]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/downloader/http.py
Update changelog.
[youtubedl] / youtube_dl / downloader / http.py
index 8a6638cc2d47c4afa08104d77e0877f92749ac8a..5046878dfcd874013e737e85d32764a95737406e 100644 (file)
@@ -4,13 +4,18 @@ import errno
 import os
 import socket
 import time
+import random
 import re
 
 from .common import FileDownloader
-from ..compat import compat_urllib_error
+from ..compat import (
+    compat_str,
+    compat_urllib_error,
+)
 from ..utils import (
     ContentTooShortError,
     encodeFilename,
+    int_or_none,
     sanitize_open,
     sanitized_Request,
     write_xattr,
@@ -38,21 +43,26 @@ class HttpFD(FileDownloader):
         add_headers = info_dict.get('http_headers')
         if add_headers:
             headers.update(add_headers)
-        basic_request = sanitized_Request(url, None, headers)
-        request = sanitized_Request(url, None, headers)
 
         is_test = self.params.get('test', False)
-
-        if is_test:
-            request.add_header('Range', 'bytes=0-%s' % str(self._TEST_FILE_SIZE - 1))
+        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)
 
         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.resume_len = os.path.getsize(
+                    encodeFilename(ctx.tmpfilename))
+
+        ctx.is_resume = ctx.resume_len > 0
 
         count = 0
         retries = self.params.get('retries', 0)
@@ -64,11 +74,36 @@ class HttpFD(FileDownloader):
             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():
-            if ctx.resume_len != 0:
-                self.report_resuming_byte(ctx.resume_len)
-                request.add_header('Range', 'bytes=%d-' % ctx.resume_len)
+            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:
                 ctx.data = self.ydl.urlopen(request)
@@ -76,38 +111,49 @@ class HttpFD(FileDownloader):
                 # 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/rg3/youtube-dl/issues/6057#issuecomment-126129799)
-                if ctx.resume_len > 0:
+                # 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+)-', 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 and ctx.resume_len == int(content_range_m.group(1)):
-                            return
+                        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:
-                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
-                        ctx.data = self.ydl.urlopen(basic_request)
+                        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
-                        if (content_length is not None and
-                                (ctx.resume_len - 100 < int(content_length) < ctx.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,
@@ -130,6 +176,9 @@ class HttpFD(FileDownloader):
                             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:
@@ -160,7 +209,7 @@ class HttpFD(FileDownloader):
                     return False
 
             byte_counter = 0 + ctx.resume_len
-            block_size = self.params.get('buffersize', 1024)
+            block_size = ctx.block_size
             start = time.time()
 
             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
@@ -168,16 +217,17 @@ class HttpFD(FileDownloader):
             before = start  # start measuring
 
             def retry(e):
-                if ctx.tmpfilename != '-':
+                to_stdout = ctx.tmpfilename == '-'
+                if not to_stdout:
                     ctx.stream.close()
                 ctx.stream = None
-                ctx.resume_len = os.path.getsize(encodeFilename(ctx.tmpfilename))
+                ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
                 raise RetryDownload(e)
 
             while True:
                 try:
                     # Download and write
-                    data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
+                    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:
@@ -233,25 +283,30 @@ class HttpFD(FileDownloader):
 
                 # Progress message
                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
-                if data_len is None:
+                if ctx.data_len is None:
                     eta = None
                 else:
-                    eta = self.calc_eta(start, time.time(), data_len - ctx.resume_len, byte_counter - ctx.resume_len)
+                    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': data_len,
+                    'total_bytes': ctx.data_len,
                     'tmpfilename': ctx.tmpfilename,
                     'filename': ctx.filename,
                     'eta': eta,
                     'speed': speed,
-                    'elapsed': now - start,
+                    'elapsed': now - ctx.start_time,
                 })
 
-                if is_test and byte_counter == data_len:
+                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')
@@ -276,7 +331,7 @@ class HttpFD(FileDownloader):
                 'total_bytes': byte_counter,
                 'filename': ctx.filename,
                 'status': 'finished',
-                'elapsed': time.time() - start,
+                'elapsed': time.time() - ctx.start_time,
             })
 
             return True
@@ -284,13 +339,14 @@ class HttpFD(FileDownloader):
         while count <= retries:
             try:
                 establish_connection()
-                download()
-                return True
+                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