Files
2026-09-15 13:53:58 +02:00

805 lines
37 KiB
Python
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# streaming_providers/base/utils/mpd_rewriter.py
"""
MPD rewriter for DASH manifests.
Handles URL proxying, DRM key injection, quality filtering, and representation blocklisting.
"""
import base64
import struct
import xml.etree.ElementTree as ET
from dataclasses import dataclass
from typing import Optional, Tuple, Set, Dict
from urllib.parse import urljoin, quote, urlencode
from datetime import datetime, timezone
from email.utils import parsedate_to_datetime
from .logger import logger
from .url_resolver import URLResolver
from .drm_key_manager import KeyConfiguration
from .representation_blocklist import RepresentationBlocklist
from .video_quality import VideoQualityFilter, VideoRepresentation
from .time_utils import parse_iso_duration
@dataclass
class _RewriteState:
"""
Per-node traversal state for MPDRewriter._rewrite_node.
This used to be five separate positional parameters threaded through
every recursive call by hand. That's how the $RepresentationID$
substitution bug happened: current_rep_id was tracked as a local
variable in _rewrite_node but never added to the recursive call's
argument list, so it silently reset to None on every recursion step —
meaning it only ever "worked" for attributes living directly on the
<Representation> element itself, never for the <SegmentTemplate> child
where media/initialization templates actually live in practice.
Bundling the state into one object closes off that whole class of bug:
a recursive call either forwards the state object or it doesn't compile
(there's no way to forward "most of" a dataclass instance by accident).
"""
base_url: str
period_id: str = ""
encrypted: bool = False
kid: Optional[str] = None
rep_id: Optional[str] = None
from ..models.drm.constants import DRM_SYSTEM_NAMES
class MPDRewriter:
MPD_NAMESPACE = {"mpd": "urn:mpeg:dash:schema:mpd:2011"}
CENC_NAMESPACE = {"cenc": "urn:mpeg:cenc:2013"}
# Derive from shared DRM constants — format raw hex UUID as urn:uuid:xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
_ck_hex = DRM_SYSTEM_NAMES["CLEARKEY"] # e2719d58a985b3c9781ab030af78d30e
CLEARKEY_SCHEME_URI = (
f"urn:uuid:{_ck_hex[0:8]}-{_ck_hex[8:12]}-{_ck_hex[12:16]}-{_ck_hex[16:20]}-{_ck_hex[20:32]}"
)
def __init__(
self,
media_proxy_url: str,
provider_proxy_url: Optional[str] = None,
clearkey_keyids: Optional[dict] = None,
highest_quality_video_only: bool = False,
provider: Optional[str] = None,
channel: Optional[str] = None,
blocklist_path: str = "representation_blocklist.json",
clearkey_receiver_side: bool = False,
segment_headers: Optional[Dict[str, str]] = None,
):
self.media_proxy_url = media_proxy_url.rstrip("/")
self.provider_proxy_url = provider_proxy_url
self.key_config = KeyConfiguration(clearkey_keyids or {})
self.highest_quality_video_only = highest_quality_video_only
self.clearkey_receiver_side = clearkey_receiver_side
if self.clearkey_receiver_side and not self.key_config.keys:
logger.warning(
"clearkey_receiver_side=True but no keys provided; receiver will have no keys to decrypt with")
# Blocklist configuration
self.provider = provider
self.channel = channel
self.blocklist = RepresentationBlocklist(blocklist_path)
# Pre-calculate query params that don't change to save cycles during rewrite
self._static_params = {}
if self.provider_proxy_url:
self._static_params["proxy"] = self.provider_proxy_url
if segment_headers:
import json
encoded = base64.urlsafe_b64encode(
json.dumps(segment_headers).encode()
).decode().rstrip("=")
self._static_params["headers"] = encoded
@staticmethod
def encode_url(url: str) -> str:
return base64.urlsafe_b64encode(url.encode("utf-8")).decode("utf-8").rstrip("=")
@staticmethod
def decode_url(encoded: str) -> str:
padding = 4 - (len(encoded) % 4)
if padding != 4:
encoded += "=" * padding
return base64.urlsafe_b64decode(encoded.encode("utf-8")).decode("utf-8")
def build_proxy_url(
self,
original_url: str,
template_pattern: Optional[str] = None,
segment_type: Optional[str] = None,
is_encrypted: bool = False,
kid: Optional[str] = None,
representation_id: Optional[str] = None,
) -> str:
params = {"url": original_url, **self._static_params}
if self.key_config.keys and is_encrypted and not self.clearkey_receiver_side:
if self.key_config.single_key_mode:
# Single key mode: use the default key for everything
if segment_type == "initialization":
params["kid"] = self.key_config.default_kid
elif segment_type == "media":
params["key"] = self.key_config.default_key
else:
params["kid"] = self.key_config.default_kid
params["key"] = self.key_config.default_key
else:
# Multi-key mode: use specific KID if provided
if kid and kid in self.key_config.keys:
key = self.key_config.keys[kid]
if segment_type == "initialization":
params["kid"] = kid
elif segment_type == "media":
params["key"] = key
else:
params["kid"] = kid
params["key"] = key
else:
# Fallback to first key
if segment_type == "initialization":
params["kid"] = self.key_config.default_kid
elif segment_type == "media":
params["key"] = self.key_config.default_key
else:
params["kid"] = self.key_config.default_kid
params["key"] = self.key_config.default_key
logger.warning(f"No KID provided for encrypted segment, using fallback key")
encoded = self.encode_url(urlencode(params))
decrypt_server_side = self.key_config.keys and is_encrypted and not self.clearkey_receiver_side
endpoint = "decrypt" if decrypt_server_side else "proxy"
proxy_url = f"{self.media_proxy_url}/api/{endpoint}/{encoded}"
if template_pattern:
# Substitute $RepresentationID$ if we have it and highest_quality_video_only is enabled
if self.highest_quality_video_only and representation_id and "$RepresentationID$" in template_pattern:
template_pattern = template_pattern.replace("$RepresentationID$", representation_id)
proxy_url += f"/{quote(template_pattern, safe='.-_$%')}"
return proxy_url
def rewrite_mpd(self, mpd_content: str, manifest_url: str) -> str:
try:
root = ET.fromstring(mpd_content)
ET.register_namespace("", self.MPD_NAMESPACE["mpd"])
ET.register_namespace("cenc", self.CENC_NAMESPACE["cenc"])
# Extract MPD-level base URL using shared utility
mpd_base_url = self._extract_mpd_base_url(root, manifest_url)
# Single-pass tree preparation with BaseURL extraction
encrypted_ids, as_id_to_kid, base_url_map = self._prepare_tree_and_extract_kids(root, mpd_base_url)
# FIRST: Filter out encrypted AdaptationSets without available keys.
# Only applies to server-side decrypt mode — we can only decrypt what
# we have keys for, so remove sets whose KID we don't have a key for.
#
# Proxy-only mode (no keys): pass everything through unchanged.
# The proxy just forwards bytes; the player uses its own DRM stack
# (Widevine/PlayReady) to decrypt. Stripping encrypted sets in proxy
# mode would break multi-period manifests where the main content is
# encrypted and only a short free bumper is unencrypted.
#
# Receiver-side clearkey (keys + clearkey_receiver_side): keep all
# sets; receiver will decrypt using injected keys.
if self.key_config.keys and not self.clearkey_receiver_side:
# Server-side decrypt: remove sets we can't decrypt
self._remove_adaptationsets_without_keys(root, as_id_to_kid)
# SECOND: Filter out blocked representations that cause 500 errors
if self.provider and self.channel:
self._remove_blocked_representations(root)
# THEN: Filter to highest quality video from remaining decryptable content
best_video_info = None
if self.highest_quality_video_only:
best_video_info = VideoQualityFilter.find_best_video(root)
if best_video_info:
logger.info(
f"Filtered to highest quality video: {best_video_info.width}x{best_video_info.height} "
f"@ {best_video_info.bandwidth}bps, RepID={best_video_info.representation_id}"
)
# Remove all other video representations, keep only the best one
self._filter_to_single_best_video(root, best_video_info)
# Verify we have playable content remaining
remaining_sets = root.findall(".//mpd:AdaptationSet", self.MPD_NAMESPACE)
if not remaining_sets:
raise ValueError("No AdaptationSets remain after key filtering - manifest would be empty")
# Rewrite URLs with appropriate keys and context-aware base URLs
self._rewrite_node(
root, _RewriteState(base_url=mpd_base_url), encrypted_ids, as_id_to_kid, base_url_map
)
rewritten = ET.tostring(root, encoding="unicode", method="xml")
if not rewritten.startswith("<?xml"):
rewritten = '<?xml version="1.0" encoding="UTF-8"?>\n' + rewritten
return rewritten
except Exception as e:
logger.error(f"Failed to rewrite MPD: {e}")
raise
def _filter_to_single_best_video(self, root: ET.Element, best_video: VideoRepresentation):
"""
Remove all video representations except the best one.
Keeps audio/subtitles unchanged.
"""
removal_count = 0
adaptation_sets_to_remove = []
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
period_id = period.get("id", "")
for adaptation_set in period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE):
if not VideoQualityFilter.is_video_adaptation_set(adaptation_set):
continue
as_id = adaptation_set.get("id", str(id(adaptation_set)))
# Check if this adaptation set contains the best representation
representations_to_remove = []
contains_best = False
for representation in adaptation_set.findall("mpd:Representation", self.MPD_NAMESPACE):
rep_id = representation.get("id", "")
# Is this the best representation?
if (period_id == best_video.period_id and
as_id == best_video.as_id and
rep_id == best_video.representation_id):
contains_best = True
else:
representations_to_remove.append(representation)
# Remove non-best representations
for rep in representations_to_remove:
adaptation_set.remove(rep)
removal_count += 1
# If this adaptation set no longer has any representations, mark for removal
if not contains_best:
adaptation_sets_to_remove.append(adaptation_set)
# Remove empty video adaptation sets
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
for adaptation_set in list(period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE)):
if adaptation_set in adaptation_sets_to_remove:
period.remove(adaptation_set)
logger.info(f"Removed {removal_count} video representation(s), kept highest quality only")
# Clean up SupplementalProperty elements that reference removed AdaptationSet IDs.
# These are used for codec-switching hints (e.g. AVC↔HEVC) and become dangling
# after we discard all but one video AdaptationSet.
self._cleanup_codec_switch_properties(root)
def _cleanup_codec_switch_properties(self, root: ET.Element):
"""Remove or update SupplementalProperty elements for codec switching."""
CODEC_SWITCH_SCHEME = "urn:mpeg:dash:adaptation-set-switching:2016"
# Collect all surviving AdaptationSet IDs
surviving_as_ids: Set[str] = set()
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
for adaptation_set in period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE):
as_id = adaptation_set.get("id")
if as_id:
surviving_as_ids.add(as_id)
# Clean up SupplementalProperty elements
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
for adaptation_set in period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE):
for sp in list(adaptation_set.findall("mpd:SupplementalProperty", self.MPD_NAMESPACE)):
if sp.get("schemeIdUri") == CODEC_SWITCH_SCHEME:
raw_value = sp.get("value", "")
# Keep only IDs that still exist in the tree
kept = [
v.strip()
for v in raw_value.split(",")
if v.strip() in surviving_as_ids
]
if not kept:
# No surviving peers — remove the property entirely
adaptation_set.remove(sp)
else:
sp.set("value", ",".join(kept))
def _prepare_tree_and_extract_kids(self, root: ET.Element, mpd_base_url: str) -> Tuple[
Set[str], Dict[str, str], Dict[str, str]]:
"""
Single-pass optimization: clean tree, identify encrypted sets, extract KIDs and BaseURLs.
Returns: (encrypted_adaptation_set_ids, as_id_to_kid_mapping, base_url_mapping)
"""
encrypted_ids = set()
as_id_to_kid = {}
base_url_map = {} # Maps period_id:as_id -> resolved base URL
# Process all Periods (handles multi-period manifests correctly)
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
period_id = period.get("id", "")
# Extract Period-level BaseURL BEFORE removing it
period_base_url = mpd_base_url
period_base_elem = period.find("mpd:BaseURL", self.MPD_NAMESPACE)
if period_base_elem is not None and period_base_elem.text:
period_base_text = period_base_elem.text.strip()
period_base_url = self._urljoin_preserve_query(mpd_base_url, period_base_text)
logger.debug(f"Period {period_id} BaseURL: {period_base_url}")
# Store period-level base URL in base_url_map with a special key
period_key = f"period_{period_id}" if period_id else "period_root"
base_url_map[period_key] = period_base_url
# Remove Period-level BaseURL elements
for bu in list(period.findall("mpd:BaseURL", self.MPD_NAMESPACE)):
period.remove(bu)
for adaptation_set in period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE):
as_id = adaptation_set.get("id")
if not as_id:
as_id = str(id(adaptation_set))
unique_id = f"{period_id}_{as_id}" if period_id else as_id
# Extract AdaptationSet-level BaseURL BEFORE removing it
as_base_url = period_base_url
as_base_elem = adaptation_set.find("mpd:BaseURL", self.MPD_NAMESPACE)
if as_base_elem is not None and as_base_elem.text:
as_base_text = as_base_elem.text.strip()
as_base_url = self._urljoin_preserve_query(period_base_url, as_base_text)
logger.debug(f"AdaptationSet {unique_id} BaseURL: {as_base_url}")
# Store the resolved base URL for this AdaptationSet
base_url_map[unique_id] = as_base_url
# Remove AdaptationSet-level BaseURL elements
for bu in list(adaptation_set.findall("mpd:BaseURL", self.MPD_NAMESPACE)):
adaptation_set.remove(bu)
# Check if encrypted
cp_elements = adaptation_set.findall("mpd:ContentProtection", self.MPD_NAMESPACE)
has_content_protection = len(cp_elements) > 0
# Also check Representation-level ContentProtection
for representation in adaptation_set.findall("mpd:Representation", self.MPD_NAMESPACE):
rep_cp = representation.findall("mpd:ContentProtection", self.MPD_NAMESPACE)
if rep_cp:
has_content_protection = True
break
if has_content_protection:
encrypted_ids.add(unique_id)
# Extract KID
extracted_kid = self._extract_kid_from_adaptationset(adaptation_set)
if extracted_kid:
as_id_to_kid[unique_id] = extracted_kid
# logger.debug(f"AdaptationSet {unique_id} KID: {extracted_kid[:8]}...")
# Handle ContentProtection elements based on mode:
# - server-side decrypt: strip all (player sees plain stream)
# - receiver-side clearkey: replace with ClearKey signaling so player can decrypt
# - proxy-only (no keys): preserve as-is so player can handle DRM itself
if self.key_config.keys and not self.clearkey_receiver_side:
# Server-side decrypt mode: strip everything
for cp in list(adaptation_set.findall("mpd:ContentProtection", self.MPD_NAMESPACE)):
adaptation_set.remove(cp)
for representation in adaptation_set.findall("mpd:Representation", self.MPD_NAMESPACE):
for cp in list(representation.findall("mpd:ContentProtection", self.MPD_NAMESPACE)):
representation.remove(cp)
elif self.clearkey_receiver_side and has_content_protection:
# Receiver-side clearkey mode: strip original DRM signaling, inject ClearKey
for cp in list(adaptation_set.findall("mpd:ContentProtection", self.MPD_NAMESPACE)):
adaptation_set.remove(cp)
for representation in adaptation_set.findall("mpd:Representation", self.MPD_NAMESPACE):
for cp in list(representation.findall("mpd:ContentProtection", self.MPD_NAMESPACE)):
representation.remove(cp)
# Inject ClearKey ContentProtection element
kid = as_id_to_kid.get(unique_id)
self._inject_clearkey_content_protection(adaptation_set, kid)
# else: proxy-only, no keys — leave ContentProtection untouched
return encrypted_ids, as_id_to_kid, base_url_map
def _extract_kid_from_adaptationset(self, adaptation_set: ET.Element) -> Optional[str]:
"""Extract KID from ContentProtection elements in an AdaptationSet."""
# Try cenc:default_KID first
for cp in adaptation_set.findall("mpd:ContentProtection", self.MPD_NAMESPACE):
default_kid = cp.get("{urn:mpeg:cenc:2013}default_KID")
if default_kid:
normalized = default_kid.replace("-", "").lower()
# logger.debug(f"Extracted KID from default_KID attribute: {normalized}")
return normalized
# Try PSSH box
for cp in adaptation_set.findall("mpd:ContentProtection", self.MPD_NAMESPACE):
pssh_elem = cp.find("cenc:pssh", self.CENC_NAMESPACE)
if pssh_elem is not None and pssh_elem.text:
try:
pssh_data = base64.b64decode(pssh_elem.text)
logger.debug(f"PSSH box size: {len(pssh_data)} bytes")
if len(pssh_data) >= 36:
version = pssh_data[8] if len(pssh_data) > 8 else 0
system_id = pssh_data[12:28].hex() if len(pssh_data) >= 28 else "unknown"
logger.debug(f"PSSH version: {version}, System ID: {system_id}")
if version > 0:
kid_count = struct.unpack(">I", pssh_data[28:32])[0]
logger.debug(f"KID count from header: {kid_count}")
if kid_count > 0 and len(pssh_data) >= 48:
kid_bytes = pssh_data[32:48]
kid_hex = kid_bytes.hex().lower()
# logger.debug(f"Extracted KID from PSSH header: {kid_hex}")
return kid_hex
except Exception as e:
logger.debug(f"Error extracting KID from PSSH: {e}")
return None
def _inject_clearkey_content_protection(self, adaptation_set: ET.Element, kid: Optional[str]):
"""
Inject a ClearKey ContentProtection element so the receiver can decrypt.
If a KID is available (extracted from the original ContentProtection), it is
formatted as a UUID and set as cenc:default_KID. Otherwise only the scheme URI
is signaled and the receiver must derive the KID from the stream itself.
"""
ET.register_namespace("cenc", "urn:mpeg:cenc:2013")
attribs = {"schemeIdUri": self.CLEARKEY_SCHEME_URI}
if kid:
# Format 32-char hex KID as UUID (xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx)
kid_uuid = f"{kid[0:8]}-{kid[8:12]}-{kid[12:16]}-{kid[16:20]}-{kid[20:32]}"
attribs["{urn:mpeg:cenc:2013}default_KID"] = kid_uuid
logger.debug(f"Injecting ClearKey ContentProtection with KID={kid_uuid}")
else:
logger.debug("Injecting ClearKey ContentProtection without KID (not found in original)")
cp_elem = ET.Element("ContentProtection", attribs)
# Insert at the front of the AdaptationSet so players encounter it early
adaptation_set.insert(0, cp_elem)
def _remove_adaptationsets_without_keys(self, root: ET.Element, as_id_to_kid: Dict[str, str]):
"""Remove encrypted AdaptationSets for which we don't have keys."""
removed_count = 0
# Debug: Log what keys we have
# logger.debug(f"Available keys: {list(self.key_config.keys.keys())}")
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
period_id = period.get("id", "")
adaptationsets_to_remove = []
for adaptation_set in period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE):
as_id = adaptation_set.get("id", str(id(adaptation_set)))
unique_id = f"{period_id}_{as_id}" if period_id else as_id
if unique_id in as_id_to_kid:
kid = as_id_to_kid[unique_id]
if kid not in self.key_config.keys:
logger.warning(
f"Removing AdaptationSet {unique_id} - "
f"missing key for KID: {kid[:8]}..."
)
adaptationsets_to_remove.append(adaptation_set)
removed_count += 1
for adaptation_set in adaptationsets_to_remove:
period.remove(adaptation_set)
if removed_count > 0:
logger.info(f"Removed {removed_count} AdaptationSet(s) without available keys")
def _remove_blocked_representations(self, root: ET.Element):
"""Remove representations that are blocklisted for this provider/channel."""
blocked_ids = self.blocklist.get_blocked_ids(self.provider, self.channel)
if not blocked_ids:
return
logger.info(
f"Applying representation blocklist for {self.provider}/{self.channel}: "
f"{len(blocked_ids)} representation(s) blocked"
)
total_reps_removed = 0
total_as_removed = 0
for period in root.findall(".//mpd:Period", self.MPD_NAMESPACE):
adaptationsets_to_remove = []
for adaptation_set in period.findall("mpd:AdaptationSet", self.MPD_NAMESPACE):
representations = adaptation_set.findall("mpd:Representation", self.MPD_NAMESPACE)
representations_to_remove = []
# Check each representation in this AdaptationSet
for representation in representations:
rep_id = representation.get("id", "")
if rep_id in blocked_ids:
logger.info(
f"Blocking representation '{rep_id}' for {self.provider}/{self.channel} "
f"(known to cause 500 errors)"
)
representations_to_remove.append(representation)
# If all representations are blocked, mark the entire AdaptationSet for removal
if representations_to_remove and len(representations_to_remove) == len(representations):
as_id = adaptation_set.get("id", "unknown")
logger.info(
f"Removing entire AdaptationSet '{as_id}' - "
f"all {len(representations)} representation(s) are blocked"
)
adaptationsets_to_remove.append(adaptation_set)
total_as_removed += 1
else:
# Remove only the blocked representations
for representation in representations_to_remove:
adaptation_set.remove(representation)
total_reps_removed += 1
# Remove marked AdaptationSets
for adaptation_set in adaptationsets_to_remove:
period.remove(adaptation_set)
if total_reps_removed > 0 or total_as_removed > 0:
logger.info(
f"Blocklist filtering complete: removed {total_reps_removed} representation(s) "
f"and {total_as_removed} AdaptationSet(s) for {self.provider}/{self.channel}"
)
def _rewrite_node(
self,
element: ET.Element,
state: _RewriteState,
encrypted_ids: Set[str],
as_id_to_kid: Dict[str, str],
base_url_map: Dict[str, str],
):
"""Recursive node rewriter with KID-aware key selection and context-aware base URLs."""
# Track period ID as we traverse
if element.tag.endswith("Period"):
period_id = element.get("id", "")
base_url = state.base_url
# When entering a new period, check if we have a period-level base URL stored
period_key = f"period_{period_id}" if period_id else "period_root"
if period_key in base_url_map:
# Update base_url to the period-specific base URL
base_url = base_url_map[period_key]
# logger.debug(f"Period {period_id} using stored base URL: {base_url}")
child_state = _RewriteState(
base_url=base_url,
period_id=period_id,
encrypted=state.encrypted,
kid=state.kid,
rep_id=state.rep_id,
)
# Process all children of this period
for child in list(element):
self._rewrite_node(child, child_state, encrypted_ids, as_id_to_kid, base_url_map)
return # Don't process further - we've handled all children
# Track representation ID for template substitution
rep_id = state.rep_id
if element.tag.endswith("Representation"):
rep_id = element.get("id", "")
base_url = state.base_url
encrypted = state.encrypted
kid = state.kid
# Update state when entering an AdaptationSet
if element.tag.endswith("AdaptationSet"):
as_id = element.get("id", str(id(element)))
# Use same unique ID logic as _prepare_tree_and_extract_kids
unique_id = f"{state.period_id}_{as_id}" if state.period_id else as_id
encrypted = unique_id in encrypted_ids
# Update base_url to the AdaptationSet-specific base URL
if unique_id in base_url_map:
base_url = base_url_map[unique_id]
# logger.debug(f"AdaptationSet {unique_id} using stored base URL: {base_url}")
# Get specific KID for this AdaptationSet (multi-key mode only)
if encrypted and not self.key_config.single_key_mode:
kid = as_id_to_kid.get(unique_id)
# This is the state that applies to *this* element and everything
# below it — including rep_id, which is the piece that previously
# got dropped on recursion (see _RewriteState docstring).
current_state = _RewriteState(
base_url=base_url,
period_id=state.period_id,
encrypted=encrypted,
kid=kid,
rep_id=rep_id,
)
# Handle BaseURL elements - THESE MUST BE REWRITTEN TO PROXY URLS
if element.tag.endswith("BaseURL") and element.text:
raw_url = element.text.strip()
if raw_url:
# Resolve the relative BaseURL against the current base
resolved_cdn_url = self._urljoin_preserve_query(current_state.base_url, raw_url)
# Rewrite the BaseURL text to a proxy URL
# In receiver-side clearkey mode, this will use /api/proxy/
element.text = self.build_proxy_url(
resolved_cdn_url, None, None,
current_state.encrypted, current_state.kid,
representation_id=current_state.rep_id
)
# ------------------------------------------------------------------
# Standard SegmentTemplate / SegmentList attribute rewriting.
# Handles media/initialization/sourceURL attributes on SegmentTemplate,
# SegmentList, etc.
#
attr_map = {
"media": "media",
"initialization": "initialization",
"sourceURL": None,
}
for attr, seg_type in attr_map.items():
if attr in element.attrib:
val = element.attrib[attr]
if not val:
continue
resolved = self._urljoin_preserve_query(current_state.base_url, val)
if "$" in resolved:
# Use shared utility for splitting template URLs
path, pattern = URLResolver.split_template_url(resolved)
element.attrib[attr] = self.build_proxy_url(
path, pattern, seg_type, current_state.encrypted, current_state.kid,
representation_id=current_state.rep_id
)
else:
element.attrib[attr] = self.build_proxy_url(
resolved, None, seg_type, current_state.encrypted, current_state.kid,
representation_id=current_state.rep_id
)
# Handle SegmentURL (always 'media' type)
if element.tag.endswith("SegmentURL") and "media" in element.attrib:
resolved = self._urljoin_preserve_query(current_state.base_url, element.attrib["media"])
path, pattern = (
URLResolver.split_template_url(resolved)
if "$" in resolved
else (resolved, None)
)
element.attrib["media"] = self.build_proxy_url(
path, pattern, "media", current_state.encrypted, current_state.kid,
representation_id=current_state.rep_id
)
# Recurse to children, forwarding the full state (including rep_id)
for child in element:
self._rewrite_node(child, current_state, encrypted_ids, as_id_to_kid, base_url_map)
@staticmethod
def _urljoin_preserve_query(base: str, url: str) -> str:
"""
Like urljoin(base, url) but preserves the query string of *base* when
the relative *url* itself carries no query string.
Why this is needed:
Some CDNs (e.g. Discovery+/Max) embed the auth token in the manifest
URL's query string (e.g. ?manifest-params=…). The segment paths
inside such manifests are plain relative paths like "v/0_9cd1a5/v13.mp4"
with no query component. Standard urljoin() discards the base query
string when joining with a relative path (RFC 3986 §5.3 correct
behaviour in general, wrong here). The CDN requires the same token on
every segment request, so we must re-append it.
Rules (mirrors RFC 3986 §5.3 reference-resolution):
- If *url* is absolute (has a scheme) → plain urljoin, no query merging.
- If *url* already has its own query string → plain urljoin; the
relative URL's query takes full precedence.
- If *url* has no query string and *base* does → urljoin then re-attach
base's query string to the result.
"""
from urllib.parse import urlparse, urlunparse
parsed_url = urlparse(url)
# Absolute URL: standard behaviour, no query merging.
if parsed_url.scheme:
return urljoin(base, url)
# URL contains $ template variables (e.g. $RepresentationID$, $Number$):
# these are SegmentTemplate patterns that manage their own query strings.
# Never re-attach the base query — it would be injected into the middle of
# the template pattern and corrupt the CDN URL the proxy encodes.
if "$" in url:
return urljoin(base, url)
# URL already has its own query string — it takes full precedence.
if parsed_url.query:
return urljoin(base, url)
joined = urljoin(base, url)
# Re-attach the base query string if it has one and the result lost it.
# This handles CDNs that embed auth tokens in the manifest URL query string
# (e.g. ?manifest-params=...) which must be forwarded to every segment URL.
base_query = urlparse(base).query
if base_query:
parsed_joined = urlparse(joined)
if not parsed_joined.query:
# urlunparse is typed as returning str | bytes depending on input;
# our inputs are always str so cast to make the type checker happy.
joined = str(urlunparse(parsed_joined._replace(query=base_query)))
return joined
def _extract_mpd_base_url(self, root: ET.Element, manifest_url: str) -> str:
"""
Extract and resolve MPD-level BaseURL.
Uses shared URLResolver utility.
"""
base_url_elem = root.find("mpd:BaseURL", self.MPD_NAMESPACE)
base_url_text = None
if base_url_elem is not None and base_url_elem.text:
base_url_text = base_url_elem.text.strip()
# Use shared utility - it handles special service prefixes.
# Note: URLResolver intentionally strips the manifest URL's query string
# when deriving the base directory URL. This is correct — query strings
# like ?manifest-params= are only for the .mpd fetch itself and must NOT
# be forwarded to segment requests.
return URLResolver.resolve_base_url_with_element(manifest_url, base_url_text)
@staticmethod
def extract_cache_ttl(headers: dict) -> int:
cache_control = headers.get("Cache-Control", headers.get("cache-control", ""))
if "max-age=" in cache_control:
try:
for directive in cache_control.split(","):
directive = directive.strip()
if directive.startswith("max-age="):
return int(directive.split("=")[1])
except (ValueError, IndexError):
pass
expires = headers.get("Expires", headers.get("expires"))
if expires:
try:
expires_dt = parsedate_to_datetime(expires)
now = datetime.now(timezone.utc)
ttl = int((expires_dt - now).total_seconds())
if ttl > 0:
return ttl
except Exception:
pass
return 300
@staticmethod
def extract_mpd_update_period(mpd_content: str) -> Optional[int]:
try:
root = ET.fromstring(mpd_content)
if root.attrib.get("type") == "dynamic":
update_period = root.attrib.get("minimumUpdatePeriod")
if update_period:
return parse_iso_duration(update_period)
except Exception:
pass
return None