|
- // 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
-
- // Note: golang compiler seems so behave differently depending on the current
- // locale, so some conversions might behave differently. For that reason, we don't
- // depend on particular conversions but in expected behaviors.
-
- res, err = ToUTF8WithErr([]byte{0x41, 0x42, 0x43})
- assert.NoError(t, err)
- assert.Equal(t, "ABC", res)
-
- // "áéíóú"
- res, err = ToUTF8WithErr([]byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
- assert.NoError(t, err)
- assert.Equal(t, []byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba}, []byte(res))
-
- // "áéíóú"
- res, err = ToUTF8WithErr([]byte{0xef, 0xbb, 0xbf, 0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3,
- 0xc3, 0xba})
- assert.NoError(t, err)
- assert.Equal(t, []byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba}, []byte(res))
-
- res, err = ToUTF8WithErr([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63,
- 0xF3, 0x6D, 0x6F, 0x20, 0xF1, 0x6F, 0x73, 0x41, 0x41, 0x41, 0x2e})
- assert.NoError(t, err)
- stringMustStartWith(t, "Hola,", res)
- stringMustEndWith(t, "AAA.", res)
-
- res, err = ToUTF8WithErr([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63,
- 0xF3, 0x6D, 0x6F, 0x20, 0x07, 0xA4, 0x6F, 0x73, 0x41, 0x41, 0x41, 0x2e})
- assert.NoError(t, err)
- stringMustStartWith(t, "Hola,", res)
- stringMustEndWith(t, "AAA.", res)
-
- res, err = ToUTF8WithErr([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63,
- 0xF3, 0x6D, 0x6F, 0x20, 0x81, 0xA4, 0x6F, 0x73, 0x41, 0x41, 0x41, 0x2e})
- assert.NoError(t, err)
- stringMustStartWith(t, "Hola,", res)
- stringMustEndWith(t, "AAA.", res)
-
- // Japanese (Shift-JIS)
- // 日属秘ぞしちゅ。
- res, err = ToUTF8WithErr([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82,
- 0xBF, 0x82, 0xE3, 0x81, 0x42})
- assert.NoError(t, err)
- assert.Equal(t, []byte{0xE6, 0x97, 0xA5, 0xE5, 0xB1, 0x9E, 0xE7, 0xA7, 0x98, 0xE3,
- 0x81, 0x9E, 0xE3, 0x81, 0x97, 0xE3, 0x81, 0xA1, 0xE3, 0x82, 0x85, 0xE3, 0x80, 0x82},
- []byte(res))
-
- res, err = ToUTF8WithErr([]byte{0x00, 0x00, 0x00, 0x00})
- assert.NoError(t, err)
- assert.Equal(t, []byte{0x00, 0x00, 0x00, 0x00}, []byte(res))
- }
-
- func TestToUTF8WithFallback(t *testing.T) {
- // "ABC"
- res := ToUTF8WithFallback([]byte{0x41, 0x42, 0x43})
- assert.Equal(t, []byte{0x41, 0x42, 0x43}, res)
-
- // "áéíóú"
- res = ToUTF8WithFallback([]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)
-
- // UTF8 BOM + "áéíóú"
- res = ToUTF8WithFallback([]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)
-
- // "Hola, así cómo ños"
- 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{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xC3, 0xAD, 0x20, 0x63,
- 0xC3, 0xB3, 0x6D, 0x6F, 0x20, 0xC3, 0xB1, 0x6F, 0x73}, res)
-
- // "Hola, así cómo "
- minmatch := []byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xC3, 0xAD, 0x20, 0x63, 0xC3, 0xB3, 0x6D, 0x6F, 0x20}
-
- 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)
- // "日属秘ぞしちゅ。"
- res = ToUTF8WithFallback([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82, 0xBF, 0x82, 0xE3, 0x81, 0x42})
- assert.Equal(t, []byte{0xE6, 0x97, 0xA5, 0xE5, 0xB1, 0x9E, 0xE7, 0xA7, 0x98, 0xE3,
- 0x81, 0x9E, 0xE3, 0x81, 0x97, 0xE3, 0x81, 0xA1, 0xE3, 0x82, 0x85, 0xE3, 0x80, 0x82}, res)
-
- res = ToUTF8WithFallback([]byte{0x00, 0x00, 0x00, 0x00})
- assert.Equal(t, []byte{0x00, 0x00, 0x00, 0x00}, res)
- }
-
- func TestToUTF8(t *testing.T) {
-
- // Note: golang compiler seems so behave differently depending on the current
- // locale, so some conversions might behave differently. For that reason, we don't
- // depend on particular conversions but in expected behaviors.
-
- res := ToUTF8(string([]byte{0x41, 0x42, 0x43}))
- assert.Equal(t, "ABC", res)
-
- // "áéíóú"
- res = ToUTF8(string([]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}, []byte(res))
-
- // BOM + "áéíóú"
- res = ToUTF8(string([]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}, []byte(res))
-
- // Latin1
- // Hola, así cómo ños
- res = ToUTF8(string([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63,
- 0xF3, 0x6D, 0x6F, 0x20, 0xF1, 0x6F, 0x73}))
- assert.Equal(t, []byte{0x48, 0x6f, 0x6c, 0x61, 0x2c, 0x20, 0x61, 0x73, 0xc3, 0xad, 0x20, 0x63,
- 0xc3, 0xb3, 0x6d, 0x6f, 0x20, 0xc3, 0xb1, 0x6f, 0x73}, []byte(res))
-
- // Latin1
- // Hola, así cómo \x07ños
- res = ToUTF8(string([]byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xED, 0x20, 0x63,
- 0xF3, 0x6D, 0x6F, 0x20, 0x07, 0xA4, 0x6F, 0x73}))
- // Hola,
- bytesMustStartWith(t, []byte{0x48, 0x6F, 0x6C, 0x61, 0x2C}, []byte(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)
- // 日属秘ぞしちゅ。
- res = ToUTF8(string([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82,
- 0xBF, 0x82, 0xE3, 0x81, 0x42}))
- assert.Equal(t, []byte{0xE6, 0x97, 0xA5, 0xE5, 0xB1, 0x9E, 0xE7, 0xA7, 0x98, 0xE3,
- 0x81, 0x9E, 0xE3, 0x81, 0x97, 0xE3, 0x81, 0xA1, 0xE3, 0x82, 0x85, 0xE3, 0x80, 0x82},
- []byte(res))
-
- res = ToUTF8("\x00\x00\x00\x00")
- assert.Equal(t, []byte{0x00, 0x00, 0x00, 0x00}, []byte(res))
- }
-
- func TestToUTF8DropErrors(t *testing.T) {
- // "ABC"
- res := ToUTF8DropErrors([]byte{0x41, 0x42, 0x43})
- assert.Equal(t, []byte{0x41, 0x42, 0x43}, res)
-
- // "áéíóú"
- res = ToUTF8DropErrors([]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)
-
- // UTF8 BOM + "áéíóú"
- res = ToUTF8DropErrors([]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)
-
- // "Hola, así cómo ños"
- 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{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73}, res[:8])
- assert.Equal(t, []byte{0x73}, res[len(res)-1:])
-
- // "Hola, así cómo "
- minmatch := []byte{0x48, 0x6F, 0x6C, 0x61, 0x2C, 0x20, 0x61, 0x73, 0xC3, 0xAD, 0x20, 0x63, 0xC3, 0xB3, 0x6D, 0x6F, 0x20}
-
- 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)
- // "日属秘ぞしちゅ。"
- res = ToUTF8DropErrors([]byte{0x93, 0xFA, 0x91, 0xAE, 0x94, 0xE9, 0x82, 0xBC, 0x82, 0xB5, 0x82, 0xBF, 0x82, 0xE3, 0x81, 0x42})
- assert.Equal(t, []byte{0xE6, 0x97, 0xA5, 0xE5, 0xB1, 0x9E, 0xE7, 0xA7, 0x98, 0xE3,
- 0x81, 0x9E, 0xE3, 0x81, 0x97, 0xE3, 0x81, 0xA1, 0xE3, 0x82, 0x85, 0xE3, 0x80, 0x82}, 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)
- }
-
- func stringMustStartWith(t *testing.T, expected string, value string) {
- assert.Equal(t, expected, string(value[:len(expected)]))
- }
-
- func stringMustEndWith(t *testing.T, expected string, value string) {
- assert.Equal(t, expected, string(value[len(value)-len(expected):]))
- }
-
- func bytesMustStartWith(t *testing.T, expected []byte, value []byte) {
- assert.Equal(t, expected, value[:len(expected)])
- }
|