import collections import contextlib import itertools import json import math import operator import re from .utils import ( NO_DEFAULT, ExtractorError, js_to_json, remove_quotes, truncate_string, unified_timestamp, write_string, ) def _js_bit_op(op): def wrapped(a, b): def zeroise(x): return 0 if x in (None, JS_Undefined) else x return op(zeroise(a), zeroise(b)) 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 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, '|': _js_bit_op(operator.or_), '^': _js_bit_op(operator.xor), '&': _js_bit_op(operator.and_), '===': operator.is_, '==': _js_eq_op(operator.eq), '!==': operator.is_not, '!=': _js_eq_op(operator.ne), '<=': _js_comp_op(operator.le), '>=': _js_comp_op(operator.ge), '<': _js_comp_op(operator.lt), '>': _js_comp_op(operator.gt), '>>': _js_bit_op(operator.rshift), '<<': _js_bit_op(operator.lshift), '+': _js_arith_op(operator.add), '-': _js_arith_op(operator.sub), '*': _js_arith_op(operator.mul), '/': _js_div, '%': _js_mod, '**': _js_exp, } _COMP_OPERATORS = {'===', '!==', '==', '!=', '<=', '>=', '<', '>'} _NAME_RE = r'[a-zA-Z_$][\w$]*' _MATCHING_PARENS = dict(zip(*zip('()', '{}', '[]'))) _QUOTES = '\'"/' class JS_Undefined: pass class JS_Break(ExtractorError): def __init__(self): ExtractorError.__init__(self, 'Invalid break') class JS_Continue(ExtractorError): def __init__(self): 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: if key in scope: scope[key] = value return self.maps[0][key] = value def __delitem__(self, key): raise NotImplementedError('Deleting is not supported') class Debugger: import sys ENABLED = False and 'pytest' in sys.modules @staticmethod def write(*args, level=100): write_string(f'[debug] JS: {" " * (100 - level)}' f'{" ".join(truncate_string(str(x), 50, 50) for x in args)}\n') @classmethod def wrap_interpreter(cls, f): 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) if cls.ENABLED and stmt.strip(): cls.write(['->', '=>'][should_ret], repr(ret), '<-|', stmt, level=allow_recursion) return ret, should_ret return interpret_statement 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 class Exception(ExtractorError): def __init__(self, msg, expr=None, *args, **kwargs): if expr is not None: msg = f'{msg.rstrip()} in: {truncate_string(expr, 50, 50)}' super().__init__(msg, *args, **kwargs) def _named_object(self, namespace, obj): self.__named_object_counter += 1 name = f'__yt_dlp_jsinterp_obj{self.__named_object_counter}' 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, 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): 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 == ' ' and after_op) if char != delim[pos] or any(counters.values()) or in_quote: pos = 0 continue elif pos != delim_len: pos += 1 continue yield expr[start: idx - delim_len] start, pos = idx + 1, 0 splits += 1 if max_split and splits >= max_split: break yield expr[start:] @classmethod def _separate_at_paren(cls, expr, delim): separated = list(cls._separate(expr, delim, 1)) if len(separated) < 2: raise cls.Exception(f'No terminating paren {delim}', expr) return separated[0][1:].strip(), separated[1].strip() def _operator(self, op, left_val, right_expr, expr, local_vars, allow_recursion): if op in ('||', '&&'): 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 = _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): return right_val try: return _OPERATORS[op](left_val, right_val) 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, 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): try: return json.dumps(obj) except TypeError: return self._named_object(namespace, obj) @Debugger.wrap_interpreter def interpret_statement(self, stmt, local_vars, allow_recursion=100): if allow_recursion < 0: raise self.Exception('Recursion limit reached') allow_recursion -= 1 should_return = False sub_statements = list(self._separate(stmt, ';')) or [''] expr = stmt = sub_statements.pop().strip() for sub_stmt in sub_statements: ret, should_return = self.interpret_statement(sub_stmt, local_vars, allow_recursion) if should_return: return ret, should_return m = re.match(r'(?P(?:var|const|let)\s)|return(?:\s+|(?=["\'])|$)|(?Pthrow\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) 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:], ')') expr = unified_timestamp( self.interpret_expression(left, local_vars, allow_recursion), False) if not expr: raise self.Exception(f'Failed to parse date {left!r}', expr) expr = self._dump(int(expr * 1000), local_vars) + right else: raise self.Exception(f'Unsupported object {obj}', expr) if expr.startswith('void '): left = self.interpret_expression(expr[5:], local_vars, allow_recursion) return None, should_return if expr.startswith('{'): inner, outer = self._separate_at_paren(expr, '}') # Look for Map first 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 else: expr = self._dump(inner, local_vars) + outer if expr.startswith('('): 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 else: expr = self._dump(inner, local_vars) + outer if expr.startswith('['): 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(rf'''(?x) (?Ptry|finally)\s*| (?Pcatch\s*(?P\(\s*{_NAME_RE}\s*\)))| (?Pswitch)\s*\(| (?Pfor)\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:], '' try: ret, should_abort = self.interpret_statement(try_expr, local_vars, allow_recursion) if should_abort: return ret, True except JS_Throw as e: local_vars[self._EXC_NAME] = e.error except Exception as e: # XXX: This works for now, but makes debugging future issues very hard local_vars[self._EXC_NAME] = e ret, should_abort = self.interpret_statement(expr, local_vars, allow_recursion) return ret, should_abort or should_return elif m and m.group('catch'): catch_expr, expr = self._separate_at_paren(expr[m.end():], '}') if self._EXC_NAME in local_vars: catch_vars = local_vars.new_child({m.group('err'): local_vars.pop(self._EXC_NAME)}) ret, should_abort = self.interpret_statement(catch_expr, catch_vars, allow_recursion) 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('for'): constructor, remaining = self._separate_at_paren(expr[m.end() - 1:], ')') if remaining.startswith('{'): 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:], ')') body, expr = self._separate_at_paren(remaining, '}') body = 'switch(%s){%s}' % (switch_val, body) else: body, expr = remaining, '' start, cndn, increment = self._separate(constructor, ';') self.interpret_expression(start, local_vars, allow_recursion) while True: 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) if should_abort: return ret, True except JS_Break: break 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:], ')') 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:] for default in (False, True): matched = False for item in items: case, stmt = (i.strip() for i in self._separate(item, ':', 1)) if default: matched = matched or case == 'default' elif not matched: matched = (case != 'default' and switch_val == self.interpret_expression(case, local_vars, allow_recursion)) if not matched: continue try: ret, should_abort = self.interpret_statement(stmt, local_vars, allow_recursion) if should_abort: return ret except JS_Break: break if matched: break 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)) 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\+\+|--)(?P{_NAME_RE})| (?P{_NAME_RE})(?P\+\+|--)''', expr): var = m.group('var1') or m.group('var2') start, end = m.span() sign = m.group('pre_sign') or m.group('post_sign') ret = local_vars[var] local_vars[var] += 1 if sign[0] == '+' else -1 if m.group('pre_sign'): ret = local_vars[var] expr = expr[:start] + self._dump(ret, local_vars) + expr[end:] if not expr: return None, should_return m = re.match(fr'''(?x) (?P (?P{_NAME_RE})(?:\[(?P[^\]]+?)\])?\s* (?P{"|".join(map(re.escape, set(_OPERATORS) - _COMP_OPERATORS))})? =(?!=)(?P.*)$ )|(?P (?!if|return|true|false|null|undefined)(?P{_NAME_RE})$ )|(?P (?P{_NAME_RE})\[(?P.+)\]$ )|(?P (?P{_NAME_RE})(?:(?P\?)?\.(?P[^(]+)|\[(?P[^\]]+)\])\s* )|(?P (?P{_NAME_RE})\((?P.*)\)$ )''', expr) if m and m.group('assign'): left_val = local_vars.get(m.group('out')) if not m.group('index'): 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 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) if not isinstance(idx, (int, float)): raise self.Exception(f'List index {idx} must be integer', expr) idx = int(idx) left_val[idx] = self._operator( 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(): return int(expr), should_return elif expr == 'break': raise JS_Break() elif expr == 'continue': raise JS_Continue() elif expr == 'undefined': return JS_Undefined, should_return elif m and m.group('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 if m and m.group('indexing'): val = local_vars[m.group('in')] idx = self.interpret_expression(m.group('idx'), local_vars, allow_recursion) return self._index(val, idx), should_return for op in _OPERATORS: separated = list(self._separate(expr, op)) right_expr = 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, left_val, right_expr, expr, local_vars, allow_recursion), should_return if m and m.group('attribute'): 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, ')') else: arg_str, remaining = None, arg_str def assertion(cndn, msg): """ assert, but without risk of getting optimized out """ if not cndn: raise self.Exception(f'{member} {msg}', expr) def eval_method(): if (variable, member) == ('console', 'debug'): if Debugger.ENABLED: Debugger.write(self.interpret_expression(f'[{arg_str}]', local_vars, allow_recursion)) return types = { 'String': str, 'Math': float, } obj = local_vars.get(variable, types.get(variable, NO_DEFAULT)) if obj is NO_DEFAULT: if variable not in self._objects: 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, nullish) # Function call argvals = [ self.interpret_expression(v, local_vars, allow_recursion) for v in self._separate(arg_str)] if obj == str: if member == 'fromCharCode': assertion(argvals, 'takes one or more arguments') return ''.join(map(chr, argvals)) raise self.Exception(f'Unsupported String method {member}', expr) elif obj == float: if member == 'pow': assertion(len(argvals) == 2, 'takes two arguments') return argvals[0] ** argvals[1] raise self.Exception(f'Unsupported Math method {member}', expr) if member == 'split': assertion(argvals, 'takes one or more arguments') assertion(len(argvals) == 1, 'with limit argument is not implemented') return obj.split(argvals[0]) if argvals[0] else list(obj) elif member == 'join': assertion(isinstance(obj, list), 'must be applied on a list') assertion(len(argvals) == 1, 'takes exactly one argument') return argvals[0].join(obj) elif member == 'reverse': assertion(not argvals, 'does not take any arguments') obj.reverse() return obj elif member == 'slice': assertion(isinstance(obj, list), 'must be applied on a list') assertion(len(argvals) == 1, 'takes exactly one argument') return obj[argvals[0]:] elif member == 'splice': assertion(isinstance(obj, list), 'must be applied on a list') assertion(argvals, 'takes one or more arguments') index, howMany = map(int, (argvals + [len(obj)])[:2]) if index < 0: index += len(obj) add_items = argvals[2:] res = [] for i in range(index, min(index + howMany, len(obj))): res.append(obj.pop(index)) for i, item in enumerate(add_items): obj.insert(index + i, item) return res elif member == 'unshift': assertion(isinstance(obj, list), 'must be applied on a list') assertion(argvals, 'takes one or more arguments') for item in reversed(argvals): obj.insert(0, item) return obj elif member == 'pop': assertion(isinstance(obj, list), 'must be applied on a list') assertion(not argvals, 'does not take any arguments') if not obj: return return obj.pop() elif member == 'push': assertion(argvals, 'takes one or more arguments') obj.extend(argvals) return obj elif member == 'forEach': assertion(argvals, 'takes one or more arguments') assertion(len(argvals) <= 2, 'takes at-most 2 arguments') f, this = (argvals + [''])[:2] return [f((item, idx, obj), {'this': this}, allow_recursion) for idx, item in enumerate(obj)] elif member == 'indexOf': assertion(argvals, 'takes one or more arguments') assertion(len(argvals) <= 2, 'takes at-most 2 arguments') idx, start = (argvals + [0])[:2] try: return obj.index(idx, start) except ValueError: return -1 idx = int(member) if isinstance(obj, list) else member return obj[idx](argvals, allow_recursion=allow_recursion) if remaining: ret, should_abort = self.interpret_statement( self._named_object(local_vars, eval_method()) + remaining, local_vars, allow_recursion) return ret, should_return or should_abort else: return eval_method(), should_return elif m and m.group('function'): fname = m.group('fname') argvals = [self.interpret_expression(v, local_vars, allow_recursion) for v in self._separate(m.group('args'))] if fname in local_vars: return local_vars[fname](argvals, allow_recursion=allow_recursion), should_return elif fname not in self._functions: self._functions[fname] = self.extract_function(fname) return self._functions[fname](argvals, allow_recursion=allow_recursion), should_return raise self.Exception( f'Unsupported JS expression {truncate_string(expr, 20, 20) if expr != stmt else ""}', stmt) def interpret_expression(self, expr, local_vars, allow_recursion): ret, should_return = self.interpret_statement(expr, local_vars, allow_recursion) if should_return: raise self.Exception('Cannot return from an expression', expr) return ret def extract_object(self, objname): _FUNC_NAME_RE = r'''(?:[a-zA-Z$0-9]+|"[a-zA-Z$0-9]+"|'[a-zA-Z$0-9]+')''' obj = {} obj_m = re.search( r'''(?x) (?(%s\s*:\s*function\s*\(.*?\)\s*{.*?}(?:,\s*)?)*) }\s*; ''' % (re.escape(objname), _FUNC_NAME_RE), self.code) if not obj_m: raise self.Exception(f'Could not find object {objname}') fields = obj_m.group('fields') # Currently, it only supports function definitions fields_m = re.finditer( r'''(?x) (?P%s)\s*:\s*function\s*\((?P(?:%s|,)*)\){(?P[^}]+)} ''' % (_FUNC_NAME_RE, _NAME_RE), fields) for f in fields_m: argnames = f.group('args').split(',') obj[remove_quotes(f.group('key'))] = self.build_function(argnames, f.group('code')) return obj def extract_function_code(self, funcname): """ @returns argnames, code """ func_m = re.search( r'''(?xs) (?: function\s+%(name)s| [{;,]\s*%(name)s\s*=\s*function| (?:var|const|let)\s+%(name)s\s*=\s*function )\s* \((?P[^)]*)\)\s* (?P{.+})''' % {'name': re.escape(funcname)}, self.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 def extract_function(self, funcname): return self.extract_function_from_code(*self.extract_function_code(funcname)) def extract_function_from_code(self, argnames, code, *global_stack): local_vars = {} while True: mobj = re.search(r'function\((?P[^)]*)\)\s*{', code) if mobj is None: break start, body_start = mobj.span() 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)) code = code[:start] + name + remaining return self.build_function(argnames, code, local_vars, *global_stack) def call_function(self, funcname, *args): return self.extract_function(funcname)(args) def build_function(self, argnames, code, *global_stack): global_stack = list(global_stack) or [{}] argnames = tuple(argnames) def resf(args, kwargs={}, allow_recursion=100): 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) if should_abort: return ret return resf