-9")
+		fmt.Fprint(w, "G\xfcltekin")
+	}))
+	defer ts.Close()
+
+	req, _ := NewRequest("GET", ts.URL, nil)
+	res, _ := NewClient().DoRequestClient(req, DefaultMaxBody, false)
+
+	if string(res.Body) != "Gültekin" {
+		t.Fatal(string(res.Body))
+	}
+}
+
+func TestCharsetFromBody(t *testing.T) {
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "text/plain")
+		fmt.Fprint(w, "G\xfcltekin")
+	}))
+	defer ts.Close()
+
+	req, _ := NewRequest("GET", ts.URL, nil)
+	res, _ := NewClient().DoRequestClient(req, DefaultMaxBody, false)
+
+	if string(res.Body) != "Gültekin" {
+		t.Fatal(string(res.Body))
+	}
+}
+
+func TestCharsetProvidedWithRequest(t *testing.T) {
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "text/plain")
+		fmt.Fprint(w, "G\xfcltekin")
+	}))
+	defer ts.Close()
+
+	req, _ := NewRequest("GET", ts.URL, nil)
+	req.Encoding = "windows-1254"
+	res, _ := NewClient().DoRequestClient(req, DefaultMaxBody, false)
+
+	if string(res.Body) != "Gültekin" {
+		t.Fatal(string(res.Body))
+	}
+}

client/request.go

@@ -12,6 +12,7 @@ type Request struct {
 	Synchronized bool
 	Rendered     bool
 	Cancelled    bool
+	Encoding     string
 }
 
 // Cancel request