[nytimes] improve extraction

youtube_dl/extractor/nytimes.py

@@ -1,26 +1,37 @@
 from __future__ import unicode_literals
 
+import hmac
+import hashlib
+import base64
+
 from .common import InfoExtractor
 from ..utils import (
     float_or_none,
     int_or_none,
     parse_iso8601,
+    mimetype2ext,
+    determine_ext,
 )
 
 
 class NYTimesBaseIE(InfoExtractor):
+    _SECRET = b'pX(2MbU2);4N{7J8)>YwKRJ+/pQ3JkiU2Q^V>mFYv6g6gYvt6v'
+
     def _extract_video_from_id(self, video_id):
+        # Authorization generation algorithm is reverse engineered from `signer` in
+        # http://graphics8.nytimes.com/video/vhs/vhs-2.x.min.js
+        path = '/svc/video/api/v3/video/' + video_id
+        hm = hmac.new(self._SECRET, (path + ':vhs').encode(), hashlib.sha512).hexdigest()
+        video_data = self._download_json('http://www.nytimes.com' + path, video_id, 'Downloading video JSON', headers={
+            'Authorization': 'NYTV ' + base64.b64encode(hm.encode()).decode(),
+            'X-NYTV': 'vhs',
+        }, fatal=False)
+        if not video_data:
             video_data = self._download_json(
-            'http://www.nytimes.com/svc/video/api/v2/video/%s' % video_id,
+                'http://www.nytimes.com/svc/video/api/v2/video/' + video_id,
                 video_id, 'Downloading video JSON')
 
         title = video_data['headline']
-        description = video_data.get('summary')
-        duration = float_or_none(video_data.get('duration'), 1000)
-
-        uploader = video_data.get('byline')
-        publication_date = video_data.get('publication_date')
-        timestamp = parse_iso8601(publication_date[:-8]) if publication_date else None
 
         def get_file_size(file_size):
             if isinstance(file_size, int):
@@ -28,35 +39,59 @@ class NYTimesBaseIE(InfoExtractor):
             elif isinstance(file_size, dict):
                 return int(file_size.get('value', 0))
             else:
-                return 0
+                return None
 
-        formats = [
+        urls = []
-            {
+        formats = []
-                'url': video['url'],
+        for video in video_data.get('renditions', []):
-                'format_id': video.get('type'),
+            video_url = video.get('url')
-                'vcodec': video.get('video_codec'),
+            format_id = video.get('type')
+            if not video_url or format_id == 'thumbs' or video_url in urls:
+                continue
+            urls.append(video_url)
+            ext = mimetype2ext(video.get('mimetype')) or determine_ext(video_url)
+            if ext == 'm3u8':
+                formats.extend(self._extract_m3u8_formats(
+                    video_url, video_id, 'mp4', 'm3u8_native',
+                    m3u8_id=format_id or 'hls', fatal=False))
+            elif ext == 'mpd':
+                continue
+            #     formats.extend(self._extract_mpd_formats(
+            #         video_url, video_id, format_id or 'dash', fatal=False))
+            else:
+                formats.append({
+                    'url': video_url,
+                    'format_id': format_id,
+                    'vcodec': video.get('videoencoding') or video.get('video_codec'),
                     'width': int_or_none(video.get('width')),
                     'height': int_or_none(video.get('height')),
-                'filesize': get_file_size(video.get('fileSize')),
+                    'filesize': get_file_size(video.get('file_size') or video.get('fileSize')),
-            } for video in video_data['renditions'] if video.get('url')
+                    'tbr': int_or_none(video.get('bitrate'), 1000),
-        ]
+                    'ext': ext,
+                })
         self._sort_formats(formats)
 
-        thumbnails = [
+        thumbnails = []
-            {
+        for image in video_data.get('images', []):
-                'url': 'http://www.nytimes.com/%s' % image['url'],
+            image_url = image.get('url')
+            if not image_url:
+                continue
+            thumbnails.append({
+                'url': 'http://www.nytimes.com/' + image_url,
                 'width': int_or_none(image.get('width')),
                 'height': int_or_none(image.get('height')),
-            } for image in video_data.get('images', []) if image.get('url')
+            })
-        ]
+
+        publication_date = video_data.get('publication_date')
+        timestamp = parse_iso8601(publication_date[:-8]) if publication_date else None
 
         return {
             'id': video_id,
             'title': title,
-            'description': description,
+            'description': video_data.get('summary'),
             'timestamp': timestamp,
-            'uploader': uploader,
+            'uploader': video_data.get('byline'),
-            'duration': duration,
+            'duration': float_or_none(video_data.get('duration'), 1000),
             'formats': formats,
             'thumbnails': thumbnails,
         }
@@ -67,7 +102,7 @@ class NYTimesIE(NYTimesBaseIE):
 
     _TESTS = [{
         'url': 'http://www.nytimes.com/video/opinion/100000002847155/verbatim-what-is-a-photocopier.html?playlistId=100000001150263',
-        'md5': '18a525a510f942ada2720db5f31644c0',
+        'md5': 'd665342765db043f7e225cff19df0f2d',
         'info_dict': {
             'id': '100000002847155',
             'ext': 'mov',