Convert files to utf-8 for indexing (#7814)

* Convert files to utf-8 for indexing

* Move utf8 functions to modules/base

* Bump repoIndexerLatestVersion to 3

* Add tests for base/encoding.go

* Changes to pass gosimple

* Move UTF8 funcs into new modules/charset package

2 files changed, 343 insertions, 0 deletions

diff --git a/modules/charset/charset.go b/modules/charset/charset.go
new file mode 100644
index 0000000000..f43fe387ed
--- /dev/null
+++ b/modules/charset/charset.go
@@ -0,0 +1,152 @@
+// Copyright 2014 The Gogs Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package charset
+
+import (
+	"bytes"
+	"fmt"
+	"unicode/utf8"
+
+	"code.gitea.io/gitea/modules/log"
+	"code.gitea.io/gitea/modules/setting"
+
+	"github.com/gogits/chardet"
+	"golang.org/x/net/html/charset"
+	"golang.org/x/text/transform"
+)
+
+// UTF8BOM is the utf-8 byte-order marker
+var UTF8BOM = []byte{'\xef', '\xbb', '\xbf'}
+
+// ToUTF8WithErr converts content to UTF8 encoding
+func ToUTF8WithErr(content []byte) (string, error) {
+	charsetLabel, err := DetectEncoding(content)
+	if err != nil {
+		return "", err
+	} else if charsetLabel == "UTF-8" {
+		return string(RemoveBOMIfPresent(content)), nil
+	}
+
+	encoding, _ := charset.Lookup(charsetLabel)
+	if encoding == nil {
+		return string(content), fmt.Errorf("Unknown encoding: %s", charsetLabel)
+	}
+
+	// If there is an error, we concatenate the nicely decoded part and the
+	// original left over. This way we won't lose data.
+	result, n, err := transform.Bytes(encoding.NewDecoder(), content)
+	if err != nil {
+		result = append(result, content[n:]...)
+	}
+
+	result = RemoveBOMIfPresent(result)
+
+	return string(result), err
+}
+
+// ToUTF8WithFallback detects the encoding of content and coverts to UTF-8 if possible
+func ToUTF8WithFallback(content []byte) []byte {
+	charsetLabel, err := DetectEncoding(content)
+	if err != nil || charsetLabel == "UTF-8" {
+		return RemoveBOMIfPresent(content)
+	}
+
+	encoding, _ := charset.Lookup(charsetLabel)
+	if encoding == nil {
+		return content
+	}
+
+	// If there is an error, we concatenate the nicely decoded part and the
+	// original left over. This way we won't lose data.
+	result, n, err := transform.Bytes(encoding.NewDecoder(), content)
+	if err != nil {
+		return append(result, content[n:]...)
+	}
+
+	return RemoveBOMIfPresent(result)
+}
+
+// ToUTF8 converts content to UTF8 encoding and ignore error
+func ToUTF8(content string) string {
+	res, _ := ToUTF8WithErr([]byte(content))
+	return res
+}
+
+// ToUTF8DropErrors makes sure the return string is valid utf-8; attempts conversion if possible
+func ToUTF8DropErrors(content []byte) []byte {
+	charsetLabel, err := DetectEncoding(content)
+	if err != nil || charsetLabel == "UTF-8" {
+		return RemoveBOMIfPresent(content)
+	}
+
+	encoding, _ := charset.Lookup(charsetLabel)
+	if encoding == nil {
+		return content
+	}
+
+	// We ignore any non-decodable parts from the file.
+	// Some parts might be lost
+	var decoded []byte
+	decoder := encoding.NewDecoder()
+	idx := 0
+	for {
+		result, n, err := transform.Bytes(decoder, content[idx:])
+		decoded = append(decoded, result...)
+		if err == nil {
+			break
+		}
+		decoded = append(decoded, ' ')
+		idx = idx + n + 1
+		if idx >= len(content) {
+			break
+		}
+	}
+
+	return RemoveBOMIfPresent(decoded)
+}
+
+// RemoveBOMIfPresent removes a UTF-8 BOM from a []byte
+func RemoveBOMIfPresent(content []byte) []byte {
+	if len(content) > 2 && bytes.Equal(content[0:3], UTF8BOM) {
+		return content[3:]
+	}
+	return content
+}
+
+// DetectEncoding detect the encoding of content
+func DetectEncoding(content []byte) (string, error) {
+	if utf8.Valid(content) {
+		log.Debug("Detected encoding: utf-8 (fast)")
+		return "UTF-8", nil
+	}
+
+	textDetector := chardet.NewTextDetector()
+	var detectContent []byte
+	if len(content) < 1024 {
+		// Check if original content is valid
+		if _, err := textDetector.DetectBest(content); err != nil {
+			return "", err
+		}
+		times := 1024 / len(content)
+		detectContent = make([]byte, 0, times*len(content))
+		for i := 0; i < times; i++ {
+			detectContent = append(detectContent, content...)
+		}
+	} else {
+		detectContent = content
+	}
+	result, err := textDetector.DetectBest(detectContent)
+	if err != nil {
+		return "", err
+	}
+	// FIXME: to properly decouple this function the fallback ANSI charset should be passed as an argument
+	if result.Charset != "UTF-8" && len(setting.Repository.AnsiCharset) > 0 {
+		log.Debug("Using default AnsiCharset: %s", setting.Repository.AnsiCharset)
+		return setting.Repository.AnsiCharset, err
+	}
+
+	log.Debug("Detected encoding: %s", result.Charset)
+	return result.Charset, err
+}
diff --git a/modules/charset/charset_test.go b/modules/charset/charset_test.go
new file mode 100644
index 0000000000..fc9ec6a979
--- /dev/null
+++ b/modules/charset/charset_test.go
@@ -0,0 +1,191 @@
+// Copyright 2019 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package charset
+
+import (
+	"testing"
+
+	"code.gitea.io/gitea/modules/setting"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestRemoveBOMIfPresent(t *testing.T) {
+	res := RemoveBOMIfPresent([]byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, []byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba}, res)
+
+	res = RemoveBOMIfPresent([]byte{0xef, 0xbb, 0xbf, 0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, []byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba}, res)
+}
+
+func TestToUTF8WithErr(t *testing.T) {
+	var res string
+	var err error
+
+	res, err = ToUTF8WithErr([]byte{0x41, 0x42, 0x43})
+	assert.Equal(t, "ABC", res)
+	assert.NoError(t, err)
+
+	res, err = ToUTF8WithErr([]byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, "áéíóú", res)
+	assert.NoError(t, err)
+
+	res, err = ToUTF8WithErr([]byte{0xef, 0xbb, 0xbf, 0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, "áéíóú", res)
+	assert.NoError(t, err)
+
+	// This test FAILS
+	res, err = ToUTF8WithErr([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0xF1, 0x6F, 0x73})
+	assert.Equal(t, "Hola, así cómo ños", res)
+	assert.NoError(t, err)
+
+	res, err = ToUTF8WithErr([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0x07, 0xA4, 0x6F, 0x73})
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Regexp(t, "^Hola, así cómo", res)
+	assert.NoError(t, err)
+
+	res, err = ToUTF8WithErr([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0x81, 0xA4, 0x6F, 0x73})
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Regexp(t, "^Hola, así cómo", res)
+	assert.NoError(t, err)
+
+	// Japanese (Shift-JIS)
+	res, err = ToUTF8WithErr([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82, 0xBF, 0x82, 0xE3, 0x81, 0x42})
+	assert.Equal(t, "日属秘ぞしちゅ。", res)
+	assert.NoError(t, err)
+
+	res, err = ToUTF8WithErr([]byte{0x00, 0x00, 0x00, 0x00})
+	assert.Equal(t, "\x00\x00\x00\x00", res)
+	assert.NoError(t, err)
+}
+
+func TestToUTF8WithFallback(t *testing.T) {
+	res := ToUTF8WithFallback([]byte{0x41, 0x42, 0x43})
+	assert.Equal(t, []byte("ABC"), res)
+
+	res = ToUTF8WithFallback([]byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, []byte("áéíóú"), res)
+
+	res = ToUTF8WithFallback([]byte{0xef, 0xbb, 0xbf, 0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, []byte("áéíóú"), res)
+
+	res = ToUTF8WithFallback([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0xF1, 0x6F, 0x73})
+	assert.Equal(t, []byte("Hola, así cómo ños"), res)
+
+	minmatch := []byte("Hola, así cómo ")
+
+	res = ToUTF8WithFallback([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0x07, 0xA4, 0x6F, 0x73})
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Equal(t, minmatch, res[0:len(minmatch)])
+
+	res = ToUTF8WithFallback([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0x81, 0xA4, 0x6F, 0x73})
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Equal(t, minmatch, res[0:len(minmatch)])
+
+	// Japanese (Shift-JIS)
+	res = ToUTF8WithFallback([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82, 0xBF, 0x82, 0xE3, 0x81, 0x42})
+	assert.Equal(t, []byte("日属秘ぞしちゅ。"), res)
+
+	res = ToUTF8WithFallback([]byte{0x00, 0x00, 0x00, 0x00})
+	assert.Equal(t, []byte{0x00, 0x00, 0x00, 0x00}, res)
+}
+
+func TestToUTF8(t *testing.T) {
+	res := ToUTF8("ABC")
+	assert.Equal(t, "ABC", res)
+
+	res = ToUTF8("áéíóú")
+	assert.Equal(t, "áéíóú", res)
+
+	// With utf-8 BOM
+	res = ToUTF8("\ufeffáéíóú")
+	assert.Equal(t, "áéíóú", res)
+
+	res = ToUTF8("Hola, así cómo ños")
+	assert.Equal(t, "Hola, así cómo ños", res)
+
+	res = ToUTF8("Hola, así cómo \x07ños")
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Regexp(t, "^Hola, así cómo", res)
+
+	// This test FAILS
+	// res = ToUTF8("Hola, así cómo \x81ños")
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	// assert.Regexp(t, "^Hola, así cómo", res)
+
+	// Japanese (Shift-JIS)
+	res = ToUTF8("\x93\xFA\x91\xAE\x94\xE9\x82\xBC\x82\xB5\x82\xBF\x82\xE3\x81\x42")
+	assert.Equal(t, "日属秘ぞしちゅ。", res)
+
+	res = ToUTF8("\x00\x00\x00\x00")
+	assert.Equal(t, "\x00\x00\x00\x00", res)
+}
+
+func TestToUTF8DropErrors(t *testing.T) {
+	res := ToUTF8DropErrors([]byte{0x41, 0x42, 0x43})
+	assert.Equal(t, []byte("ABC"), res)
+
+	res = ToUTF8DropErrors([]byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, []byte("áéíóú"), res)
+
+	res = ToUTF8DropErrors([]byte{0xef, 0xbb, 0xbf, 0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
+	assert.Equal(t, []byte("áéíóú"), res)
+
+	res = ToUTF8DropErrors([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0xF1, 0x6F, 0x73})
+	assert.Equal(t, []byte("Hola, así cómo ños"), res)
+
+	minmatch := []byte("Hola, así cómo ")
+
+	res = ToUTF8DropErrors([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0x07, 0xA4, 0x6F, 0x73})
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Equal(t, minmatch, res[0:len(minmatch)])
+
+	res = ToUTF8DropErrors([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63, 0xF3, 0x6D, 0x6F, 0x20, 0x81, 0xA4, 0x6F, 0x73})
+	// Do not fail for differences in invalid cases, as the library might change the conversion criteria for those
+	assert.Equal(t, minmatch, res[0:len(minmatch)])
+
+	// Japanese (Shift-JIS)
+	res = ToUTF8DropErrors([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82, 0xBF, 0x82, 0xE3, 0x81, 0x42})
+	assert.Equal(t, []byte("日属秘ぞしちゅ。"), res)
+
+	res = ToUTF8DropErrors([]byte{0x00, 0x00, 0x00, 0x00})
+	assert.Equal(t, []byte{0x00, 0x00, 0x00, 0x00}, res)
+}
+
+func TestDetectEncoding(t *testing.T) {
+	testSuccess := func(b []byte, expected string) {
+		encoding, err := DetectEncoding(b)
+		assert.NoError(t, err)
+		assert.Equal(t, expected, encoding)
+	}
+	// utf-8
+	b := []byte("just some ascii")
+	testSuccess(b, "UTF-8")
+
+	// utf-8-sig: "hey" (with BOM)
+	b = []byte{0xef, 0xbb, 0xbf, 0x68, 0x65, 0x79}
+	testSuccess(b, "UTF-8")
+
+	// utf-16: "hey<accented G>"
+	b = []byte{0xff, 0xfe, 0x68, 0x00, 0x65, 0x00, 0x79, 0x00, 0xf4, 0x01}
+	testSuccess(b, "UTF-16LE")
+
+	// iso-8859-1: d<accented e>cor<newline>
+	b = []byte{0x44, 0xe9, 0x63, 0x6f, 0x72, 0x0a}
+	encoding, err := DetectEncoding(b)
+	assert.NoError(t, err)
+	// due to a race condition in `chardet` library, it could either detect
+	// "ISO-8859-1" or "IS0-8859-2" here. Technically either is correct, so
+	// we accept either.
+	assert.Contains(t, encoding, "ISO-8859")
+
+	setting.Repository.AnsiCharset = "placeholder"
+	testSuccess(b, "placeholder")
+
+	// invalid bytes
+	b = []byte{0xfa}
+	_, err = DetectEncoding(b)
+	assert.Error(t, err)
+}