diff options
author | Luke Shumaker <lukeshu@lukeshu.com> | 2023-02-16 19:06:46 -0700 |
---|---|---|
committer | Luke Shumaker <lukeshu@lukeshu.com> | 2023-02-18 22:45:54 -0700 |
commit | 00187950437a10952b82353405e5ba4b4515fb29 (patch) | |
tree | 826c4ff76310bf6f58e79f37f2107c329810aaa8 | |
parent | a87d6cbbb51a19071c5c742ef3c91bbb90a727c6 (diff) |
reencode: Don't normalize the capitalization of \uXXXX hex escapes
-rw-r--r-- | ReleaseNotes.md | 5 | ||||
-rw-r--r-- | compat/json/compat.go | 2 | ||||
-rw-r--r-- | compat/json/compat_test.go | 54 | ||||
-rw-r--r-- | encode_escape.go | 39 | ||||
-rw-r--r-- | internal/jsonstring/encode_string.go | 60 | ||||
-rw-r--r-- | reencode.go | 3 |
6 files changed, 121 insertions, 42 deletions
diff --git a/ReleaseNotes.md b/ReleaseNotes.md index 73df694..a8496e0 100644 --- a/ReleaseNotes.md +++ b/ReleaseNotes.md @@ -39,6 +39,11 @@ BackslashEscapeMode to allow emitted strings to contain invalid UTF-8. + + Feature: ReEncoder: No longer unconditionally normalizes + `\uXXXX` hex characters to lower-case; now this is controlled + by the `BackslashEscaper` (and the default is now to leave the + capitalization alone). + + Change: EscapeDefault, EscapeDefaultNonHTMLSafe: No longer force long Unicode `\uXXXX` sequences for the U+FFFD Unicode replacement character. diff --git a/compat/json/compat.go b/compat/json/compat.go index d33f278..3a9bd6c 100644 --- a/compat/json/compat.go +++ b/compat/json/compat.go @@ -157,7 +157,7 @@ func HTMLEscape(dst *bytes.Buffer, src []byte) { case lowmemjson.BackslashEscapeNone: dst.WriteRune(c) case lowmemjson.BackslashEscapeUnicode: - _ = jsonstring.WriteStringUnicodeEscape(dst, c) + _ = jsonstring.WriteStringUnicodeEscape(dst, c, mode) default: panic(fmt.Errorf("lowmemjson.EscapeHTMLSafe returned an unexpected escape mode=%d", mode)) } diff --git a/compat/json/compat_test.go b/compat/json/compat_test.go index c83ca7e..29a8b37 100644 --- a/compat/json/compat_test.go +++ b/compat/json/compat_test.go @@ -18,7 +18,10 @@ func TestCompatHTMLEscape(t *testing.T) { Out string } testcases := map[string]testcase{ - "invalid": {In: `x`, Out: `x`}, + "invalid": {In: `x`, Out: `x`}, + "hex-lower": {In: `"\uabcd"`, Out: `"\uabcd"`}, + "hex-upper": {In: `"\uABCD"`, Out: `"\uABCD"`}, + "hex-mixed": {In: `"\uAbCd"`, Out: `"\uAbCd"`}, } for tcName, tc := range testcases { tc := tc @@ -39,11 +42,14 @@ func TestCompatValid(t *testing.T) { Exp bool } testcases := map[string]testcase{ - "empty": {In: ``, Exp: false}, - "num": {In: `1`, Exp: true}, - "trunc": {In: `{`, Exp: false}, - "object": {In: `{}`, Exp: true}, - "non-utf8": {In: "\"\x85\xcd\"", Exp: false}, // https://github.com/golang/go/issues/58517 + "empty": {In: ``, Exp: false}, + "num": {In: `1`, Exp: true}, + "trunc": {In: `{`, Exp: false}, + "object": {In: `{}`, Exp: true}, + "non-utf8": {In: "\"\x85\xcd\"", Exp: false}, // https://github.com/golang/go/issues/58517 + "hex-lower": {In: `"\uabcd"`, Exp: true}, + "hex-upper": {In: `"\uABCD"`, Exp: true}, + "hex-mixed": {In: `"\uAbCd"`, Exp: true}, } for tcName, tc := range testcases { tc := tc @@ -64,10 +70,13 @@ func TestCompatCompact(t *testing.T) { Err string } testcases := map[string]testcase{ - "trunc": {In: `{`, Out: ``, Err: `unexpected end of JSON input`}, - "object": {In: `{}`, Out: `{}`}, - "non-utf8": {In: "\"\x85\xcd\"", Out: "\"\x85\xcd\""}, - "float": {In: `1.200e003`, Out: `1.200e003`}, + "trunc": {In: `{`, Out: ``, Err: `unexpected end of JSON input`}, + "object": {In: `{}`, Out: `{}`}, + "non-utf8": {In: "\"\x85\xcd\"", Out: "\"\x85\xcd\""}, + "float": {In: `1.200e003`, Out: `1.200e003`}, + "hex-lower": {In: `"\uabcd"`, Out: `"\uabcd"`}, + "hex-upper": {In: `"\uABCD"`, Out: `"\uABCD"`}, + "hex-mixed": {In: `"\uAbCd"`, Out: `"\uAbCd"`}, } for tcName, tc := range testcases { tc := tc @@ -94,17 +103,20 @@ func TestCompatIndent(t *testing.T) { Err string } testcases := map[string]testcase{ - "trunc": {In: `{`, Out: ``, Err: `unexpected end of JSON input`}, - "object": {In: `{}`, Out: `{}`}, - "non-utf8": {In: "\"\x85\xcd\"", Out: "\"\x85\xcd\""}, - "float": {In: `1.200e003`, Out: `1.200e003`}, - "tailws0": {In: `0`, Out: `0`}, - "tailws1": {In: `0 `, Out: `0 `}, - "tailws2": {In: `0 `, Out: `0 `}, - "tailws3": {In: "0\n", Out: "0\n"}, - "headws1": {In: ` 0`, Out: `0`}, - "objws1": {In: `{"a" : 1}`, Out: "{\n>.\"a\": 1\n>}"}, - "objws2": {In: "{\"a\"\n:\n1}", Out: "{\n>.\"a\": 1\n>}"}, + "trunc": {In: `{`, Out: ``, Err: `unexpected end of JSON input`}, + "object": {In: `{}`, Out: `{}`}, + "non-utf8": {In: "\"\x85\xcd\"", Out: "\"\x85\xcd\""}, + "float": {In: `1.200e003`, Out: `1.200e003`}, + "tailws0": {In: `0`, Out: `0`}, + "tailws1": {In: `0 `, Out: `0 `}, + "tailws2": {In: `0 `, Out: `0 `}, + "tailws3": {In: "0\n", Out: "0\n"}, + "headws1": {In: ` 0`, Out: `0`}, + "objws1": {In: `{"a" : 1}`, Out: "{\n>.\"a\": 1\n>}"}, + "objws2": {In: "{\"a\"\n:\n1}", Out: "{\n>.\"a\": 1\n>}"}, + "hex-lower": {In: `"\uabcd"`, Out: `"\uabcd"`}, + "hex-upper": {In: `"\uABCD"`, Out: `"\uABCD"`}, + "hex-mixed": {In: `"\uAbCd"`, Out: `"\uAbCd"`}, } for tcName, tc := range testcases { tc := tc diff --git a/encode_escape.go b/encode_escape.go index c9e2bc9..664c762 100644 --- a/encode_escape.go +++ b/encode_escape.go @@ -36,7 +36,8 @@ const ( // - as a short "well-known" `\X` backslash sequence (where `X` is a // single-character) // -// - as a long Unicode `\uXXXX` backslash sequence +// - as a long Unicode `\uXXXX` backslash sequence (with 16 +// permutations of capitalization) // // - as a raw byte; this allows you to emit invalid JSON; JSON must // be valid UTF-8, but this allows you to emit arbitrary binary @@ -47,8 +48,29 @@ type BackslashEscapeMode = jsonstring.BackslashEscapeMode const ( BackslashEscapeNone = jsonstring.BackslashEscapeNone BackslashEscapeShort = jsonstring.BackslashEscapeShort - BackslashEscapeUnicode = jsonstring.BackslashEscapeUnicode BackslashEscapeRawByte = jsonstring.BackslashEscapeRawByte + + BackslashEscapeUnicodeXXXX = jsonstring.BackslashEscapeUnicodeXXXX + BackslashEscapeUnicodeXXXx = jsonstring.BackslashEscapeUnicodeXXXx + BackslashEscapeUnicodeXXxX = jsonstring.BackslashEscapeUnicodeXXxX + BackslashEscapeUnicodeXXxx = jsonstring.BackslashEscapeUnicodeXXxx + BackslashEscapeUnicodeXxXX = jsonstring.BackslashEscapeUnicodeXxXX + BackslashEscapeUnicodeXxXx = jsonstring.BackslashEscapeUnicodeXxXx + BackslashEscapeUnicodeXxxX = jsonstring.BackslashEscapeUnicodeXxxX + BackslashEscapeUnicodeXxxx = jsonstring.BackslashEscapeUnicodeXxxx + BackslashEscapeUnicodexXXX = jsonstring.BackslashEscapeUnicodexXXX + BackslashEscapeUnicodexXXx = jsonstring.BackslashEscapeUnicodexXXx + BackslashEscapeUnicodexXxX = jsonstring.BackslashEscapeUnicodexXxX + BackslashEscapeUnicodexXxx = jsonstring.BackslashEscapeUnicodexXxx + BackslashEscapeUnicodexxXX = jsonstring.BackslashEscapeUnicodexxXX + BackslashEscapeUnicodexxXx = jsonstring.BackslashEscapeUnicodexxXx + BackslashEscapeUnicodexxxX = jsonstring.BackslashEscapeUnicodexxxX + BackslashEscapeUnicodexxxx = jsonstring.BackslashEscapeUnicodexxxx + + BackslashEscapeUnicodeMin = jsonstring.BackslashEscapeUnicodeMin + BackslashEscapeUnicodeMax = jsonstring.BackslashEscapeUnicodeMax + + BackslashEscapeUnicode = jsonstring.BackslashEscapeUnicode // back-compat ) func hexToInt(c byte) rune { @@ -72,13 +94,24 @@ func hexToRune(a, b, c, d byte) rune { hexToInt(d)<<0 } +func hexToMode(a, b, c, d byte) BackslashEscapeMode { + // The 0b0010_0000 bit is the ASCII "lowercase bit". + return BackslashEscapeUnicodeMin + BackslashEscapeMode(0| + ((a&0b0010_0000)>>2)| + ((b&0b0010_0000)>>3)| + ((c&0b0010_0000)>>4)| + ((d&0b0010_0000)>>5)) +} + // A BackslashEscaper controls how a ReEncoder emits a character in a // JSON string. The `rune` argument is the character being // considered, and the `BackslashEscapeMode` argument is how it was // originally encoded in the input. // // The ReEncoder will panic if a BackslashEscaper returns an unknown -// BackslashEscapeMode. +// BackslashEscapeMode. However, a BackslashEscaper should be +// permissive of BackslashEscapeModes it doesn't recognize; it is safe +// to just return them unmodified. type BackslashEscaper = func(rune, BackslashEscapeMode) BackslashEscapeMode // EscapePreserve is a BackslashEscaper that preserves the original diff --git a/internal/jsonstring/encode_string.go b/internal/jsonstring/encode_string.go index 2488cb2..1416b3e 100644 --- a/internal/jsonstring/encode_string.go +++ b/internal/jsonstring/encode_string.go @@ -31,22 +31,49 @@ type BackslashEscapeMode uint8 const ( BackslashEscapeNone BackslashEscapeMode = iota BackslashEscapeShort - BackslashEscapeUnicode BackslashEscapeRawByte + + // It is significant to the implementation that if X=binary-0 + // and x=binary-1, then these "BackslashEscapeUnicode" + // constants are counting in-order from 0 to 15. + + BackslashEscapeUnicodeXXXX + BackslashEscapeUnicodeXXXx + BackslashEscapeUnicodeXXxX + BackslashEscapeUnicodeXXxx + BackslashEscapeUnicodeXxXX + BackslashEscapeUnicodeXxXx + BackslashEscapeUnicodeXxxX + BackslashEscapeUnicodeXxxx + BackslashEscapeUnicodexXXX + BackslashEscapeUnicodexXXx + BackslashEscapeUnicodexXxX + BackslashEscapeUnicodexXxx + BackslashEscapeUnicodexxXX + BackslashEscapeUnicodexxXx + BackslashEscapeUnicodexxxX + BackslashEscapeUnicodexxxx + + BackslashEscapeUnicodeMin = BackslashEscapeUnicodeXXXX + BackslashEscapeUnicodeMax = BackslashEscapeUnicodexxxx + + BackslashEscapeUnicode = BackslashEscapeUnicodexxxx // back-compat ) // BackslashEscaper is describe in the main lowmemjson package docs. type BackslashEscaper = func(rune, BackslashEscapeMode) BackslashEscapeMode -func WriteStringUnicodeEscape(w io.Writer, c rune) error { - const alphabet = "0123456789abcdef" +func WriteStringUnicodeEscape(w io.Writer, c rune, mode BackslashEscapeMode) error { + const alphabet = "0123456789ABCDEF" + _mode := byte(mode - BackslashEscapeUnicodeMin) buf := [6]byte{ '\\', 'u', - alphabet[(c>>12)&0xf], - alphabet[(c>>8)&0xf], - alphabet[(c>>4)&0xf], - alphabet[(c>>0)&0xf], + // The 0b0010_0000 bit is the ASCII "lowercase bit". + alphabet[(c>>12)&0xf] | ((_mode << 2) & 0b0010_0000), + alphabet[(c>>8)&0xf] | ((_mode << 3) & 0b0010_0000), + alphabet[(c>>4)&0xf] | ((_mode << 4) & 0b0010_0000), + alphabet[(c>>0)&0xf] | ((_mode << 5) & 0b0010_0000), } _, err := noescape.Write(w, buf[:]) return err @@ -84,7 +111,7 @@ func WriteStringChar(w fastio.AllWriter, c rune, escape BackslashEscapeMode) err case '\b', '\f', '\n', '\r', '\t': // short-escape if possible return writeStringShortEscape(w, c) default: - return WriteStringUnicodeEscape(w, c) + return WriteStringUnicodeEscape(w, c, BackslashEscapeUnicode) } case c == '"' || c == '\\': // override, gotta escape these return writeStringShortEscape(w, c) @@ -100,14 +127,6 @@ func WriteStringChar(w fastio.AllWriter, c rune, escape BackslashEscapeMode) err _, err := w.WriteRune(c) return err } - case BackslashEscapeUnicode: - switch { - case c > 0xFFFF: // override, can't escape these (TODO: unless we use UTF-16 surrogates?) - _, err := w.WriteRune(c) - return err - default: // obey - return WriteStringUnicodeEscape(w, c) - } case BackslashEscapeRawByte: switch { case c < utf8.RuneSelf: @@ -118,6 +137,15 @@ func WriteStringChar(w fastio.AllWriter, c rune, escape BackslashEscapeMode) err return w.WriteByte(byte(c)) } default: + if BackslashEscapeUnicodeMin <= escape && escape <= BackslashEscapeUnicodeMax { + switch { + case c > 0xFFFF: // override, can't escape these (TODO: unless we use UTF-16 surrogates?) + _, err := w.WriteRune(c) + return err + default: // obey + return WriteStringUnicodeEscape(w, c, escape) + } + } panic(fmt.Errorf("escaper returned an invalid escape mode=%d", escape)) } } diff --git a/reencode.go b/reencode.go index 1943b9c..7439bf0 100644 --- a/reencode.go +++ b/reencode.go @@ -441,8 +441,9 @@ func (enc *ReEncoder) handleRuneType(c rune, t jsonparse.RuneType, stackSize int enc.uhex[2] = byte(c) return nil case jsonparse.RuneTypeStringEscUD: + mode := hexToMode(enc.uhex[0], enc.uhex[1], enc.uhex[2], byte(c)) c = hexToRune(enc.uhex[0], enc.uhex[1], enc.uhex[2], byte(c)) - return enc.out.HandleRune(c, jsonparse.RuneTypeStringChar, BackslashEscapeUnicode, stackSize) + return enc.out.HandleRune(c, jsonparse.RuneTypeStringChar, mode, stackSize) case jsonparse.RuneTypeError: panic(fmt.Errorf("should not happen: handleRune called with %#v", t)) default: |