1 // Copyright 2011 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 http 6 7 import ( 8 "bytes" 9 "encoding/binary" 10 ) 11 12 // The algorithm uses at most sniffLen bytes to make its decision. 13 const sniffLen = 512 14 15 // DetectContentType implements the algorithm described 16 // at http://mimesniff.spec.whatwg.org/ to determine the 17 // Content-Type of the given data. It considers at most the 18 // first 512 bytes of data. DetectContentType always returns 19 // a valid MIME type: if it cannot determine a more specific one, it 20 // returns "application/octet-stream". 21 func DetectContentType(data []byte) string { 22 if len(data) > sniffLen { 23 data = data[:sniffLen] 24 } 25 26 // Index of the first non-whitespace byte in data. 27 firstNonWS := 0 28 for ; firstNonWS < len(data) && isWS(data[firstNonWS]); firstNonWS++ { 29 } 30 31 for _, sig := range sniffSignatures { 32 if ct := sig.match(data, firstNonWS); ct != "" { 33 return ct 34 } 35 } 36 37 return "application/octet-stream" // fallback 38 } 39 40 func isWS(b byte) bool { 41 switch b { 42 case '\t', '\n', '\x0c', '\r', ' ': 43 return true 44 } 45 return false 46 } 47 48 type sniffSig interface { 49 // match returns the MIME type of the data, or "" if unknown. 50 match(data []byte, firstNonWS int) string 51 } 52 53 // Data matching the table in section 6. 54 var sniffSignatures = []sniffSig{ 55 htmlSig("<!DOCTYPE HTML"), 56 htmlSig("<HTML"), 57 htmlSig("<HEAD"), 58 htmlSig("<SCRIPT"), 59 htmlSig("<IFRAME"), 60 htmlSig("<H1"), 61 htmlSig("<DIV"), 62 htmlSig("<FONT"), 63 htmlSig("<TABLE"), 64 htmlSig("<A"), 65 htmlSig("<STYLE"), 66 htmlSig("<TITLE"), 67 htmlSig("<B"), 68 htmlSig("<BODY"), 69 htmlSig("<BR"), 70 htmlSig("<P"), 71 htmlSig("<!--"), 72 73 &maskedSig{mask: []byte("\xFF\xFF\xFF\xFF\xFF"), pat: []byte("<?xml"), skipWS: true, ct: "text/xml; charset=utf-8"}, 74 75 &exactSig{[]byte("%PDF-"), "application/pdf"}, 76 &exactSig{[]byte("%!PS-Adobe-"), "application/postscript"}, 77 78 // UTF BOMs. 79 &maskedSig{mask: []byte("\xFF\xFF\x00\x00"), pat: []byte("\xFE\xFF\x00\x00"), ct: "text/plain; charset=utf-16be"}, 80 &maskedSig{mask: []byte("\xFF\xFF\x00\x00"), pat: []byte("\xFF\xFE\x00\x00"), ct: "text/plain; charset=utf-16le"}, 81 &maskedSig{mask: []byte("\xFF\xFF\xFF\x00"), pat: []byte("\xEF\xBB\xBF\x00"), ct: "text/plain; charset=utf-8"}, 82 83 &exactSig{[]byte("GIF87a"), "image/gif"}, 84 &exactSig{[]byte("GIF89a"), "image/gif"}, 85 &exactSig{[]byte("\x89\x50\x4E\x47\x0D\x0A\x1A\x0A"), "image/png"}, 86 &exactSig{[]byte("\xFF\xD8\xFF"), "image/jpeg"}, 87 &exactSig{[]byte("BM"), "image/bmp"}, 88 &maskedSig{ 89 mask: []byte("\xFF\xFF\xFF\xFF\x00\x00\x00\x00\xFF\xFF\xFF\xFF\xFF\xFF"), 90 pat: []byte("RIFF\x00\x00\x00\x00WEBPVP"), 91 ct: "image/webp", 92 }, 93 &exactSig{[]byte("\x00\x00\x01\x00"), "image/vnd.microsoft.icon"}, 94 &exactSig{[]byte("\x4F\x67\x67\x53\x00"), "application/ogg"}, 95 &maskedSig{ 96 mask: []byte("\xFF\xFF\xFF\xFF\x00\x00\x00\x00\xFF\xFF\xFF\xFF"), 97 pat: []byte("RIFF\x00\x00\x00\x00WAVE"), 98 ct: "audio/wave", 99 }, 100 &exactSig{[]byte("\x1A\x45\xDF\xA3"), "video/webm"}, 101 &exactSig{[]byte("\x52\x61\x72\x20\x1A\x07\x00"), "application/x-rar-compressed"}, 102 &exactSig{[]byte("\x50\x4B\x03\x04"), "application/zip"}, 103 &exactSig{[]byte("\x1F\x8B\x08"), "application/x-gzip"}, 104 105 // TODO(dsymonds): Re-enable this when the spec is sorted w.r.t. MP4. 106 //mp4Sig(0), 107 108 textSig(0), // should be last 109 } 110 111 type exactSig struct { 112 sig []byte 113 ct string 114 } 115 116 func (e *exactSig) match(data []byte, firstNonWS int) string { 117 if bytes.HasPrefix(data, e.sig) { 118 return e.ct 119 } 120 return "" 121 } 122 123 type maskedSig struct { 124 mask, pat []byte 125 skipWS bool 126 ct string 127 } 128 129 func (m *maskedSig) match(data []byte, firstNonWS int) string { 130 if m.skipWS { 131 data = data[firstNonWS:] 132 } 133 if len(data) < len(m.mask) { 134 return "" 135 } 136 for i, mask := range m.mask { 137 db := data[i] & mask 138 if db != m.pat[i] { 139 return "" 140 } 141 } 142 return m.ct 143 } 144 145 type htmlSig []byte 146 147 func (h htmlSig) match(data []byte, firstNonWS int) string { 148 data = data[firstNonWS:] 149 if len(data) < len(h)+1 { 150 return "" 151 } 152 for i, b := range h { 153 db := data[i] 154 if 'A' <= b && b <= 'Z' { 155 db &= 0xDF 156 } 157 if b != db { 158 return "" 159 } 160 } 161 // Next byte must be space or right angle bracket. 162 if db := data[len(h)]; db != ' ' && db != '>' { 163 return "" 164 } 165 return "text/html; charset=utf-8" 166 } 167 168 var mp4ftype = []byte("ftyp") 169 170 type mp4Sig int 171 172 func (mp4Sig) match(data []byte, firstNonWS int) string { 173 // c.f. section 6.1. 174 if len(data) < 8 { 175 return "" 176 } 177 boxSize := int(binary.BigEndian.Uint32(data[:4])) 178 if boxSize%4 != 0 || len(data) < boxSize { 179 return "" 180 } 181 if !bytes.Equal(data[4:8], mp4ftype) { 182 return "" 183 } 184 for st := 8; st < boxSize; st += 4 { 185 if st == 12 { 186 // minor version number 187 continue 188 } 189 seg := string(data[st : st+3]) 190 switch seg { 191 case "mp4", "iso", "M4V", "M4P", "M4B": 192 return "video/mp4" 193 /* The remainder are not in the spec. 194 case "M4A": 195 return "audio/mp4" 196 case "3gp": 197 return "video/3gpp" 198 case "jp2": 199 return "image/jp2" // JPEG 2000 200 */ 201 } 202 } 203 return "" 204 } 205 206 type textSig int 207 208 func (textSig) match(data []byte, firstNonWS int) string { 209 // c.f. section 5, step 4. 210 for _, b := range data[firstNonWS:] { 211 switch { 212 case 0x00 <= b && b <= 0x08, 213 b == 0x0B, 214 0x0E <= b && b <= 0x1A, 215 0x1C <= b && b <= 0x1F: 216 return "" 217 } 218 } 219 return "text/plain; charset=utf-8" 220 } 221
