write_string,
)
-_NAME_RE = r'[a-zA-Z_$][\w$]*'
+
+def _js_bit_op(op):
+ def zeroise(x):
+ return 0 if x in (None, JS_Undefined) else x
+
+ def wrapped(a, b):
+ return op(zeroise(a), zeroise(b)) & 0xffffffff
+
+ return wrapped
+
+
+def _js_arith_op(op):
+
+ def wrapped(a, b):
+ if JS_Undefined in (a, b):
+ return float('nan')
+ return op(a or 0, b or 0)
+
+ return wrapped
+
+
+def _js_div(a, b):
+ if JS_Undefined in (a, b) or not (a and b):
+ return float('nan')
+ return (a or 0) / b if b else float('inf')
+
+
+def _js_mod(a, b):
+ if JS_Undefined in (a, b) or not b:
+ return float('nan')
+ return (a or 0) % b
+
+
+def _js_exp(a, b):
+ if not b:
+ return 1 # even 0 ** 0 !!
+ elif JS_Undefined in (a, b):
+ return float('nan')
+ return (a or 0) ** b
+
+
+def _js_eq_op(op):
+
+ def wrapped(a, b):
+ if {a, b} <= {None, JS_Undefined}:
+ return op(a, a)
+ return op(a, b)
+
+ return wrapped
+
+
+def _js_comp_op(op):
+
+ def wrapped(a, b):
+ if JS_Undefined in (a, b):
+ return False
+ if isinstance(a, str) or isinstance(b, str):
+ return op(str(a or 0), str(b or 0))
+ return op(a or 0, b or 0)
+
+ return wrapped
+
+
+def _js_ternary(cndn, if_true=True, if_false=False):
+ """Simulate JS's ternary operator (cndn?if_true:if_false)"""
+ if cndn in (False, None, 0, '', JS_Undefined):
+ return if_false
+ with contextlib.suppress(TypeError):
+ if math.isnan(cndn): # NB: NaN cannot be checked by membership
+ return if_false
+ return if_true
+
# Ref: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Operator_Precedence
_OPERATORS = { # None => Defined in JSInterpreter._operator
'?': None,
-
+ '??': None,
'||': None,
'&&': None,
- '&': operator.and_,
- '|': operator.or_,
- '^': operator.xor,
+
+ '|': _js_bit_op(operator.or_),
+ '^': _js_bit_op(operator.xor),
+ '&': _js_bit_op(operator.and_),
'===': operator.is_,
'!==': operator.is_not,
- '==': operator.eq,
- '!=': operator.ne,
-
- '<=': operator.le,
- '>=': operator.ge,
- '<': operator.lt,
- '>': operator.gt,
+ '==': _js_eq_op(operator.eq),
+ '!=': _js_eq_op(operator.ne),
- '>>': operator.rshift,
- '<<': operator.lshift,
+ '<=': _js_comp_op(operator.le),
+ '>=': _js_comp_op(operator.ge),
+ '<': _js_comp_op(operator.lt),
+ '>': _js_comp_op(operator.gt),
- '+': operator.add,
- '-': operator.sub,
+ '>>': _js_bit_op(operator.rshift),
+ '<<': _js_bit_op(operator.lshift),
- '*': operator.mul,
- '/': operator.truediv,
- '%': operator.mod,
+ '+': _js_arith_op(operator.add),
+ '-': _js_arith_op(operator.sub),
- '**': operator.pow,
+ '*': _js_arith_op(operator.mul),
+ '%': _js_mod,
+ '/': _js_div,
+ '**': _js_exp,
}
_COMP_OPERATORS = {'===', '!==', '==', '!=', '<=', '>=', '<', '>'}
-_MATCHING_PARENS = dict(zip('({[', ')}]'))
-_QUOTES = '\'"'
+_NAME_RE = r'[a-zA-Z_$][\w$]*'
+_MATCHING_PARENS = dict(zip(*zip('()', '{}', '[]')))
+_QUOTES = '\'"/'
-def _ternary(cndn, if_true=True, if_false=False):
- """Simulate JS's ternary operator (cndn?if_true:if_false)"""
- if cndn in (False, None, 0, ''):
- return if_false
- with contextlib.suppress(TypeError):
- if math.isnan(cndn): # NB: NaN cannot be checked by membership
- return if_false
- return if_true
+class JS_Undefined:
+ pass
class JS_Break(ExtractorError):
ExtractorError.__init__(self, 'Invalid continue')
+class JS_Throw(ExtractorError):
+ def __init__(self, e):
+ self.error = e
+ ExtractorError.__init__(self, f'Uncaught exception {e}')
+
+
class LocalNameSpace(collections.ChainMap):
def __setitem__(self, key, value):
for scope in self.maps:
def interpret_statement(self, stmt, local_vars, allow_recursion, *args, **kwargs):
if cls.ENABLED and stmt.strip():
cls.write(stmt, level=allow_recursion)
- ret, should_ret = f(self, stmt, local_vars, allow_recursion, *args, **kwargs)
+ try:
+ ret, should_ret = f(self, stmt, local_vars, allow_recursion, *args, **kwargs)
+ except Exception as e:
+ if cls.ENABLED:
+ if isinstance(e, ExtractorError):
+ e = e.orig_msg
+ cls.write('=> Raises:', e, '<-|', stmt, level=allow_recursion)
+ raise
if cls.ENABLED and stmt.strip():
cls.write(['->', '=>'][should_ret], repr(ret), '<-|', stmt, level=allow_recursion)
return ret, should_ret
class JSInterpreter:
__named_object_counter = 0
+ _RE_FLAGS = {
+ # special knowledge: Python's re flags are bitmask values, current max 128
+ # invent new bitmask values well above that for literal parsing
+ # TODO: new pattern class to execute matches with these flags
+ 'd': 1024, # Generate indices for substring matches
+ 'g': 2048, # Global search
+ 'i': re.I, # Case-insensitive search
+ 'm': re.M, # Multi-line search
+ 's': re.S, # Allows . to match newline characters
+ 'u': re.U, # Treat a pattern as a sequence of unicode code points
+ 'y': 4096, # Perform a "sticky" search that matches starting at the current position in the target string
+ }
+
+ _EXC_NAME = '__yt_dlp_exception__'
+
def __init__(self, code, objects=None):
self.code, self._functions = code, {}
self._objects = {} if objects is None else objects
namespace[name] = obj
return name
+ @classmethod
+ def _regex_flags(cls, expr):
+ flags = 0
+ if not expr:
+ return flags, expr
+ for idx, ch in enumerate(expr):
+ if ch not in cls._RE_FLAGS:
+ break
+ flags |= cls._RE_FLAGS[ch]
+ return flags, expr[idx + 1:]
+
@staticmethod
def _separate(expr, delim=',', max_split=None):
+ OP_CHARS = '+-*/%&|^=<>!,;{}:'
if not expr:
return
counters = {k: 0 for k in _MATCHING_PARENS.values()}
start, splits, pos, delim_len = 0, 0, 0, len(delim) - 1
- in_quote, escaping = None, False
+ in_quote, escaping, after_op, in_regex_char_group = None, False, True, False
for idx, char in enumerate(expr):
if not in_quote and char in _MATCHING_PARENS:
counters[_MATCHING_PARENS[char]] += 1
elif not in_quote and char in counters:
counters[char] -= 1
- elif not escaping and char in _QUOTES and in_quote in (char, None):
- in_quote = None if in_quote else char
+ elif not escaping:
+ if char in _QUOTES and in_quote in (char, None):
+ if in_quote or after_op or char != '/':
+ in_quote = None if in_quote and not in_regex_char_group else char
+ elif in_quote == '/' and char in '[]':
+ in_regex_char_group = char == '['
escaping = not escaping and in_quote and char == '\\'
+ after_op = not in_quote and char in OP_CHARS or (char.isspace() and after_op)
if char != delim[pos] or any(counters.values()) or in_quote:
pos = 0
yield expr[start:]
@classmethod
- def _separate_at_paren(cls, expr, delim):
+ def _separate_at_paren(cls, expr, delim=None):
+ if delim is None:
+ delim = expr and _MATCHING_PARENS[expr[0]]
separated = list(cls._separate(expr, delim, 1))
if len(separated) < 2:
raise cls.Exception(f'No terminating paren {delim}', expr)
def _operator(self, op, left_val, right_expr, expr, local_vars, allow_recursion):
if op in ('||', '&&'):
- if (op == '&&') ^ _ternary(left_val):
+ if (op == '&&') ^ _js_ternary(left_val):
return left_val # short circuiting
+ elif op == '??':
+ if left_val not in (None, JS_Undefined):
+ return left_val
elif op == '?':
- right_expr = _ternary(left_val, *self._separate(right_expr, ':', 1))
+ right_expr = _js_ternary(left_val, *self._separate(right_expr, ':', 1))
right_val = self.interpret_expression(right_expr, local_vars, allow_recursion)
if not _OPERATORS.get(op):
except Exception as e:
raise self.Exception(f'Failed to evaluate {left_val!r} {op} {right_val!r}', expr, cause=e)
- def _index(self, obj, idx):
+ def _index(self, obj, idx, allow_undefined=False):
if idx == 'length':
return len(obj)
try:
return obj[int(idx)] if isinstance(obj, list) else obj[idx]
except Exception as e:
+ if allow_undefined:
+ return JS_Undefined
raise self.Exception(f'Cannot get index {idx}', repr(obj), cause=e)
def _dump(self, obj, namespace):
if should_return:
return ret, should_return
- m = re.match(r'(?P<var>(?:var|const|let)\s)|return(?:\s+|$)', stmt)
+ m = re.match(r'(?P<var>(?:var|const|let)\s)|return(?:\s+|(?=["\'])|$)|(?P<throw>throw\s+)', stmt)
if m:
expr = stmt[len(m.group(0)):].strip()
+ if m.group('throw'):
+ raise JS_Throw(self.interpret_expression(expr, local_vars, allow_recursion))
should_return = not m.group('var')
if not expr:
return None, should_return
if expr[0] in _QUOTES:
inner, outer = self._separate(expr, expr[0], 1)
- inner = json.loads(js_to_json(f'{inner}{expr[0]}', strict=True))
+ if expr[0] == '/':
+ flags, outer = self._regex_flags(outer)
+ inner = re.compile(inner[1:], flags=flags)
+ else:
+ inner = json.loads(js_to_json(f'{inner}{expr[0]}', strict=True))
if not outer:
return inner, should_return
expr = self._named_object(local_vars, inner) + outer
if expr.startswith('new '):
obj = expr[4:]
if obj.startswith('Date('):
- left, right = self._separate_at_paren(obj[4:], ')')
+ left, right = self._separate_at_paren(obj[4:])
expr = unified_timestamp(
self.interpret_expression(left, local_vars, allow_recursion), False)
if not expr:
return None, should_return
if expr.startswith('{'):
- inner, outer = self._separate_at_paren(expr, '}')
+ inner, outer = self._separate_at_paren(expr)
+ # try for object expression (Map)
+ sub_expressions = [list(self._separate(sub_expr.strip(), ':', 1)) for sub_expr in self._separate(inner)]
+ if all(len(sub_expr) == 2 for sub_expr in sub_expressions):
+ def dict_item(key, val):
+ val = self.interpret_expression(val, local_vars, allow_recursion)
+ if re.match(_NAME_RE, key):
+ return key, val
+ return self.interpret_expression(key, local_vars, allow_recursion), val
+
+ return dict(dict_item(k, v) for k, v in sub_expressions), should_return
+
inner, should_abort = self.interpret_statement(inner, local_vars, allow_recursion)
if not outer or should_abort:
return inner, should_abort or should_return
expr = self._dump(inner, local_vars) + outer
if expr.startswith('('):
- inner, outer = self._separate_at_paren(expr, ')')
+ inner, outer = self._separate_at_paren(expr)
inner, should_abort = self.interpret_statement(inner, local_vars, allow_recursion)
if not outer or should_abort:
return inner, should_abort or should_return
expr = self._dump(inner, local_vars) + outer
if expr.startswith('['):
- inner, outer = self._separate_at_paren(expr, ']')
+ inner, outer = self._separate_at_paren(expr)
name = self._named_object(local_vars, [
self.interpret_expression(item, local_vars, allow_recursion)
for item in self._separate(inner)])
expr = name + outer
- m = re.match(r'(?P<try>try|finally)\s*|(?:(?P<catch>catch)|(?P<for>for)|(?P<switch>switch))\s*\(', expr)
- if m and m.group('try'):
- if expr[m.end()] == '{':
- try_expr, expr = self._separate_at_paren(expr[m.end():], '}')
- else:
- try_expr, expr = expr[m.end() - 1:], ''
- ret, should_abort = self.interpret_statement(try_expr, local_vars, allow_recursion)
+ m = re.match(r'''(?x)
+ (?P<try>try)\s*\{|
+ (?P<switch>switch)\s*\(|
+ (?P<for>for)\s*\(
+ ''', expr)
+ md = m.groupdict() if m else {}
+ if md.get('try'):
+ try_expr, expr = self._separate_at_paren(expr[m.end() - 1:])
+ err = None
+ try:
+ ret, should_abort = self.interpret_statement(try_expr, local_vars, allow_recursion)
+ if should_abort:
+ return ret, True
+ except Exception as e:
+ # XXX: This works for now, but makes debugging future issues very hard
+ err = e
+
+ pending = (None, False)
+ m = re.match(r'catch\s*(?P<err>\(\s*{_NAME_RE}\s*\))?\{{'.format(**globals()), expr)
+ if m:
+ sub_expr, expr = self._separate_at_paren(expr[m.end() - 1:])
+ if err:
+ catch_vars = {}
+ if m.group('err'):
+ catch_vars[m.group('err')] = err.error if isinstance(err, JS_Throw) else err
+ catch_vars = local_vars.new_child(catch_vars)
+ err, pending = None, self.interpret_statement(sub_expr, catch_vars, allow_recursion)
+
+ m = re.match(r'finally\s*\{', expr)
+ if m:
+ sub_expr, expr = self._separate_at_paren(expr[m.end() - 1:])
+ ret, should_abort = self.interpret_statement(sub_expr, local_vars, allow_recursion)
+ if should_abort:
+ return ret, True
+
+ ret, should_abort = pending
if should_abort:
return ret, True
- ret, should_abort = self.interpret_statement(expr, local_vars, allow_recursion)
- return ret, should_abort or should_return
- elif m and m.group('catch'):
- # We ignore the catch block
- _, expr = self._separate_at_paren(expr, '}')
- ret, should_abort = self.interpret_statement(expr, local_vars, allow_recursion)
- return ret, should_abort or should_return
+ if err:
+ raise err
- elif m and m.group('for'):
- constructor, remaining = self._separate_at_paren(expr[m.end() - 1:], ')')
+ elif md.get('for'):
+ constructor, remaining = self._separate_at_paren(expr[m.end() - 1:])
if remaining.startswith('{'):
- body, expr = self._separate_at_paren(remaining, '}')
+ body, expr = self._separate_at_paren(remaining)
else:
switch_m = re.match(r'switch\s*\(', remaining) # FIXME
if switch_m:
- switch_val, remaining = self._separate_at_paren(remaining[switch_m.end() - 1:], ')')
+ switch_val, remaining = self._separate_at_paren(remaining[switch_m.end() - 1:])
body, expr = self._separate_at_paren(remaining, '}')
body = 'switch(%s){%s}' % (switch_val, body)
else:
start, cndn, increment = self._separate(constructor, ';')
self.interpret_expression(start, local_vars, allow_recursion)
while True:
- if not _ternary(self.interpret_expression(cndn, local_vars, allow_recursion)):
+ if not _js_ternary(self.interpret_expression(cndn, local_vars, allow_recursion)):
break
try:
ret, should_abort = self.interpret_statement(body, local_vars, allow_recursion)
except JS_Continue:
pass
self.interpret_expression(increment, local_vars, allow_recursion)
- ret, should_abort = self.interpret_statement(expr, local_vars, allow_recursion)
- return ret, should_abort or should_return
- elif m and m.group('switch'):
- switch_val, remaining = self._separate_at_paren(expr[m.end() - 1:], ')')
+ elif md.get('switch'):
+ switch_val, remaining = self._separate_at_paren(expr[m.end() - 1:])
switch_val = self.interpret_expression(switch_val, local_vars, allow_recursion)
body, expr = self._separate_at_paren(remaining, '}')
items = body.replace('default:', 'case default:').split('case ')[1:]
break
if matched:
break
+
+ if md:
ret, should_abort = self.interpret_statement(expr, local_vars, allow_recursion)
return ret, should_abort or should_return
# Comma separated statements
sub_expressions = list(self._separate(expr))
- expr = sub_expressions.pop().strip() if sub_expressions else ''
- for sub_expr in sub_expressions:
- ret, should_abort = self.interpret_statement(sub_expr, local_vars, allow_recursion)
- if should_abort:
- return ret, True
+ if len(sub_expressions) > 1:
+ for sub_expr in sub_expressions:
+ ret, should_abort = self.interpret_statement(sub_expr, local_vars, allow_recursion)
+ if should_abort:
+ return ret, True
+ return ret, False
for m in re.finditer(rf'''(?x)
(?P<pre_sign>\+\+|--)(?P<var1>{_NAME_RE})|
(?P<assign>
(?P<out>{_NAME_RE})(?:\[(?P<index>[^\]]+?)\])?\s*
(?P<op>{"|".join(map(re.escape, set(_OPERATORS) - _COMP_OPERATORS))})?
- =(?P<expr>.*)$
+ =(?!=)(?P<expr>.*)$
)|(?P<return>
- (?!if|return|true|false|null|undefined)(?P<name>{_NAME_RE})$
+ (?!if|return|true|false|null|undefined|NaN)(?P<name>{_NAME_RE})$
)|(?P<indexing>
(?P<in>{_NAME_RE})\[(?P<idx>.+)\]$
)|(?P<attribute>
- (?P<var>{_NAME_RE})(?:\.(?P<member>[^(]+)|\[(?P<member2>[^\]]+)\])\s*
+ (?P<var>{_NAME_RE})(?:(?P<nullish>\?)?\.(?P<member>[^(]+)|\[(?P<member2>[^\]]+)\])\s*
)|(?P<function>
(?P<fname>{_NAME_RE})\((?P<args>.*)\)$
)''', expr)
local_vars[m.group('out')] = self._operator(
m.group('op'), left_val, m.group('expr'), expr, local_vars, allow_recursion)
return local_vars[m.group('out')], should_return
- elif left_val is None:
+ elif left_val in (None, JS_Undefined):
raise self.Exception(f'Cannot index undefined variable {m.group("out")}', expr)
idx = self.interpret_expression(m.group('index'), local_vars, allow_recursion)
raise self.Exception(f'List index {idx} must be integer', expr)
idx = int(idx)
left_val[idx] = self._operator(
- m.group('op'), left_val[idx], m.group('expr'), expr, local_vars, allow_recursion)
+ m.group('op'), self._index(left_val, idx), m.group('expr'), expr, local_vars, allow_recursion)
return left_val[idx], should_return
elif expr.isdigit():
raise JS_Break()
elif expr == 'continue':
raise JS_Continue()
+ elif expr == 'undefined':
+ return JS_Undefined, should_return
+ elif expr == 'NaN':
+ return float('NaN'), should_return
elif m and m.group('return'):
- return local_vars[m.group('name')], should_return
+ return local_vars.get(m.group('name'), JS_Undefined), should_return
with contextlib.suppress(ValueError):
return json.loads(js_to_json(expr, strict=True)), should_return
for op in _OPERATORS:
separated = list(self._separate(expr, op))
right_expr = separated.pop()
- while op in '<>*-' and len(separated) > 1 and not separated[-1].strip():
- separated.pop()
+ while True:
+ if op in '?<>*-' and len(separated) > 1 and not separated[-1].strip():
+ separated.pop()
+ elif not (separated and op == '?' and right_expr.startswith('.')):
+ break
right_expr = f'{op}{right_expr}'
if op != '-':
right_expr = f'{separated.pop()}{op}{right_expr}'
if not separated:
continue
left_val = self.interpret_expression(op.join(separated), local_vars, allow_recursion)
- return self._operator(op, 0 if left_val is None else left_val,
- right_expr, expr, local_vars, allow_recursion), should_return
+ return self._operator(op, left_val, right_expr, expr, local_vars, allow_recursion), should_return
if m and m.group('attribute'):
- variable = m.group('var')
- member = m.group('member')
+ variable, member, nullish = m.group('var', 'member', 'nullish')
if not member:
member = self.interpret_expression(m.group('member2'), local_vars, allow_recursion)
arg_str = expr[m.end():]
if arg_str.startswith('('):
- arg_str, remaining = self._separate_at_paren(arg_str, ')')
+ arg_str, remaining = self._separate_at_paren(arg_str)
else:
arg_str, remaining = None, arg_str
obj = local_vars.get(variable, types.get(variable, NO_DEFAULT))
if obj is NO_DEFAULT:
if variable not in self._objects:
- self._objects[variable] = self.extract_object(variable)
- obj = self._objects[variable]
+ try:
+ self._objects[variable] = self.extract_object(variable)
+ except self.Exception:
+ if not nullish:
+ raise
+ obj = self._objects.get(variable, JS_Undefined)
+
+ if nullish and obj is JS_Undefined:
+ return JS_Undefined
# Member access
if arg_str is None:
- return self._index(obj, member)
+ return self._index(obj, member, nullish)
# Function call
argvals = [
return obj.index(idx, start)
except ValueError:
return -1
+ elif member == 'charCodeAt':
+ assertion(isinstance(obj, str), 'must be applied on a string')
+ assertion(len(argvals) == 1, 'takes exactly one argument')
+ idx = argvals[0] if isinstance(argvals[0], int) else 0
+ if idx >= len(obj):
+ return None
+ return ord(obj[idx])
idx = int(member) if isinstance(obj, list) else member
return obj[idx](argvals, allow_recursion=allow_recursion)
\((?P<args>[^)]*)\)\s*
(?P<code>{.+})''' % {'name': re.escape(funcname)},
self.code)
- code, _ = self._separate_at_paren(func_m.group('code'), '}')
+ code, _ = self._separate_at_paren(func_m.group('code'))
if func_m is None:
raise self.Exception(f'Could not find JS function "{funcname}"')
return [x.strip() for x in func_m.group('args').split(',')], code
if mobj is None:
break
start, body_start = mobj.span()
- body, remaining = self._separate_at_paren(code[body_start - 1:], '}')
+ body, remaining = self._separate_at_paren(code[body_start - 1:])
name = self._named_object(local_vars, self.extract_function_from_code(
[x.strip() for x in mobj.group('args').split(',')],
body, local_vars, *global_stack))
global_stack[0].update(itertools.zip_longest(argnames, args, fillvalue=None))
global_stack[0].update(kwargs)
var_stack = LocalNameSpace(*global_stack)
- ret, should_abort = self.interpret_statement(code.replace('\n', ''), var_stack, allow_recursion - 1)
+ ret, should_abort = self.interpret_statement(code.replace('\n', ' '), var_stack, allow_recursion - 1)
if should_abort:
return ret
return resf