From ba515388b8dc608a07b4390b601e227c1fce6fa9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Sergey=20M=E2=80=A4?= Date: Sun, 28 Jan 2018 13:21:45 +0700 Subject: [PATCH] Introduce --http-chunk-size --- test/test_downloader_http.py | 123 ++++++++++++++++++++++++++++++++++ youtube_dl/__init__.py | 6 ++ youtube_dl/downloader/http.py | 101 ++++++++++++++++++++++------ youtube_dl/options.py | 5 ++ 4 files changed, 213 insertions(+), 22 deletions(-) create mode 100644 test/test_downloader_http.py diff --git a/test/test_downloader_http.py b/test/test_downloader_http.py new file mode 100644 index 000000000..95d152072 --- /dev/null +++ b/test/test_downloader_http.py @@ -0,0 +1,123 @@ +#!/usr/bin/env python +# coding: utf-8 +from __future__ import unicode_literals + +# Allow direct execution +import os +import re +import sys +import unittest +sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) + +from youtube_dl import YoutubeDL +from youtube_dl.compat import compat_http_server +from youtube_dl.downloader.http import HttpFD +from youtube_dl.utils import encodeFilename +import ssl +import threading + +TEST_DIR = os.path.dirname(os.path.abspath(__file__)) + + +def http_server_port(httpd): + if os.name == 'java' and isinstance(httpd.socket, ssl.SSLSocket): + # In Jython SSLSocket is not a subclass of socket.socket + sock = httpd.socket.sock + else: + sock = httpd.socket + return sock.getsockname()[1] + + +TEST_SIZE = 10 * 1024 + + +class HTTPTestRequestHandler(compat_http_server.BaseHTTPRequestHandler): + def log_message(self, format, *args): + pass + + def send_content_range(self, total=None): + range_header = self.headers.get('Range') + start = end = None + if range_header: + mobj = re.search(r'^bytes=(\d+)-(\d+)', range_header) + if mobj: + start = int(mobj.group(1)) + end = int(mobj.group(2)) + valid_range = start is not None and end is not None + if valid_range: + content_range = 'bytes %d-%d' % (start, end) + if total: + content_range += '/%d' % total + self.send_header('Content-Range', content_range) + return (end - start + 1) if valid_range else total + + def serve(self, range=True, content_length=True): + self.send_response(200) + self.send_header('Content-Type', 'video/mp4') + size = TEST_SIZE + if range: + size = self.send_content_range(TEST_SIZE) + if content_length: + self.send_header('Content-Length', size) + self.end_headers() + self.wfile.write(b'#' * size) + + def do_GET(self): + if self.path == '/regular': + self.serve() + elif self.path == '/no-content-length': + self.serve(content_length=False) + elif self.path == '/no-range': + self.serve(range=False) + elif self.path == '/no-range-no-content-length': + self.serve(range=False, content_length=False) + else: + assert False + + +class FakeLogger(object): + def debug(self, msg): + pass + + def warning(self, msg): + pass + + def error(self, msg): + pass + + +class TestHttpFD(unittest.TestCase): + def setUp(self): + self.httpd = compat_http_server.HTTPServer( + ('127.0.0.1', 0), HTTPTestRequestHandler) + self.port = http_server_port(self.httpd) + self.server_thread = threading.Thread(target=self.httpd.serve_forever) + self.server_thread.daemon = True + self.server_thread.start() + + def download(self, params, ep): + params['logger'] = FakeLogger() + ydl = YoutubeDL(params) + downloader = HttpFD(ydl, params) + filename = 'testfile.mp4' + self.assertTrue(downloader.real_download(filename, { + 'url': 'http://127.0.0.1:%d/%s' % (self.port, ep), + })) + self.assertEqual(os.path.getsize(encodeFilename(filename)), TEST_SIZE) + os.remove(encodeFilename(filename)) + + def download_all(self, params): + for ep in ('regular', 'no-content-length', 'no-range', 'no-range-no-content-length'): + self.download(params, ep) + + def test_regular(self): + self.download_all({}) + + def test_chunked(self): + self.download_all({ + 'http_chunk_size': 1000, + }) + + +if __name__ == '__main__': + unittest.main() diff --git a/youtube_dl/__init__.py b/youtube_dl/__init__.py index ba684a075..9bb952457 100644 --- a/youtube_dl/__init__.py +++ b/youtube_dl/__init__.py @@ -191,6 +191,11 @@ def parse_retries(retries): if numeric_buffersize is None: parser.error('invalid buffer size specified') opts.buffersize = numeric_buffersize + if opts.http_chunk_size is not None: + numeric_chunksize = FileDownloader.parse_bytes(opts.http_chunk_size) + if not numeric_chunksize: + parser.error('invalid http chunk size specified') + opts.http_chunk_size = numeric_chunksize if opts.playliststart <= 0: raise ValueError('Playlist start must be positive') if opts.playlistend not in (-1, None) and opts.playlistend < opts.playliststart: @@ -346,6 +351,7 @@ def parse_retries(retries): 'keep_fragments': opts.keep_fragments, 'buffersize': opts.buffersize, 'noresizebuffer': opts.noresizebuffer, + 'http_chunk_size': opts.http_chunk_size, 'continuedl': opts.continue_dl, 'noprogress': opts.noprogress, 'progress_with_newline': opts.progress_with_newline, diff --git a/youtube_dl/downloader/http.py b/youtube_dl/downloader/http.py index 3ff26ff70..dc2b37beb 100644 --- a/youtube_dl/downloader/http.py +++ b/youtube_dl/downloader/http.py @@ -7,10 +7,14 @@ 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, @@ -42,17 +46,22 @@ class DownloadContext(dict): 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 ( + 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() 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 +73,33 @@ 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(): - if ctx.resume_len != 0: - self.report_resuming_byte(ctx.resume_len) - request.add_header('Range', 'bytes=%d-' % ctx.resume_len) + 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 chunk_size > 0: + range_start = 0 + else: + range_start = None + ctx.is_resume = False + range_end = range_start + chunk_size - 1 if 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 + if has_range: + set_range(request, range_start, range_end) # Establish connection try: ctx.data = self.ydl.urlopen(request) @@ -77,25 +108,35 @@ def establish_connection(): # 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: + 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 chunk_size or + # Chunked download and requested piece or + # its part is promised to be served + 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 @@ -130,6 +171,15 @@ def establish_connection(): ctx.resume_len = 0 ctx.open_mode = 'wb' return + elif err.code == 302: + if not chunk_size: + raise + # HTTP Error 302: The HTTP server returned a redirect error that would lead to an infinite loop. + # may happen during chunk downloading. This is usually fixed + # with a retry. + 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 +210,7 @@ def download(): 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 @@ -233,25 +283,30 @@ def retry(e): # 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: break + if not is_test and 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 @@ def retry(e): 'total_bytes': byte_counter, 'filename': ctx.filename, 'status': 'finished', - 'elapsed': time.time() - start, + 'elapsed': time.time() - ctx.start_time, }) return True @@ -290,6 +345,8 @@ def retry(e): if count <= retries: self.report_retry(e.source_error, count, retries) continue + except NextFragment: + continue except SucceedDownload: return True diff --git a/youtube_dl/options.py b/youtube_dl/options.py index 4c0455044..7d1bbc021 100644 --- a/youtube_dl/options.py +++ b/youtube_dl/options.py @@ -478,6 +478,11 @@ def _comma_separated_values_options_callback(option, opt_str, value, parser): '--no-resize-buffer', action='store_true', dest='noresizebuffer', default=False, help='Do not automatically adjust the buffer size. By default, the buffer size is automatically resized from an initial value of SIZE.') + downloader.add_option( + '--http-chunk-size', + dest='http_chunk_size', metavar='SIZE', default=None, + help='Size of a chunk for chunk-based HTTP downloading (e.g. 10485760 or 10M) (default is disabled). ' + 'May be useful for bypassing bandwidth throttling imposed by a webserver (experimental)') downloader.add_option( '--test', action='store_true', dest='test', default=False,