1// Copyright 2010 The Go Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style
3// license that can be found in the LICENSE file.
4
5package json
6
7import "bytes"
8
9// HTMLEscape appends to dst the JSON-encoded src with <, >, &, U+2028 and U+2029
10// characters inside string literals changed to \u003c, \u003e, \u0026, \u2028, \u2029
11// so that the JSON will be safe to embed inside HTML <script> tags.
12// For historical reasons, web browsers don't honor standard HTML
13// escaping within <script> tags, so an alternative JSON encoding must be used.
14func HTMLEscape(dst *bytes.Buffer, src []byte) {
15	dst.Grow(len(src))
16	dst.Write(appendHTMLEscape(dst.AvailableBuffer(), src))
17}
18
19func appendHTMLEscape(dst, src []byte) []byte {
20	// The characters can only appear in string literals,
21	// so just scan the string one byte at a time.
22	start := 0
23	for i, c := range src {
24		if c == '<' || c == '>' || c == '&' {
25			dst = append(dst, src[start:i]...)
26			dst = append(dst, '\\', 'u', '0', '0', hex[c>>4], hex[c&0xF])
27			start = i + 1
28		}
29		// Convert U+2028 and U+2029 (E2 80 A8 and E2 80 A9).
30		if c == 0xE2 && i+2 < len(src) && src[i+1] == 0x80 && src[i+2]&^1 == 0xA8 {
31			dst = append(dst, src[start:i]...)
32			dst = append(dst, '\\', 'u', '2', '0', '2', hex[src[i+2]&0xF])
33			start = i + len("\u2029")
34		}
35	}
36	return append(dst, src[start:]...)
37}
38
39// Compact appends to dst the JSON-encoded src with
40// insignificant space characters elided.
41func Compact(dst *bytes.Buffer, src []byte) error {
42	dst.Grow(len(src))
43	b := dst.AvailableBuffer()
44	b, err := appendCompact(b, src, false)
45	dst.Write(b)
46	return err
47}
48
49func appendCompact(dst, src []byte, escape bool) ([]byte, error) {
50	origLen := len(dst)
51	scan := newScanner()
52	defer freeScanner(scan)
53	start := 0
54	for i, c := range src {
55		if escape && (c == '<' || c == '>' || c == '&') {
56			if start < i {
57				dst = append(dst, src[start:i]...)
58			}
59			dst = append(dst, '\\', 'u', '0', '0', hex[c>>4], hex[c&0xF])
60			start = i + 1
61		}
62		// Convert U+2028 and U+2029 (E2 80 A8 and E2 80 A9).
63		if escape && c == 0xE2 && i+2 < len(src) && src[i+1] == 0x80 && src[i+2]&^1 == 0xA8 {
64			if start < i {
65				dst = append(dst, src[start:i]...)
66			}
67			dst = append(dst, '\\', 'u', '2', '0', '2', hex[src[i+2]&0xF])
68			start = i + 3
69		}
70		v := scan.step(scan, c)
71		if v >= scanSkipSpace {
72			if v == scanError {
73				break
74			}
75			if start < i {
76				dst = append(dst, src[start:i]...)
77			}
78			start = i + 1
79		}
80	}
81	if scan.eof() == scanError {
82		return dst[:origLen], scan.err
83	}
84	if start < len(src) {
85		dst = append(dst, src[start:]...)
86	}
87	return dst, nil
88}
89
90func appendNewline(dst []byte, prefix, indent string, depth int) []byte {
91	dst = append(dst, '\n')
92	dst = append(dst, prefix...)
93	for i := 0; i < depth; i++ {
94		dst = append(dst, indent...)
95	}
96	return dst
97}
98
99// indentGrowthFactor specifies the growth factor of indenting JSON input.
100// Empirically, the growth factor was measured to be between 1.4x to 1.8x
101// for some set of compacted JSON with the indent being a single tab.
102// Specify a growth factor slightly larger than what is observed
103// to reduce probability of allocation in appendIndent.
104// A factor no higher than 2 ensures that wasted space never exceeds 50%.
105const indentGrowthFactor = 2
106
107// Indent appends to dst an indented form of the JSON-encoded src.
108// Each element in a JSON object or array begins on a new,
109// indented line beginning with prefix followed by one or more
110// copies of indent according to the indentation nesting.
111// The data appended to dst does not begin with the prefix nor
112// any indentation, to make it easier to embed inside other formatted JSON data.
113// Although leading space characters (space, tab, carriage return, newline)
114// at the beginning of src are dropped, trailing space characters
115// at the end of src are preserved and copied to dst.
116// For example, if src has no trailing spaces, neither will dst;
117// if src ends in a trailing newline, so will dst.
118func Indent(dst *bytes.Buffer, src []byte, prefix, indent string) error {
119	dst.Grow(indentGrowthFactor * len(src))
120	b := dst.AvailableBuffer()
121	b, err := appendIndent(b, src, prefix, indent)
122	dst.Write(b)
123	return err
124}
125
126func appendIndent(dst, src []byte, prefix, indent string) ([]byte, error) {
127	origLen := len(dst)
128	scan := newScanner()
129	defer freeScanner(scan)
130	needIndent := false
131	depth := 0
132	for _, c := range src {
133		scan.bytes++
134		v := scan.step(scan, c)
135		if v == scanSkipSpace {
136			continue
137		}
138		if v == scanError {
139			break
140		}
141		if needIndent && v != scanEndObject && v != scanEndArray {
142			needIndent = false
143			depth++
144			dst = appendNewline(dst, prefix, indent, depth)
145		}
146
147		// Emit semantically uninteresting bytes
148		// (in particular, punctuation in strings) unmodified.
149		if v == scanContinue {
150			dst = append(dst, c)
151			continue
152		}
153
154		// Add spacing around real punctuation.
155		switch c {
156		case '{', '[':
157			// delay indent so that empty object and array are formatted as {} and [].
158			needIndent = true
159			dst = append(dst, c)
160		case ',':
161			dst = append(dst, c)
162			dst = appendNewline(dst, prefix, indent, depth)
163		case ':':
164			dst = append(dst, c, ' ')
165		case '}', ']':
166			if needIndent {
167				// suppress indent in empty object/array
168				needIndent = false
169			} else {
170				depth--
171				dst = appendNewline(dst, prefix, indent, depth)
172			}
173			dst = append(dst, c)
174		default:
175			dst = append(dst, c)
176		}
177	}
178	if scan.eof() == scanError {
179		return dst[:origLen], scan.err
180	}
181	return dst, nil
182}
183