1// Copyright 2009 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
5// Package base64 implements base64 encoding as specified by RFC 4648.
6package base64
7
8import (
9	"bytes"
10	"io"
11	"strconv"
12	"strings"
13)
14
15/*
16 * Encodings
17 */
18
19// An Encoding is a radix 64 encoding/decoding scheme, defined by a
20// 64-character alphabet.  The most common encoding is the "base64"
21// encoding defined in RFC 4648 and used in MIME (RFC 2045) and PEM
22// (RFC 1421).  RFC 4648 also defines an alternate encoding, which is
23// the standard encoding with - and _ substituted for + and /.
24type Encoding struct {
25	encode    string
26	decodeMap [256]byte
27}
28
29const encodeStd = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"
30const encodeURL = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_"
31
32// NewEncoding returns a new Encoding defined by the given alphabet,
33// which must be a 64-byte string.
34func NewEncoding(encoder string) *Encoding {
35	e := new(Encoding)
36	e.encode = encoder
37	for i := 0; i < len(e.decodeMap); i++ {
38		e.decodeMap[i] = 0xFF
39	}
40	for i := 0; i < len(encoder); i++ {
41		e.decodeMap[encoder[i]] = byte(i)
42	}
43	return e
44}
45
46// StdEncoding is the standard base64 encoding, as defined in
47// RFC 4648.
48var StdEncoding = NewEncoding(encodeStd)
49
50// URLEncoding is the alternate base64 encoding defined in RFC 4648.
51// It is typically used in URLs and file names.
52var URLEncoding = NewEncoding(encodeURL)
53
54var removeNewlinesMapper = func(r rune) rune {
55	if r == '\r' || r == '\n' {
56		return -1
57	}
58	return r
59}
60
61/*
62 * Encoder
63 */
64
65// Encode encodes src using the encoding enc, writing
66// EncodedLen(len(src)) bytes to dst.
67//
68// The encoding pads the output to a multiple of 4 bytes,
69// so Encode is not appropriate for use on individual blocks
70// of a large data stream.  Use NewEncoder() instead.
71func (enc *Encoding) Encode(dst, src []byte) {
72	if len(src) == 0 {
73		return
74	}
75
76	for len(src) > 0 {
77		dst[0] = 0
78		dst[1] = 0
79		dst[2] = 0
80		dst[3] = 0
81
82		// Unpack 4x 6-bit source blocks into a 4 byte
83		// destination quantum
84		switch len(src) {
85		default:
86			dst[3] |= src[2] & 0x3F
87			dst[2] |= src[2] >> 6
88			fallthrough
89		case 2:
90			dst[2] |= (src[1] << 2) & 0x3F
91			dst[1] |= src[1] >> 4
92			fallthrough
93		case 1:
94			dst[1] |= (src[0] << 4) & 0x3F
95			dst[0] |= src[0] >> 2
96		}
97
98		// Encode 6-bit blocks using the base64 alphabet
99		for j := 0; j < 4; j++ {
100			dst[j] = enc.encode[dst[j]]
101		}
102
103		// Pad the final quantum
104		if len(src) < 3 {
105			dst[3] = '='
106			if len(src) < 2 {
107				dst[2] = '='
108			}
109			break
110		}
111
112		src = src[3:]
113		dst = dst[4:]
114	}
115}
116
117// EncodeToString returns the base64 encoding of src.
118func (enc *Encoding) EncodeToString(src []byte) string {
119	buf := make([]byte, enc.EncodedLen(len(src)))
120	enc.Encode(buf, src)
121	return string(buf)
122}
123
124type encoder struct {
125	err  error
126	enc  *Encoding
127	w    io.Writer
128	buf  [3]byte    // buffered data waiting to be encoded
129	nbuf int        // number of bytes in buf
130	out  [1024]byte // output buffer
131}
132
133func (e *encoder) Write(p []byte) (n int, err error) {
134	if e.err != nil {
135		return 0, e.err
136	}
137
138	// Leading fringe.
139	if e.nbuf > 0 {
140		var i int
141		for i = 0; i < len(p) && e.nbuf < 3; i++ {
142			e.buf[e.nbuf] = p[i]
143			e.nbuf++
144		}
145		n += i
146		p = p[i:]
147		if e.nbuf < 3 {
148			return
149		}
150		e.enc.Encode(e.out[0:], e.buf[0:])
151		if _, e.err = e.w.Write(e.out[0:4]); e.err != nil {
152			return n, e.err
153		}
154		e.nbuf = 0
155	}
156
157	// Large interior chunks.
158	for len(p) >= 3 {
159		nn := len(e.out) / 4 * 3
160		if nn > len(p) {
161			nn = len(p)
162		}
163		nn -= nn % 3
164		if nn > 0 {
165			e.enc.Encode(e.out[0:], p[0:nn])
166			if _, e.err = e.w.Write(e.out[0 : nn/3*4]); e.err != nil {
167				return n, e.err
168			}
169		}
170		n += nn
171		p = p[nn:]
172	}
173
174	// Trailing fringe.
175	for i := 0; i < len(p); i++ {
176		e.buf[i] = p[i]
177	}
178	e.nbuf = len(p)
179	n += len(p)
180	return
181}
182
183// Close flushes any pending output from the encoder.
184// It is an error to call Write after calling Close.
185func (e *encoder) Close() error {
186	// If there's anything left in the buffer, flush it out
187	if e.err == nil && e.nbuf > 0 {
188		e.enc.Encode(e.out[0:], e.buf[0:e.nbuf])
189		e.nbuf = 0
190		_, e.err = e.w.Write(e.out[0:4])
191	}
192	return e.err
193}
194
195// NewEncoder returns a new base64 stream encoder.  Data written to
196// the returned writer will be encoded using enc and then written to w.
197// Base64 encodings operate in 4-byte blocks; when finished
198// writing, the caller must Close the returned encoder to flush any
199// partially written blocks.
200func NewEncoder(enc *Encoding, w io.Writer) io.WriteCloser {
201	return &encoder{enc: enc, w: w}
202}
203
204// EncodedLen returns the length in bytes of the base64 encoding
205// of an input buffer of length n.
206func (enc *Encoding) EncodedLen(n int) int { return (n + 2) / 3 * 4 }
207
208/*
209 * Decoder
210 */
211
212type CorruptInputError int64
213
214func (e CorruptInputError) Error() string {
215	return "illegal base64 data at input byte " + strconv.FormatInt(int64(e), 10)
216}
217
218// decode is like Decode but returns an additional 'end' value, which
219// indicates if end-of-message padding was encountered and thus any
220// additional data is an error. This method assumes that src has been
221// stripped of all supported whitespace ('\r' and '\n').
222func (enc *Encoding) decode(dst, src []byte) (n int, end bool, err error) {
223	olen := len(src)
224	for len(src) > 0 && !end {
225		// Decode quantum using the base64 alphabet
226		var dbuf [4]byte
227		dlen := 4
228
229		for j := 0; j < 4; {
230			if len(src) == 0 {
231				return n, false, CorruptInputError(olen - len(src) - j)
232			}
233			in := src[0]
234			src = src[1:]
235			if in == '=' && j >= 2 && len(src) < 4 {
236				// We've reached the end and there's padding
237				if len(src)+j < 4-1 {
238					// not enough padding
239					return n, false, CorruptInputError(olen)
240				}
241				if len(src) > 0 && src[0] != '=' {
242					// incorrect padding
243					return n, false, CorruptInputError(olen - len(src) - 1)
244				}
245				dlen, end = j, true
246				break
247			}
248			dbuf[j] = enc.decodeMap[in]
249			if dbuf[j] == 0xFF {
250				return n, false, CorruptInputError(olen - len(src) - 1)
251			}
252			j++
253		}
254
255		// Pack 4x 6-bit source blocks into 3 byte destination
256		// quantum
257		switch dlen {
258		case 4:
259			dst[2] = dbuf[2]<<6 | dbuf[3]
260			fallthrough
261		case 3:
262			dst[1] = dbuf[1]<<4 | dbuf[2]>>2
263			fallthrough
264		case 2:
265			dst[0] = dbuf[0]<<2 | dbuf[1]>>4
266		}
267		dst = dst[3:]
268		n += dlen - 1
269	}
270
271	return n, end, nil
272}
273
274// Decode decodes src using the encoding enc.  It writes at most
275// DecodedLen(len(src)) bytes to dst and returns the number of bytes
276// written.  If src contains invalid base64 data, it will return the
277// number of bytes successfully written and CorruptInputError.
278// New line characters (\r and \n) are ignored.
279func (enc *Encoding) Decode(dst, src []byte) (n int, err error) {
280	src = bytes.Map(removeNewlinesMapper, src)
281	n, _, err = enc.decode(dst, src)
282	return
283}
284
285// DecodeString returns the bytes represented by the base64 string s.
286func (enc *Encoding) DecodeString(s string) ([]byte, error) {
287	s = strings.Map(removeNewlinesMapper, s)
288	dbuf := make([]byte, enc.DecodedLen(len(s)))
289	n, err := enc.Decode(dbuf, []byte(s))
290	return dbuf[:n], err
291}
292
293type decoder struct {
294	err    error
295	enc    *Encoding
296	r      io.Reader
297	end    bool       // saw end of message
298	buf    [1024]byte // leftover input
299	nbuf   int
300	out    []byte // leftover decoded output
301	outbuf [1024 / 4 * 3]byte
302}
303
304func (d *decoder) Read(p []byte) (n int, err error) {
305	if d.err != nil {
306		return 0, d.err
307	}
308
309	// Use leftover decoded output from last read.
310	if len(d.out) > 0 {
311		n = copy(p, d.out)
312		d.out = d.out[n:]
313		return n, nil
314	}
315
316	// Read a chunk.
317	nn := len(p) / 3 * 4
318	if nn < 4 {
319		nn = 4
320	}
321	if nn > len(d.buf) {
322		nn = len(d.buf)
323	}
324	nn, d.err = io.ReadAtLeast(d.r, d.buf[d.nbuf:nn], 4-d.nbuf)
325	d.nbuf += nn
326	if d.err != nil || d.nbuf < 4 {
327		return 0, d.err
328	}
329
330	// Decode chunk into p, or d.out and then p if p is too small.
331	nr := d.nbuf / 4 * 4
332	nw := d.nbuf / 4 * 3
333	if nw > len(p) {
334		nw, d.end, d.err = d.enc.decode(d.outbuf[0:], d.buf[0:nr])
335		d.out = d.outbuf[0:nw]
336		n = copy(p, d.out)
337		d.out = d.out[n:]
338	} else {
339		n, d.end, d.err = d.enc.decode(p, d.buf[0:nr])
340	}
341	d.nbuf -= nr
342	for i := 0; i < d.nbuf; i++ {
343		d.buf[i] = d.buf[i+nr]
344	}
345
346	if d.err == nil {
347		d.err = err
348	}
349	return n, d.err
350}
351
352type newlineFilteringReader struct {
353	wrapped io.Reader
354}
355
356func (r *newlineFilteringReader) Read(p []byte) (int, error) {
357	n, err := r.wrapped.Read(p)
358	for n > 0 {
359		offset := 0
360		for i, b := range p[0:n] {
361			if b != '\r' && b != '\n' {
362				if i != offset {
363					p[offset] = b
364				}
365				offset++
366			}
367		}
368		if offset > 0 {
369			return offset, err
370		}
371		// Previous buffer entirely whitespace, read again
372		n, err = r.wrapped.Read(p)
373	}
374	return n, err
375}
376
377// NewDecoder constructs a new base64 stream decoder.
378func NewDecoder(enc *Encoding, r io.Reader) io.Reader {
379	return &decoder{enc: enc, r: &newlineFilteringReader{r}}
380}
381
382// DecodedLen returns the maximum length in bytes of the decoded data
383// corresponding to n bytes of base64-encoded data.
384func (enc *Encoding) DecodedLen(n int) int { return n / 4 * 3 }
385