source file: /System/Library/Frameworks/Python.framework/Versions/2.3/lib/python2.3/email/Header.py
file stats: 243 lines, 44 executed: 18.1% covered
1. # Copyright (C) 2002 Python Software Foundation 2. # Author: che@debian.org (Ben Gertzfield), barry@zope.com (Barry Warsaw) 3. 4. """Header encoding and decoding functionality.""" 5. 6. import re 7. import binascii 8. from types import StringType, UnicodeType 9. 10. import email.quopriMIME 11. import email.base64MIME 12. from email.Errors import HeaderParseError 13. from email.Charset import Charset 14. 15. try: 16. from email._compat22 import _floordiv 17. except SyntaxError: 18. # Python 2.1 spells integer division differently 19. from email._compat21 import _floordiv 20. 21. try: 22. True, False 23. except NameError: 24. True = 1 25. False = 0 26. 27. CRLFSPACE = '\r\n ' 28. CRLF = '\r\n' 29. NL = '\n' 30. SPACE = ' ' 31. USPACE = u' ' 32. SPACE8 = ' ' * 8 33. EMPTYSTRING = '' 34. UEMPTYSTRING = u'' 35. 36. MAXLINELEN = 76 37. 38. ENCODE = 1 39. DECODE = 2 40. 41. USASCII = Charset('us-ascii') 42. UTF8 = Charset('utf-8') 43. 44. # Match encoded-word strings in the form =?charset?q?Hello_World?= 45. ecre = re.compile(r''' 46. =\? # literal =? 47. (?P<charset>[^?]*?) # non-greedy up to the next ? is the charset 48. \? # literal ? 49. (?P<encoding>[qb]) # either a "q" or a "b", case insensitive 50. \? # literal ? 51. (?P<encoded>.*?) # non-greedy up to the next ?= is the encoded string 52. \?= # literal ?= 53. ''', re.VERBOSE | re.IGNORECASE) 54. 55. pcre = re.compile('([,;])') 56. 57. # Field name regexp, including trailing colon, but not separating whitespace, 58. # according to RFC 2822. Character range is from tilde to exclamation mark. 59. # For use with .match() 60. fcre = re.compile(r'[\041-\176]+:$') 61. 62. 63. 64. # Helpers 65. _max_append = email.quopriMIME._max_append 66. 67. 68. 69. def decode_header(header): 70. """Decode a message header value without converting charset. 71. 72. Returns a list of (decoded_string, charset) pairs containing each of the 73. decoded parts of the header. Charset is None for non-encoded parts of the 74. header, otherwise a lower-case string containing the name of the character 75. set specified in the encoded string. 76. 77. An email.Errors.HeaderParseError may be raised when certain decoding error 78. occurs (e.g. a base64 decoding exception). 79. """ 80. # If no encoding, just return the header 81. header = str(header) 82. if not ecre.search(header): 83. return [(header, None)] 84. decoded = [] 85. dec = '' 86. for line in header.splitlines(): 87. # This line might not have an encoding in it 88. if not ecre.search(line): 89. decoded.append((line, None)) 90. continue 91. parts = ecre.split(line) 92. while parts: 93. unenc = parts.pop(0).strip() 94. if unenc: 95. # Should we continue a long line? 96. if decoded and decoded[-1][1] is None: 97. decoded[-1] = (decoded[-1][0] + SPACE + unenc, None) 98. else: 99. decoded.append((unenc, None)) 100. if parts: 101. charset, encoding = [s.lower() for s in parts[0:2]] 102. encoded = parts[2] 103. dec = None 104. if encoding == 'q': 105. dec = email.quopriMIME.header_decode(encoded) 106. elif encoding == 'b': 107. try: 108. dec = email.base64MIME.decode(encoded) 109. except binascii.Error: 110. # Turn this into a higher level exception. BAW: Right 111. # now we throw the lower level exception away but 112. # when/if we get exception chaining, we'll preserve it. 113. raise HeaderParseError 114. if dec is None: 115. dec = encoded 116. 117. if decoded and decoded[-1][1] == charset: 118. decoded[-1] = (decoded[-1][0] + dec, decoded[-1][1]) 119. else: 120. decoded.append((dec, charset)) 121. del parts[0:3] 122. return decoded 123. 124. 125. 126. def make_header(decoded_seq, maxlinelen=None, header_name=None, 127. continuation_ws=' '): 128. """Create a Header from a sequence of pairs as returned by decode_header() 129. 130. decode_header() takes a header value string and returns a sequence of 131. pairs of the format (decoded_string, charset) where charset is the string 132. name of the character set. 133. 134. This function takes one of those sequence of pairs and returns a Header 135. instance. Optional maxlinelen, header_name, and continuation_ws are as in 136. the Header constructor. 137. """ 138. h = Header(maxlinelen=maxlinelen, header_name=header_name, 139. continuation_ws=continuation_ws) 140. for s, charset in decoded_seq: 141. # None means us-ascii but we can simply pass it on to h.append() 142. if charset is not None and not isinstance(charset, Charset): 143. charset = Charset(charset) 144. h.append(s, charset) 145. return h 146. 147. 148. 149. class Header: 150. def __init__(self, s=None, charset=None, 151. maxlinelen=None, header_name=None, 152. continuation_ws=' ', errors='strict'): 153. """Create a MIME-compliant header that can contain many character sets. 154. 155. Optional s is the initial header value. If None, the initial header 156. value is not set. You can later append to the header with .append() 157. method calls. s may be a byte string or a Unicode string, but see the 158. .append() documentation for semantics. 159. 160. Optional charset serves two purposes: it has the same meaning as the 161. charset argument to the .append() method. It also sets the default 162. character set for all subsequent .append() calls that omit the charset 163. argument. If charset is not provided in the constructor, the us-ascii 164. charset is used both as s's initial charset and as the default for 165. subsequent .append() calls. 166. 167. The maximum line length can be specified explicit via maxlinelen. For 168. splitting the first line to a shorter value (to account for the field 169. header which isn't included in s, e.g. `Subject') pass in the name of 170. the field in header_name. The default maxlinelen is 76. 171. 172. continuation_ws must be RFC 2822 compliant folding whitespace (usually 173. either a space or a hard tab) which will be prepended to continuation 174. lines. 175. 176. errors is passed through to the .append() call. 177. """ 178. if charset is None: 179. charset = USASCII 180. if not isinstance(charset, Charset): 181. charset = Charset(charset) 182. self._charset = charset 183. self._continuation_ws = continuation_ws 184. cws_expanded_len = len(continuation_ws.replace('\t', SPACE8)) 185. # BAW: I believe `chunks' and `maxlinelen' should be non-public. 186. self._chunks = [] 187. if s is not None: 188. self.append(s, charset, errors) 189. if maxlinelen is None: 190. maxlinelen = MAXLINELEN 191. if header_name is None: 192. # We don't know anything about the field header so the first line 193. # is the same length as subsequent lines. 194. self._firstlinelen = maxlinelen 195. else: 196. # The first line should be shorter to take into account the field 197. # header. Also subtract off 2 extra for the colon and space. 198. self._firstlinelen = maxlinelen - len(header_name) - 2 199. # Second and subsequent lines should subtract off the length in 200. # columns of the continuation whitespace prefix. 201. self._maxlinelen = maxlinelen - cws_expanded_len 202. 203. def __str__(self): 204. """A synonym for self.encode().""" 205. return self.encode() 206. 207. def __unicode__(self): 208. """Helper for the built-in unicode function.""" 209. uchunks = [] 210. lastcs = None 211. for s, charset in self._chunks: 212. # We must preserve spaces between encoded and non-encoded word 213. # boundaries, which means for us we need to add a space when we go 214. # from a charset to None/us-ascii, or from None/us-ascii to a 215. # charset. Only do this for the second and subsequent chunks. 216. nextcs = charset 217. if uchunks: 218. if lastcs not in (None, 'us-ascii'): 219. if nextcs in (None, 'us-ascii'): 220. uchunks.append(USPACE) 221. nextcs = None 222. elif nextcs not in (None, 'us-ascii'): 223. uchunks.append(USPACE) 224. lastcs = nextcs 225. uchunks.append(unicode(s, str(charset))) 226. return UEMPTYSTRING.join(uchunks) 227. 228. # Rich comparison operators for equality only. BAW: does it make sense to 229. # have or explicitly disable <, <=, >, >= operators? 230. def __eq__(self, other): 231. # other may be a Header or a string. Both are fine so coerce 232. # ourselves to a string, swap the args and do another comparison. 233. return other == self.encode() 234. 235. def __ne__(self, other): 236. return not self == other 237. 238. def append(self, s, charset=None, errors='strict'): 239. """Append a string to the MIME header. 240. 241. Optional charset, if given, should be a Charset instance or the name 242. of a character set (which will be converted to a Charset instance). A 243. value of None (the default) means that the charset given in the 244. constructor is used. 245. 246. s may be a byte string or a Unicode string. If it is a byte string 247. (i.e. isinstance(s, StringType) is true), then charset is the encoding 248. of that byte string, and a UnicodeError will be raised if the string 249. cannot be decoded with that charset. If s is a Unicode string, then 250. charset is a hint specifying the character set of the characters in 251. the string. In this case, when producing an RFC 2822 compliant header 252. using RFC 2047 rules, the Unicode string will be encoded using the 253. following charsets in order: us-ascii, the charset hint, utf-8. The 254. first character set not to provoke a UnicodeError is used. 255. 256. Optional `errors' is passed as the third argument to any unicode() or 257. ustr.encode() call. 258. """ 259. if charset is None: 260. charset = self._charset 261. elif not isinstance(charset, Charset): 262. charset = Charset(charset) 263. # If the charset is our faux 8bit charset, leave the string unchanged 264. if charset <> '8bit': 265. # We need to test that the string can be converted to unicode and 266. # back to a byte string, given the input and output codecs of the 267. # charset. 268. if isinstance(s, StringType): 269. # Possibly raise UnicodeError if the byte string can't be 270. # converted to a unicode with the input codec of the charset. 271. incodec = charset.input_codec or 'us-ascii' 272. ustr = unicode(s, incodec, errors) 273. # Now make sure that the unicode could be converted back to a 274. # byte string with the output codec, which may be different 275. # than the iput coded. Still, use the original byte string. 276. outcodec = charset.output_codec or 'us-ascii' 277. ustr.encode(outcodec, errors) 278. elif isinstance(s, UnicodeType): 279. # Now we have to be sure the unicode string can be converted 280. # to a byte string with a reasonable output codec. We want to 281. # use the byte string in the chunk. 282. for charset in USASCII, charset, UTF8: 283. try: 284. outcodec = charset.output_codec or 'us-ascii' 285. s = s.encode(outcodec, errors) 286. break 287. except UnicodeError: 288. pass 289. else: 290. assert False, 'utf-8 conversion failed' 291. self._chunks.append((s, charset)) 292. 293. def _split(self, s, charset, maxlinelen, splitchars): 294. # Split up a header safely for use with encode_chunks. 295. splittable = charset.to_splittable(s) 296. encoded = charset.from_splittable(splittable, True) 297. elen = charset.encoded_header_len(encoded) 298. # If the line's encoded length first, just return it 299. if elen <= maxlinelen: 300. return [(encoded, charset)] 301. # If we have undetermined raw 8bit characters sitting in a byte 302. # string, we really don't know what the right thing to do is. We 303. # can't really split it because it might be multibyte data which we 304. # could break if we split it between pairs. The least harm seems to 305. # be to not split the header at all, but that means they could go out 306. # longer than maxlinelen. 307. if charset == '8bit': 308. return [(s, charset)] 309. # BAW: I'm not sure what the right test here is. What we're trying to 310. # do is be faithful to RFC 2822's recommendation that ($2.2.3): 311. # 312. # "Note: Though structured field bodies are defined in such a way that 313. # folding can take place between many of the lexical tokens (and even 314. # within some of the lexical tokens), folding SHOULD be limited to 315. # placing the CRLF at higher-level syntactic breaks." 316. # 317. # For now, I can only imagine doing this when the charset is us-ascii, 318. # although it's possible that other charsets may also benefit from the 319. # higher-level syntactic breaks. 320. elif charset == 'us-ascii': 321. return self._split_ascii(s, charset, maxlinelen, splitchars) 322. # BAW: should we use encoded? 323. elif elen == len(s): 324. # We can split on _maxlinelen boundaries because we know that the 325. # encoding won't change the size of the string 326. splitpnt = maxlinelen 327. first = charset.from_splittable(splittable[:splitpnt], False) 328. last = charset.from_splittable(splittable[splitpnt:], False) 329. else: 330. # Binary search for split point 331. first, last = _binsplit(splittable, charset, maxlinelen) 332. # first is of the proper length so just wrap it in the appropriate 333. # chrome. last must be recursively split. 334. fsplittable = charset.to_splittable(first) 335. fencoded = charset.from_splittable(fsplittable, True) 336. chunk = [(fencoded, charset)] 337. return chunk + self._split(last, charset, self._maxlinelen, splitchars) 338. 339. def _split_ascii(self, s, charset, firstlen, splitchars): 340. chunks = _split_ascii(s, firstlen, self._maxlinelen, 341. self._continuation_ws, splitchars) 342. return zip(chunks, [charset]*len(chunks)) 343. 344. def _encode_chunks(self, newchunks, maxlinelen): 345. # MIME-encode a header with many different charsets and/or encodings. 346. # 347. # Given a list of pairs (string, charset), return a MIME-encoded 348. # string suitable for use in a header field. Each pair may have 349. # different charsets and/or encodings, and the resulting header will 350. # accurately reflect each setting. 351. # 352. # Each encoding can be email.Utils.QP (quoted-printable, for 353. # ASCII-like character sets like iso-8859-1), email.Utils.BASE64 354. # (Base64, for non-ASCII like character sets like KOI8-R and 355. # iso-2022-jp), or None (no encoding). 356. # 357. # Each pair will be represented on a separate line; the resulting 358. # string will be in the format: 359. # 360. # =?charset1?q?Mar=EDa_Gonz=E1lez_Alonso?=\n 361. # =?charset2?b?SvxyZ2VuIEL2aW5n?=" 362. chunks = [] 363. for header, charset in newchunks: 364. if not header: 365. continue 366. if charset is None or charset.header_encoding is None: 367. s = header 368. else: 369. s = charset.header_encode(header) 370. # Don't add more folding whitespace than necessary 371. if chunks and chunks[-1].endswith(' '): 372. extra = '' 373. else: 374. extra = ' ' 375. _max_append(chunks, s, maxlinelen, extra) 376. joiner = NL + self._continuation_ws 377. return joiner.join(chunks) 378. 379. def encode(self, splitchars=';, '): 380. """Encode a message header into an RFC-compliant format. 381. 382. There are many issues involved in converting a given string for use in 383. an email header. Only certain character sets are readable in most 384. email clients, and as header strings can only contain a subset of 385. 7-bit ASCII, care must be taken to properly convert and encode (with 386. Base64 or quoted-printable) header strings. In addition, there is a 387. 75-character length limit on any given encoded header field, so 388. line-wrapping must be performed, even with double-byte character sets. 389. 390. This method will do its best to convert the string to the correct 391. character set used in email, and encode and line wrap it safely with 392. the appropriate scheme for that character set. 393. 394. If the given charset is not known or an error occurs during 395. conversion, this function will return the header untouched. 396. 397. Optional splitchars is a string containing characters to split long 398. ASCII lines on, in rough support of RFC 2822's `highest level 399. syntactic breaks'. This doesn't affect RFC 2047 encoded lines. 400. """ 401. newchunks = [] 402. maxlinelen = self._firstlinelen 403. lastlen = 0 404. for s, charset in self._chunks: 405. # The first bit of the next chunk should be just long enough to 406. # fill the next line. Don't forget the space separating the 407. # encoded words. 408. targetlen = maxlinelen - lastlen - 1 409. if targetlen < charset.encoded_header_len(''): 410. # Stick it on the next line 411. targetlen = maxlinelen 412. newchunks += self._split(s, charset, targetlen, splitchars) 413. lastchunk, lastcharset = newchunks[-1] 414. lastlen = lastcharset.encoded_header_len(lastchunk) 415. return self._encode_chunks(newchunks, maxlinelen) 416. 417. 418. 419. def _split_ascii(s, firstlen, restlen, continuation_ws, splitchars): 420. lines = [] 421. maxlen = firstlen 422. for line in s.splitlines(): 423. # Ignore any leading whitespace (i.e. continuation whitespace) already 424. # on the line, since we'll be adding our own. 425. line = line.lstrip() 426. if len(line) < maxlen: 427. lines.append(line) 428. maxlen = restlen 429. continue 430. # Attempt to split the line at the highest-level syntactic break 431. # possible. Note that we don't have a lot of smarts about field 432. # syntax; we just try to break on semi-colons, then commas, then 433. # whitespace. 434. for ch in splitchars: 435. if line.find(ch) >= 0: 436. break 437. else: 438. # There's nothing useful to split the line on, not even spaces, so 439. # just append this line unchanged 440. lines.append(line) 441. maxlen = restlen 442. continue 443. # Now split the line on the character plus trailing whitespace 444. cre = re.compile(r'%s\s*' % ch) 445. if ch in ';,': 446. eol = ch 447. else: 448. eol = '' 449. joiner = eol + ' ' 450. joinlen = len(joiner) 451. wslen = len(continuation_ws.replace('\t', SPACE8)) 452. this = [] 453. linelen = 0 454. for part in cre.split(line): 455. curlen = linelen + max(0, len(this)-1) * joinlen 456. partlen = len(part) 457. onfirstline = not lines 458. # We don't want to split after the field name, if we're on the 459. # first line and the field name is present in the header string. 460. if ch == ' ' and onfirstline and \ 461. len(this) == 1 and fcre.match(this[0]): 462. this.append(part) 463. linelen += partlen 464. elif curlen + partlen > maxlen: 465. if this: 466. lines.append(joiner.join(this) + eol) 467. # If this part is longer than maxlen and we aren't already 468. # splitting on whitespace, try to recursively split this line 469. # on whitespace. 470. if partlen > maxlen and ch <> ' ': 471. subl = _split_ascii(part, maxlen, restlen, 472. continuation_ws, ' ') 473. lines.extend(subl[:-1]) 474. this = [subl[-1]] 475. else: 476. this = [part] 477. linelen = wslen + len(this[-1]) 478. maxlen = restlen 479. else: 480. this.append(part) 481. linelen += partlen 482. # Put any left over parts on a line by themselves 483. if this: 484. lines.append(joiner.join(this)) 485. return lines 486. 487. 488. 489. def _binsplit(splittable, charset, maxlinelen): 490. i = 0 491. j = len(splittable) 492. while i < j: 493. # Invariants: 494. # 1. splittable[:k] fits for all k <= i (note that we *assume*, 495. # at the start, that splittable[:0] fits). 496. # 2. splittable[:k] does not fit for any k > j (at the start, 497. # this means we shouldn't look at any k > len(splittable)). 498. # 3. We don't know about splittable[:k] for k in i+1..j. 499. # 4. We want to set i to the largest k that fits, with i <= k <= j. 500. # 501. m = (i+j+1) >> 1 # ceiling((i+j)/2); i < m <= j 502. chunk = charset.from_splittable(splittable[:m], True) 503. chunklen = charset.encoded_header_len(chunk) 504. if chunklen <= maxlinelen: 505. # m is acceptable, so is a new lower bound. 506. i = m 507. else: 508. # m is not acceptable, so final i must be < m. 509. j = m - 1 510. # i == j. Invariant #1 implies that splittable[:i] fits, and 511. # invariant #2 implies that splittable[:i+1] does not fit, so i 512. # is what we're looking for. 513. first = charset.from_splittable(splittable[:i], False) 514. last = charset.from_splittable(splittable[i:], False) 515. return first, last