http.py (16160B)
1 from __future__ import unicode_literals 2 3 import errno 4 import os 5 import socket 6 import time 7 import random 8 import re 9 10 from .common import FileDownloader 11 from ..compat import ( 12 compat_str, 13 compat_urllib_error, 14 ) 15 from ..utils import ( 16 ContentTooShortError, 17 encodeFilename, 18 int_or_none, 19 sanitize_open, 20 sanitized_Request, 21 write_xattr, 22 XAttrMetadataError, 23 XAttrUnavailableError, 24 ) 25 26 27 class HttpFD(FileDownloader): 28 def real_download(self, filename, info_dict): 29 url = info_dict['url'] 30 31 class DownloadContext(dict): 32 __getattr__ = dict.get 33 __setattr__ = dict.__setitem__ 34 __delattr__ = dict.__delitem__ 35 36 ctx = DownloadContext() 37 ctx.filename = filename 38 ctx.tmpfilename = self.temp_name(filename) 39 ctx.stream = None 40 41 # Do not include the Accept-Encoding header 42 headers = {'Youtubedl-no-compression': 'True'} 43 add_headers = info_dict.get('http_headers') 44 if add_headers: 45 headers.update(add_headers) 46 47 is_test = self.params.get('test', False) 48 chunk_size = self._TEST_FILE_SIZE if is_test else ( 49 info_dict.get('downloader_options', {}).get('http_chunk_size') 50 or self.params.get('http_chunk_size') or 0) 51 52 ctx.open_mode = 'wb' 53 ctx.resume_len = 0 54 ctx.data_len = None 55 ctx.block_size = self.params.get('buffersize', 1024) 56 ctx.start_time = time.time() 57 ctx.chunk_size = None 58 59 if self.params.get('continuedl', True): 60 # Establish possible resume length 61 if os.path.isfile(encodeFilename(ctx.tmpfilename)): 62 ctx.resume_len = os.path.getsize( 63 encodeFilename(ctx.tmpfilename)) 64 65 ctx.is_resume = ctx.resume_len > 0 66 67 count = 0 68 retries = self.params.get('retries', 0) 69 70 class SucceedDownload(Exception): 71 pass 72 73 class RetryDownload(Exception): 74 def __init__(self, source_error): 75 self.source_error = source_error 76 77 class NextFragment(Exception): 78 pass 79 80 def set_range(req, start, end): 81 range_header = 'bytes=%d-' % start 82 if end: 83 range_header += compat_str(end) 84 req.add_header('Range', range_header) 85 86 def establish_connection(): 87 ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size) 88 if not is_test and chunk_size else chunk_size) 89 if ctx.resume_len > 0: 90 range_start = ctx.resume_len 91 if ctx.is_resume: 92 self.report_resuming_byte(ctx.resume_len) 93 ctx.open_mode = 'ab' 94 elif ctx.chunk_size > 0: 95 range_start = 0 96 else: 97 range_start = None 98 ctx.is_resume = False 99 range_end = range_start + ctx.chunk_size - 1 if ctx.chunk_size else None 100 if range_end and ctx.data_len is not None and range_end >= ctx.data_len: 101 range_end = ctx.data_len - 1 102 has_range = range_start is not None 103 ctx.has_range = has_range 104 request = sanitized_Request(url, None, headers) 105 if has_range: 106 set_range(request, range_start, range_end) 107 # Establish connection 108 try: 109 try: 110 ctx.data = self.ydl.urlopen(request) 111 except (compat_urllib_error.URLError, ) as err: 112 # reason may not be available, e.g. for urllib2.HTTPError on python 2.6 113 reason = getattr(err, 'reason', None) 114 if isinstance(reason, socket.timeout): 115 raise RetryDownload(err) 116 raise err 117 # When trying to resume, Content-Range HTTP header of response has to be checked 118 # to match the value of requested Range HTTP header. This is due to a webservers 119 # that don't support resuming and serve a whole file with no Content-Range 120 # set in response despite of requested Range (see 121 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799) 122 if has_range: 123 content_range = ctx.data.headers.get('Content-Range') 124 if content_range: 125 content_range_m = re.search(r'bytes (\d+)-(\d+)?(?:/(\d+))?', content_range) 126 # Content-Range is present and matches requested Range, resume is possible 127 if content_range_m: 128 if range_start == int(content_range_m.group(1)): 129 content_range_end = int_or_none(content_range_m.group(2)) 130 content_len = int_or_none(content_range_m.group(3)) 131 accept_content_len = ( 132 # Non-chunked download 133 not ctx.chunk_size 134 # Chunked download and requested piece or 135 # its part is promised to be served 136 or content_range_end == range_end 137 or content_len < range_end) 138 if accept_content_len: 139 ctx.data_len = content_len 140 return 141 # Content-Range is either not present or invalid. Assuming remote webserver is 142 # trying to send the whole file, resume is not possible, so wiping the local file 143 # and performing entire redownload 144 self.report_unable_to_resume() 145 ctx.resume_len = 0 146 ctx.open_mode = 'wb' 147 ctx.data_len = int_or_none(ctx.data.info().get('Content-length', None)) 148 return 149 except (compat_urllib_error.HTTPError, ) as err: 150 if err.code == 416: 151 # Unable to resume (requested range not satisfiable) 152 try: 153 # Open the connection again without the range header 154 ctx.data = self.ydl.urlopen( 155 sanitized_Request(url, None, headers)) 156 content_length = ctx.data.info()['Content-Length'] 157 except (compat_urllib_error.HTTPError, ) as err: 158 if err.code < 500 or err.code >= 600: 159 raise 160 else: 161 # Examine the reported length 162 if (content_length is not None 163 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)): 164 # The file had already been fully downloaded. 165 # Explanation to the above condition: in issue #175 it was revealed that 166 # YouTube sometimes adds or removes a few bytes from the end of the file, 167 # changing the file size slightly and causing problems for some users. So 168 # I decided to implement a suggested change and consider the file 169 # completely downloaded if the file size differs less than 100 bytes from 170 # the one in the hard drive. 171 self.report_file_already_downloaded(ctx.filename) 172 self.try_rename(ctx.tmpfilename, ctx.filename) 173 self._hook_progress({ 174 'filename': ctx.filename, 175 'status': 'finished', 176 'downloaded_bytes': ctx.resume_len, 177 'total_bytes': ctx.resume_len, 178 }) 179 raise SucceedDownload() 180 else: 181 # The length does not match, we start the download over 182 self.report_unable_to_resume() 183 ctx.resume_len = 0 184 ctx.open_mode = 'wb' 185 return 186 elif err.code < 500 or err.code >= 600: 187 # Unexpected HTTP error 188 raise 189 raise RetryDownload(err) 190 except socket.error as err: 191 if err.errno != errno.ECONNRESET: 192 # Connection reset is no problem, just retry 193 raise 194 raise RetryDownload(err) 195 196 def download(): 197 data_len = ctx.data.info().get('Content-length', None) 198 199 # Range HTTP header may be ignored/unsupported by a webserver 200 # (e.g. extractor/scivee.py, extractor/bambuser.py). 201 # However, for a test we still would like to download just a piece of a file. 202 # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control 203 # block size when downloading a file. 204 if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE): 205 data_len = self._TEST_FILE_SIZE 206 207 if data_len is not None: 208 data_len = int(data_len) + ctx.resume_len 209 min_data_len = self.params.get('min_filesize') 210 max_data_len = self.params.get('max_filesize') 211 if min_data_len is not None and data_len < min_data_len: 212 self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len)) 213 return False 214 if max_data_len is not None and data_len > max_data_len: 215 self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len)) 216 return False 217 218 byte_counter = 0 + ctx.resume_len 219 block_size = ctx.block_size 220 start = time.time() 221 222 # measure time over whole while-loop, so slow_down() and best_block_size() work together properly 223 now = None # needed for slow_down() in the first loop run 224 before = start # start measuring 225 226 def retry(e): 227 to_stdout = ctx.tmpfilename == '-' 228 if ctx.stream is not None: 229 if not to_stdout: 230 ctx.stream.close() 231 ctx.stream = None 232 ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename)) 233 raise RetryDownload(e) 234 235 while True: 236 try: 237 # Download and write 238 data_block = ctx.data.read(block_size if data_len is None else min(block_size, data_len - byte_counter)) 239 # socket.timeout is a subclass of socket.error but may not have 240 # errno set 241 except socket.timeout as e: 242 retry(e) 243 except socket.error as e: 244 # SSLError on python 2 (inherits socket.error) may have 245 # no errno set but this error message 246 if e.errno in (errno.ECONNRESET, errno.ETIMEDOUT) or getattr(e, 'message', None) == 'The read operation timed out': 247 retry(e) 248 raise 249 250 byte_counter += len(data_block) 251 252 # exit loop when download is finished 253 if len(data_block) == 0: 254 break 255 256 # Open destination file just in time 257 if ctx.stream is None: 258 try: 259 ctx.stream, ctx.tmpfilename = sanitize_open( 260 ctx.tmpfilename, ctx.open_mode) 261 assert ctx.stream is not None 262 ctx.filename = self.undo_temp_name(ctx.tmpfilename) 263 self.report_destination(ctx.filename) 264 except (OSError, IOError) as err: 265 self.report_error('unable to open for writing: %s' % str(err)) 266 return False 267 268 if self.params.get('xattr_set_filesize', False) and data_len is not None: 269 try: 270 write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8')) 271 except (XAttrUnavailableError, XAttrMetadataError) as err: 272 self.report_error('unable to set filesize xattr: %s' % str(err)) 273 274 try: 275 ctx.stream.write(data_block) 276 except (IOError, OSError) as err: 277 self.to_stderr('\n') 278 self.report_error('unable to write data: %s' % str(err)) 279 return False 280 281 # Apply rate limit 282 self.slow_down(start, now, byte_counter - ctx.resume_len) 283 284 # end measuring of one loop run 285 now = time.time() 286 after = now 287 288 # Adjust block size 289 if not self.params.get('noresizebuffer', False): 290 block_size = self.best_block_size(after - before, len(data_block)) 291 292 before = after 293 294 # Progress message 295 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len) 296 if ctx.data_len is None: 297 eta = None 298 else: 299 eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len) 300 301 self._hook_progress({ 302 'status': 'downloading', 303 'downloaded_bytes': byte_counter, 304 'total_bytes': ctx.data_len, 305 'tmpfilename': ctx.tmpfilename, 306 'filename': ctx.filename, 307 'eta': eta, 308 'speed': speed, 309 'elapsed': now - ctx.start_time, 310 }) 311 312 if data_len is not None and byte_counter == data_len: 313 break 314 315 if not is_test and ctx.chunk_size and ctx.data_len is not None and byte_counter < ctx.data_len: 316 ctx.resume_len = byte_counter 317 # ctx.block_size = block_size 318 raise NextFragment() 319 320 if ctx.stream is None: 321 self.to_stderr('\n') 322 self.report_error('Did not get any data blocks') 323 return False 324 if ctx.tmpfilename != '-': 325 ctx.stream.close() 326 327 if data_len is not None and byte_counter != data_len: 328 err = ContentTooShortError(byte_counter, int(data_len)) 329 if count <= retries: 330 retry(err) 331 raise err 332 333 self.try_rename(ctx.tmpfilename, ctx.filename) 334 335 # Update file modification time 336 if self.params.get('updatetime', True): 337 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None)) 338 339 self._hook_progress({ 340 'downloaded_bytes': byte_counter, 341 'total_bytes': byte_counter, 342 'filename': ctx.filename, 343 'status': 'finished', 344 'elapsed': time.time() - ctx.start_time, 345 }) 346 347 return True 348 349 while count <= retries: 350 try: 351 establish_connection() 352 return download() 353 except RetryDownload as e: 354 count += 1 355 if count <= retries: 356 self.report_retry(e.source_error, count, retries) 357 continue 358 except NextFragment: 359 continue 360 except SucceedDownload: 361 return True 362 363 self.report_error('giving up after %s retries' % retries) 364 return False