Update README.md

[youtubedl] / youtube_dl / downloader / http.py

diff --git a/youtube_dl/downloader/http.py b/youtube_dl/downloader/http.py
index e68f20c9f46a93ebfeca2ff47dc0843f4ab94874..5046878dfcd874013e737e85d32764a95737406e 100644 (file)
--- a/youtube_dl/downloader/http.py
+++ b/youtube_dl/downloader/http.py
@@ -1,88 +1,159 @@
 from __future__ import unicode_literals
 
 from __future__ import unicode_literals
 

+import errno

 import os
 import os

+import socket

 import time
 import time

+import random
+import re

 
 from .common import FileDownloader
 from ..compat import (
 
 from .common import FileDownloader
 from ..compat import (

-    compat_urllib_request,
+    compat_str,

     compat_urllib_error,
 )
 from ..utils import (
     ContentTooShortError,
     encodeFilename,
     compat_urllib_error,
 )
 from ..utils import (
     ContentTooShortError,
     encodeFilename,

+    int_or_none,

     sanitize_open,
     sanitize_open,

-    format_bytes,
+    sanitized_Request,
+    write_xattr,
+    XAttrMetadataError,
+    XAttrUnavailableError,

 )
 
 
 class HttpFD(FileDownloader):
     def real_download(self, filename, info_dict):
         url = info_dict['url']
 )
 
 
 class HttpFD(FileDownloader):
     def real_download(self, filename, info_dict):
         url = info_dict['url']

-        tmpfilename = self.temp_name(filename)
-        stream = None
+
+        class DownloadContext(dict):
+            __getattr__ = dict.get
+            __setattr__ = dict.__setitem__
+            __delattr__ = dict.__delitem__
+
+        ctx = DownloadContext()
+        ctx.filename = filename
+        ctx.tmpfilename = self.temp_name(filename)
+        ctx.stream = None

 
         # Do not include the Accept-Encoding header
         headers = {'Youtubedl-no-compression': 'True'}
 
         # Do not include the Accept-Encoding header
         headers = {'Youtubedl-no-compression': 'True'}

-        if 'user_agent' in info_dict:
-            headers['Youtubedl-user-agent'] = info_dict['user_agent']
-        if 'http_referer' in info_dict:
-            headers['Referer'] = info_dict['http_referer']

         add_headers = info_dict.get('http_headers')
         if add_headers:
             headers.update(add_headers)
         add_headers = info_dict.get('http_headers')
         if add_headers:
             headers.update(add_headers)

-        data = info_dict.get('http_post_data')
-        http_method = info_dict.get('http_method')
-        basic_request = compat_urllib_request.Request(url, data, headers)
-        request = compat_urllib_request.Request(url, data, headers)
-        if http_method is not None:
-            basic_request.get_method = lambda: http_method
-            request.get_method = lambda: http_method

 
         is_test = self.params.get('test', False)
 
         is_test = self.params.get('test', False)

+        chunk_size = self._TEST_FILE_SIZE if is_test else (
+            info_dict.get('downloader_options', {}).get('http_chunk_size')
+            or self.params.get('http_chunk_size') or 0)

 
 

-        if is_test:
-            request.add_header('Range', 'bytes=0-%s' % str(self._TEST_FILE_SIZE - 1))
-
-        # Establish possible resume length
-        if os.path.isfile(encodeFilename(tmpfilename)):
-            resume_len = os.path.getsize(encodeFilename(tmpfilename))
-        else:
-            resume_len = 0
-
-        open_mode = 'wb'
-        if resume_len != 0:
-            if self.params.get('continuedl', False):
-                self.report_resuming_byte(resume_len)
-                request.add_header('Range', 'bytes=%d-' % resume_len)
-                open_mode = 'ab'
-            else:
-                resume_len = 0
+        ctx.open_mode = 'wb'
+        ctx.resume_len = 0
+        ctx.data_len = None
+        ctx.block_size = self.params.get('buffersize', 1024)
+        ctx.start_time = time.time()
+        ctx.chunk_size = None
+
+        if self.params.get('continuedl', True):
+            # Establish possible resume length
+            if os.path.isfile(encodeFilename(ctx.tmpfilename)):
+                ctx.resume_len = os.path.getsize(
+                    encodeFilename(ctx.tmpfilename))
+
+        ctx.is_resume = ctx.resume_len > 0

 
         count = 0
         retries = self.params.get('retries', 0)
 
         count = 0
         retries = self.params.get('retries', 0)

-        while count <= retries:
+
+        class SucceedDownload(Exception):
+            pass
+
+        class RetryDownload(Exception):
+            def __init__(self, source_error):
+                self.source_error = source_error
+
+        class NextFragment(Exception):
+            pass
+
+        def set_range(req, start, end):
+            range_header = 'bytes=%d-' % start
+            if end:
+                range_header += compat_str(end)
+            req.add_header('Range', range_header)
+
+        def establish_connection():
+            ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
+                              if not is_test and chunk_size else chunk_size)
+            if ctx.resume_len > 0:
+                range_start = ctx.resume_len
+                if ctx.is_resume:
+                    self.report_resuming_byte(ctx.resume_len)
+                ctx.open_mode = 'ab'
+            elif ctx.chunk_size > 0:
+                range_start = 0
+            else:
+                range_start = None
+            ctx.is_resume = False
+            range_end = range_start + ctx.chunk_size - 1 if ctx.chunk_size else None
+            if range_end and ctx.data_len is not None and range_end >= ctx.data_len:
+                range_end = ctx.data_len - 1
+            has_range = range_start is not None
+            ctx.has_range = has_range
+            request = sanitized_Request(url, None, headers)
+            if has_range:
+                set_range(request, range_start, range_end)

             # Establish connection
             try:
             # Establish connection
             try:

-                data = self.ydl.urlopen(request)
-                break
+                ctx.data = self.ydl.urlopen(request)
+                # When trying to resume, Content-Range HTTP header of response has to be checked
+                # to match the value of requested Range HTTP header. This is due to a webservers
+                # that don't support resuming and serve a whole file with no Content-Range
+                # set in response despite of requested Range (see
+                # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
+                if has_range:
+                    content_range = ctx.data.headers.get('Content-Range')
+                    if content_range:
+                        content_range_m = re.search(r'bytes (\d+)-(\d+)?(?:/(\d+))?', content_range)
+                        # Content-Range is present and matches requested Range, resume is possible
+                        if content_range_m:
+                            if range_start == int(content_range_m.group(1)):
+                                content_range_end = int_or_none(content_range_m.group(2))
+                                content_len = int_or_none(content_range_m.group(3))
+                                accept_content_len = (
+                                    # Non-chunked download
+                                    not ctx.chunk_size
+                                    # Chunked download and requested piece or
+                                    # its part is promised to be served
+                                    or content_range_end == range_end
+                                    or content_len < range_end)
+                                if accept_content_len:
+                                    ctx.data_len = content_len
+                                    return
+                    # Content-Range is either not present or invalid. Assuming remote webserver is
+                    # trying to send the whole file, resume is not possible, so wiping the local file
+                    # and performing entire redownload
+                    self.report_unable_to_resume()
+                    ctx.resume_len = 0
+                    ctx.open_mode = 'wb'
+                ctx.data_len = int_or_none(ctx.data.info().get('Content-length', None))
+                return

             except (compat_urllib_error.HTTPError, ) as err:
             except (compat_urllib_error.HTTPError, ) as err:

-                if (err.code < 500 or err.code >= 600) and err.code != 416:
-                    # Unexpected HTTP error
-                    raise
-                elif err.code == 416:
+                if err.code == 416:

                     # Unable to resume (requested range not satisfiable)
                     try:
                         # Open the connection again without the range header
                     # Unable to resume (requested range not satisfiable)
                     try:
                         # Open the connection again without the range header

-                        data = self.ydl.urlopen(basic_request)
-                        content_length = data.info()['Content-Length']
+                        ctx.data = self.ydl.urlopen(
+                            sanitized_Request(url, None, headers))
+                        content_length = ctx.data.info()['Content-Length']

                     except (compat_urllib_error.HTTPError, ) as err:
                         if err.code < 500 or err.code >= 600:
                             raise
                     else:
                         # Examine the reported length
                     except (compat_urllib_error.HTTPError, ) as err:
                         if err.code < 500 or err.code >= 600:
                             raise
                     else:
                         # Examine the reported length

-                        if (content_length is not None and
-                                (resume_len - 100 < int(content_length) < resume_len + 100)):
+                        if (content_length is not None
+                                and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):

                             # The file had already been fully downloaded.
                             # Explanation to the above condition: in issue #175 it was revealed that
                             # YouTube sometimes adds or removes a few bytes from the end of the file,
                             # The file had already been fully downloaded.
                             # Explanation to the above condition: in issue #175 it was revealed that
                             # YouTube sometimes adds or removes a few bytes from the end of the file,


@@ -90,139 +161,194 @@ class HttpFD(FileDownloader):
                             # I decided to implement a suggested change and consider the file
                             # completely downloaded if the file size differs less than 100 bytes from
                             # the one in the hard drive.
                             # I decided to implement a suggested change and consider the file
                             # completely downloaded if the file size differs less than 100 bytes from
                             # the one in the hard drive.

-                            self.report_file_already_downloaded(filename)
-                            self.try_rename(tmpfilename, filename)
+                            self.report_file_already_downloaded(ctx.filename)
+                            self.try_rename(ctx.tmpfilename, ctx.filename)

                             self._hook_progress({
                             self._hook_progress({

-                                'filename': filename,
+                                'filename': ctx.filename,

                                 'status': 'finished',
                                 'status': 'finished',

+                                'downloaded_bytes': ctx.resume_len,
+                                'total_bytes': ctx.resume_len,

                             })
                             })

-                            return True
+                            raise SucceedDownload()

                         else:
                             # The length does not match, we start the download over
                             self.report_unable_to_resume()
                         else:
                             # The length does not match, we start the download over
                             self.report_unable_to_resume()

-                            resume_len = 0
-                            open_mode = 'wb'
-                            break
-            # Retry
-            count += 1
-            if count <= retries:
-                self.report_retry(count, retries)
-
-        if count > retries:
-            self.report_error('giving up after %s retries' % retries)
-            return False
-
-        data_len = data.info().get('Content-length', None)
-
-        # Range HTTP header may be ignored/unsupported by a webserver
-        # (e.g. extractor/scivee.py, extractor/bambuser.py).
-        # However, for a test we still would like to download just a piece of a file.
-        # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
-        # block size when downloading a file.
-        if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
-            data_len = self._TEST_FILE_SIZE
-
-        if data_len is not None:
-            data_len = int(data_len) + resume_len
-            min_data_len = self.params.get("min_filesize", None)
-            max_data_len = self.params.get("max_filesize", None)
-            if min_data_len is not None and data_len < min_data_len:
-                self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
-                return False
-            if max_data_len is not None and data_len > max_data_len:
-                self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
-                return False
+                            ctx.resume_len = 0
+                            ctx.open_mode = 'wb'
+                            return
+                elif err.code < 500 or err.code >= 600:
+                    # Unexpected HTTP error
+                    raise
+                raise RetryDownload(err)
+            except socket.error as err:
+                if err.errno != errno.ECONNRESET:
+                    # Connection reset is no problem, just retry
+                    raise
+                raise RetryDownload(err)
+
+        def download():
+            data_len = ctx.data.info().get('Content-length', None)
+
+            # Range HTTP header may be ignored/unsupported by a webserver
+            # (e.g. extractor/scivee.py, extractor/bambuser.py).
+            # However, for a test we still would like to download just a piece of a file.
+            # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
+            # block size when downloading a file.
+            if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
+                data_len = self._TEST_FILE_SIZE

 
 

-        data_len_str = format_bytes(data_len)
-        byte_counter = 0 + resume_len
-        block_size = self.params.get('buffersize', 1024)
-        start = time.time()
+            if data_len is not None:
+                data_len = int(data_len) + ctx.resume_len
+                min_data_len = self.params.get('min_filesize')
+                max_data_len = self.params.get('max_filesize')
+                if min_data_len is not None and data_len < min_data_len:
+                    self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
+                    return False
+                if max_data_len is not None and data_len > max_data_len:
+                    self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
+                    return False

 
 

-        # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
-        now = None  # needed for slow_down() in the first loop run
-        before = start  # start measuring
-        while True:
+            byte_counter = 0 + ctx.resume_len
+            block_size = ctx.block_size
+            start = time.time()

 
 

-            # Download and write
-            data_block = data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
-            byte_counter += len(data_block)
+            # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
+            now = None  # needed for slow_down() in the first loop run
+            before = start  # start measuring

 
 

-            # exit loop when download is finished
-            if len(data_block) == 0:
-                break
+            def retry(e):
+                to_stdout = ctx.tmpfilename == '-'
+                if not to_stdout:
+                    ctx.stream.close()
+                ctx.stream = None
+                ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
+                raise RetryDownload(e)

 
 

-            # Open destination file just in time
-            if stream is None:
+            while True:

                 try:
                 try:

-                    (stream, tmpfilename) = sanitize_open(tmpfilename, open_mode)
-                    assert stream is not None
-                    filename = self.undo_temp_name(tmpfilename)
-                    self.report_destination(filename)
-                except (OSError, IOError) as err:
-                    self.report_error('unable to open for writing: %s' % str(err))
+                    # Download and write
+                    data_block = ctx.data.read(block_size if data_len is None else min(block_size, data_len - byte_counter))
+                # socket.timeout is a subclass of socket.error but may not have
+                # errno set
+                except socket.timeout as e:
+                    retry(e)
+                except socket.error as e:
+                    if e.errno not in (errno.ECONNRESET, errno.ETIMEDOUT):
+                        raise
+                    retry(e)
+
+                byte_counter += len(data_block)
+
+                # exit loop when download is finished
+                if len(data_block) == 0:
+                    break
+
+                # Open destination file just in time
+                if ctx.stream is None:
+                    try:
+                        ctx.stream, ctx.tmpfilename = sanitize_open(
+                            ctx.tmpfilename, ctx.open_mode)
+                        assert ctx.stream is not None
+                        ctx.filename = self.undo_temp_name(ctx.tmpfilename)
+                        self.report_destination(ctx.filename)
+                    except (OSError, IOError) as err:
+                        self.report_error('unable to open for writing: %s' % str(err))
+                        return False
+
+                    if self.params.get('xattr_set_filesize', False) and data_len is not None:
+                        try:
+                            write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8'))
+                        except (XAttrUnavailableError, XAttrMetadataError) as err:
+                            self.report_error('unable to set filesize xattr: %s' % str(err))
+
+                try:
+                    ctx.stream.write(data_block)
+                except (IOError, OSError) as err:
+                    self.to_stderr('\n')
+                    self.report_error('unable to write data: %s' % str(err))

                     return False
                     return False

-            try:
-                stream.write(data_block)
-            except (IOError, OSError) as err:
-                self.to_stderr('\n')
-                self.report_error('unable to write data: %s' % str(err))
-                return False

 
 

-            # Apply rate limit
-            self.slow_down(start, now, byte_counter - resume_len)
+                # Apply rate limit
+                self.slow_down(start, now, byte_counter - ctx.resume_len)

 
 

-            # end measuring of one loop run
-            now = time.time()
-            after = now
+                # end measuring of one loop run
+                now = time.time()
+                after = now

 
 

-            # Adjust block size
-            if not self.params.get('noresizebuffer', False):
-                block_size = self.best_block_size(after - before, len(data_block))
+                # Adjust block size
+                if not self.params.get('noresizebuffer', False):
+                    block_size = self.best_block_size(after - before, len(data_block))

 
 

-            before = after
+                before = after

 
 

-            # Progress message
-            speed = self.calc_speed(start, now, byte_counter - resume_len)
-            if data_len is None:
-                eta = percent = None
-            else:
-                percent = self.calc_percent(byte_counter, data_len)
-                eta = self.calc_eta(start, time.time(), data_len - resume_len, byte_counter - resume_len)
-            self.report_progress(percent, data_len_str, speed, eta)
+                # Progress message
+                speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
+                if ctx.data_len is None:
+                    eta = None
+                else:
+                    eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
+
+                self._hook_progress({
+                    'status': 'downloading',
+                    'downloaded_bytes': byte_counter,
+                    'total_bytes': ctx.data_len,
+                    'tmpfilename': ctx.tmpfilename,
+                    'filename': ctx.filename,
+                    'eta': eta,
+                    'speed': speed,
+                    'elapsed': now - ctx.start_time,
+                })
+
+                if data_len is not None and byte_counter == data_len:
+                    break
+
+            if not is_test and ctx.chunk_size and ctx.data_len is not None and byte_counter < ctx.data_len:
+                ctx.resume_len = byte_counter
+                # ctx.block_size = block_size
+                raise NextFragment()
+
+            if ctx.stream is None:
+                self.to_stderr('\n')
+                self.report_error('Did not get any data blocks')
+                return False
+            if ctx.tmpfilename != '-':
+                ctx.stream.close()
+
+            if data_len is not None and byte_counter != data_len:
+                err = ContentTooShortError(byte_counter, int(data_len))
+                if count <= retries:
+                    retry(err)
+                raise err
+
+            self.try_rename(ctx.tmpfilename, ctx.filename)
+
+            # Update file modification time
+            if self.params.get('updatetime', True):
+                info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))

 
             self._hook_progress({
                 'downloaded_bytes': byte_counter,
 
             self._hook_progress({
                 'downloaded_bytes': byte_counter,

-                'total_bytes': data_len,
-                'tmpfilename': tmpfilename,
-                'filename': filename,
-                'status': 'downloading',
-                'eta': eta,
-                'speed': speed,
+                'total_bytes': byte_counter,
+                'filename': ctx.filename,
+                'status': 'finished',
+                'elapsed': time.time() - ctx.start_time,

             })
 
             })
 

-            if is_test and byte_counter == data_len:
-                break
-
-        if stream is None:
-            self.to_stderr('\n')
-            self.report_error('Did not get any data blocks')
-            return False
-        if tmpfilename != '-':
-            stream.close()
-        self.report_finish(data_len_str, (time.time() - start))
-        if data_len is not None and byte_counter != data_len:
-            raise ContentTooShortError(byte_counter, int(data_len))
-        self.try_rename(tmpfilename, filename)
-
-        # Update file modification time
-        if self.params.get('updatetime', True):
-            info_dict['filetime'] = self.try_utime(filename, data.info().get('last-modified', None))
-
-        self._hook_progress({
-            'downloaded_bytes': byte_counter,
-            'total_bytes': byte_counter,
-            'filename': filename,
-            'status': 'finished',
-        })
-
-        return True
+            return True
+
+        while count <= retries:
+            try:
+                establish_connection()
+                return download()
+            except RetryDownload as e:
+                count += 1
+                if count <= retries:
+                    self.report_retry(e.source_error, count, retries)
+                continue
+            except NextFragment:
+                continue
+            except SucceedDownload:
+                return True
+
+        self.report_error('giving up after %s retries' % retries)
+        return False