"""Disassembler of Python byte code into mnemonics.""" import sys import types import collections import io from opcode import * from opcode import ( __all__ as _opcodes_all, _cache_format, _inline_cache_entries, _nb_ops, _intrinsic_1_descs, _intrinsic_2_descs, _specializations, _specialized_opmap, ) __all__ = ["code_info", "dis", "disassemble", "distb", "disco", "findlinestarts", "findlabels", "show_code", "get_instructions", "Instruction", "Bytecode"] + _opcodes_all del _opcodes_all _have_code = (types.MethodType, types.FunctionType, types.CodeType, classmethod, staticmethod, type) CONVERT_VALUE = opmap['CONVERT_VALUE'] SET_FUNCTION_ATTRIBUTE = opmap['SET_FUNCTION_ATTRIBUTE'] FUNCTION_ATTR_FLAGS = ('defaults', 'kwdefaults', 'annotations', 'closure') ENTER_EXECUTOR = opmap['ENTER_EXECUTOR'] LOAD_CONST = opmap['LOAD_CONST'] RETURN_CONST = opmap['RETURN_CONST'] LOAD_GLOBAL = opmap['LOAD_GLOBAL'] BINARY_OP = opmap['BINARY_OP'] JUMP_BACKWARD = opmap['JUMP_BACKWARD'] FOR_ITER = opmap['FOR_ITER'] SEND = opmap['SEND'] LOAD_ATTR = opmap['LOAD_ATTR'] LOAD_SUPER_ATTR = opmap['LOAD_SUPER_ATTR'] CALL_INTRINSIC_1 = opmap['CALL_INTRINSIC_1'] CALL_INTRINSIC_2 = opmap['CALL_INTRINSIC_2'] LOAD_FAST_LOAD_FAST = opmap['LOAD_FAST_LOAD_FAST'] STORE_FAST_LOAD_FAST = opmap['STORE_FAST_LOAD_FAST'] STORE_FAST_STORE_FAST = opmap['STORE_FAST_STORE_FAST'] CACHE = opmap["CACHE"] _all_opname = list(opname) _all_opmap = dict(opmap) for name, op in _specialized_opmap.items(): # fill opname and opmap assert op < len(_all_opname) _all_opname[op] = name _all_opmap[name] = op deoptmap = { specialized: base for base, family in _specializations.items() for specialized in family } def _try_compile(source, name): """Attempts to compile the given source, first as an expression and then as a statement if the first approach fails. Utility function to accept strings in functions that otherwise expect code objects """ try: return compile(source, name, 'eval') except SyntaxError: pass return compile(source, name, 'exec') def dis(x=None, *, file=None, depth=None, show_caches=False, adaptive=False, show_offsets=False): """Disassemble classes, methods, functions, and other compiled objects. With no argument, disassemble the last traceback. Compiled objects currently include generator objects, async generator objects, and coroutine objects, all of which store their code object in a special attribute. """ if x is None: distb(file=file, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets) return # Extract functions from methods. if hasattr(x, '__func__'): x = x.__func__ # Extract compiled code objects from... if hasattr(x, '__code__'): # ...a function, or x = x.__code__ elif hasattr(x, 'gi_code'): #...a generator object, or x = x.gi_code elif hasattr(x, 'ag_code'): #...an asynchronous generator object, or x = x.ag_code elif hasattr(x, 'cr_code'): #...a coroutine. x = x.cr_code # Perform the disassembly. if hasattr(x, '__dict__'): # Class or module items = sorted(x.__dict__.items()) for name, x1 in items: if isinstance(x1, _have_code): print("Disassembly of %s:" % name, file=file) try: dis(x1, file=file, depth=depth, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets) except TypeError as msg: print("Sorry:", msg, file=file) print(file=file) elif hasattr(x, 'co_code'): # Code object _disassemble_recursive(x, file=file, depth=depth, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets) elif isinstance(x, (bytes, bytearray)): # Raw bytecode _disassemble_bytes(x, file=file, show_caches=show_caches, show_offsets=show_offsets) elif isinstance(x, str): # Source code _disassemble_str(x, file=file, depth=depth, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets) else: raise TypeError("don't know how to disassemble %s objects" % type(x).__name__) def distb(tb=None, *, file=None, show_caches=False, adaptive=False, show_offsets=False): """Disassemble a traceback (default: last traceback).""" if tb is None: try: if hasattr(sys, 'last_exc'): tb = sys.last_exc.__traceback__ else: tb = sys.last_traceback except AttributeError: raise RuntimeError("no last traceback to disassemble") from None while tb.tb_next: tb = tb.tb_next disassemble(tb.tb_frame.f_code, tb.tb_lasti, file=file, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets) # The inspect module interrogates this dictionary to build its # list of CO_* constants. It is also used by pretty_flags to # turn the co_flags field into a human readable list. COMPILER_FLAG_NAMES = { 1: "OPTIMIZED", 2: "NEWLOCALS", 4: "VARARGS", 8: "VARKEYWORDS", 16: "NESTED", 32: "GENERATOR", 64: "NOFREE", 128: "COROUTINE", 256: "ITERABLE_COROUTINE", 512: "ASYNC_GENERATOR", } def pretty_flags(flags): """Return pretty representation of code flags.""" names = [] for i in range(32): flag = 1<" # Sentinel to represent values that cannot be calculated UNKNOWN = _Unknown() def _get_code_object(x): """Helper to handle methods, compiled or raw code objects, and strings.""" # Extract functions from methods. if hasattr(x, '__func__'): x = x.__func__ # Extract compiled code objects from... if hasattr(x, '__code__'): # ...a function, or x = x.__code__ elif hasattr(x, 'gi_code'): #...a generator object, or x = x.gi_code elif hasattr(x, 'ag_code'): #...an asynchronous generator object, or x = x.ag_code elif hasattr(x, 'cr_code'): #...a coroutine. x = x.cr_code # Handle source code. if isinstance(x, str): x = _try_compile(x, "") # By now, if we don't have a code object, we can't disassemble x. if hasattr(x, 'co_code'): return x raise TypeError("don't know how to disassemble %s objects" % type(x).__name__) def _deoptop(op): name = _all_opname[op] return _all_opmap[deoptmap[name]] if name in deoptmap else op def _get_code_array(co, adaptive): return co._co_code_adaptive if adaptive else co.co_code def code_info(x): """Formatted details of methods, functions, or code.""" return _format_code_info(_get_code_object(x)) def _format_code_info(co): lines = [] lines.append("Name: %s" % co.co_name) lines.append("Filename: %s" % co.co_filename) lines.append("Argument count: %s" % co.co_argcount) lines.append("Positional-only arguments: %s" % co.co_posonlyargcount) lines.append("Kw-only arguments: %s" % co.co_kwonlyargcount) lines.append("Number of locals: %s" % co.co_nlocals) lines.append("Stack size: %s" % co.co_stacksize) lines.append("Flags: %s" % pretty_flags(co.co_flags)) if co.co_consts: lines.append("Constants:") for i_c in enumerate(co.co_consts): lines.append("%4d: %r" % i_c) if co.co_names: lines.append("Names:") for i_n in enumerate(co.co_names): lines.append("%4d: %s" % i_n) if co.co_varnames: lines.append("Variable names:") for i_n in enumerate(co.co_varnames): lines.append("%4d: %s" % i_n) if co.co_freevars: lines.append("Free variables:") for i_n in enumerate(co.co_freevars): lines.append("%4d: %s" % i_n) if co.co_cellvars: lines.append("Cell variables:") for i_n in enumerate(co.co_cellvars): lines.append("%4d: %s" % i_n) return "\n".join(lines) def show_code(co, *, file=None): """Print details of methods, functions, or code to *file*. If *file* is not provided, the output is printed on stdout. """ print(code_info(co), file=file) Positions = collections.namedtuple( 'Positions', [ 'lineno', 'end_lineno', 'col_offset', 'end_col_offset', ], defaults=[None] * 4 ) _Instruction = collections.namedtuple( "_Instruction", [ 'opname', 'opcode', 'arg', 'argval', 'argrepr', 'offset', 'start_offset', 'starts_line', 'line_number', 'label', 'positions', 'cache_info', ], defaults=[None, None, None] ) _Instruction.opname.__doc__ = "Human readable name for operation" _Instruction.opcode.__doc__ = "Numeric code for operation" _Instruction.arg.__doc__ = "Numeric argument to operation (if any), otherwise None" _Instruction.argval.__doc__ = "Resolved arg value (if known), otherwise same as arg" _Instruction.argrepr.__doc__ = "Human readable description of operation argument" _Instruction.offset.__doc__ = "Start index of operation within bytecode sequence" _Instruction.start_offset.__doc__ = ( "Start index of operation within bytecode sequence, including extended args if present; " "otherwise equal to Instruction.offset" ) _Instruction.starts_line.__doc__ = "True if this opcode starts a source line, otherwise False" _Instruction.line_number.__doc__ = "source line number associated with this opcode (if any), otherwise None" _Instruction.label.__doc__ = "A label (int > 0) if this instruction is a jump target, otherwise None" _Instruction.positions.__doc__ = "dis.Positions object holding the span of source code covered by this instruction" _Instruction.cache_info.__doc__ = "list of (name, size, data), one for each cache entry of the instruction" _ExceptionTableEntryBase = collections.namedtuple("_ExceptionTableEntryBase", "start end target depth lasti") class _ExceptionTableEntry(_ExceptionTableEntryBase): pass _OPNAME_WIDTH = 20 _OPARG_WIDTH = 5 def _get_cache_size(opname): return _inline_cache_entries.get(opname, 0) def _get_jump_target(op, arg, offset): """Gets the bytecode offset of the jump target if this is a jump instruction. Otherwise return None. """ deop = _deoptop(op) caches = _get_cache_size(_all_opname[deop]) if deop in hasjrel: if _is_backward_jump(deop): arg = -arg target = offset + 2 + arg*2 target += 2 * caches elif deop in hasjabs: target = arg*2 else: target = None return target class Instruction(_Instruction): """Details for a bytecode operation. Defined fields: opname - human readable name for operation opcode - numeric code for operation arg - numeric argument to operation (if any), otherwise None argval - resolved arg value (if known), otherwise same as arg argrepr - human readable description of operation argument offset - start index of operation within bytecode sequence start_offset - start index of operation within bytecode sequence including extended args if present; otherwise equal to Instruction.offset starts_line - True if this opcode starts a source line, otherwise False line_number - source line number associated with this opcode (if any), otherwise None label - A label if this instruction is a jump target, otherwise None positions - Optional dis.Positions object holding the span of source code covered by this instruction cache_info - information about the format and content of the instruction's cache entries (if any) """ @property def oparg(self): """Alias for Instruction.arg.""" return self.arg @property def baseopcode(self): """Numeric code for the base operation if operation is specialized. Otherwise equal to Instruction.opcode. """ return _deoptop(self.opcode) @property def baseopname(self): """Human readable name for the base operation if operation is specialized. Otherwise equal to Instruction.opname. """ return opname[self.baseopcode] @property def cache_offset(self): """Start index of the cache entries following the operation.""" return self.offset + 2 @property def end_offset(self): """End index of the cache entries following the operation.""" return self.cache_offset + _get_cache_size(_all_opname[self.opcode])*2 @property def jump_target(self): """Bytecode index of the jump target if this is a jump operation. Otherwise return None. """ return _get_jump_target(self.opcode, self.arg, self.offset) @property def is_jump_target(self): """True if other code jumps to here, otherwise False""" return self.label is not None def __str__(self): output = io.StringIO() formatter = Formatter(file=output) formatter.print_instruction(self, False) return output.getvalue() class Formatter: def __init__(self, file=None, lineno_width=0, offset_width=0, label_width=0, line_offset=0): """Create a Formatter *file* where to write the output *lineno_width* sets the width of the line number field (0 omits it) *offset_width* sets the width of the instruction offset field *label_width* sets the width of the label field *line_offset* the line number (within the code unit) """ self.file = file self.lineno_width = lineno_width self.offset_width = offset_width self.label_width = label_width def print_instruction(self, instr, mark_as_current=False): """Format instruction details for inclusion in disassembly output.""" lineno_width = self.lineno_width offset_width = self.offset_width label_width = self.label_width new_source_line = (lineno_width > 0 and instr.starts_line and instr.offset > 0) if new_source_line: print(file=self.file) fields = [] # Column: Source code line number if lineno_width: if instr.starts_line: lineno_fmt = "%%%dd" if instr.line_number is not None else "%%%ds" lineno_fmt = lineno_fmt % lineno_width lineno = _NO_LINENO if instr.line_number is None else instr.line_number fields.append(lineno_fmt % lineno) else: fields.append(' ' * lineno_width) # Column: Label if instr.label is not None: lbl = f"L{instr.label}:" fields.append(f"{lbl:>{label_width}}") else: fields.append(' ' * label_width) # Column: Instruction offset from start of code sequence if offset_width > 0: fields.append(f"{repr(instr.offset):>{offset_width}} ") # Column: Current instruction indicator if mark_as_current: fields.append('-->') else: fields.append(' ') # Column: Opcode name fields.append(instr.opname.ljust(_OPNAME_WIDTH)) # Column: Opcode argument if instr.arg is not None: arg = repr(instr.arg) # If opname is longer than _OPNAME_WIDTH, we allow it to overflow into # the space reserved for oparg. This results in fewer misaligned opargs # in the disassembly output. opname_excess = max(0, len(instr.opname) - _OPNAME_WIDTH) fields.append(repr(instr.arg).rjust(_OPARG_WIDTH - opname_excess)) # Column: Opcode argument details if instr.argrepr: fields.append('(' + instr.argrepr + ')') print(' '.join(fields).rstrip(), file=self.file) def print_exception_table(self, exception_entries): file = self.file if exception_entries: print("ExceptionTable:", file=file) for entry in exception_entries: lasti = " lasti" if entry.lasti else "" start = entry.start_label end = entry.end_label target = entry.target_label print(f" L{start} to L{end} -> L{target} [{entry.depth}]{lasti}", file=file) class ArgResolver: def __init__(self, co_consts, names, varname_from_oparg, labels_map): self.co_consts = co_consts self.names = names self.varname_from_oparg = varname_from_oparg self.labels_map = labels_map def get_argval_argrepr(self, op, arg, offset): get_name = None if self.names is None else self.names.__getitem__ argval = None argrepr = '' deop = _deoptop(op) if arg is not None: # Set argval to the dereferenced value of the argument when # available, and argrepr to the string representation of argval. # _disassemble_bytes needs the string repr of the # raw name index for LOAD_GLOBAL, LOAD_CONST, etc. argval = arg if deop in hasconst: argval, argrepr = _get_const_info(deop, arg, self.co_consts) elif deop in hasname: if deop == LOAD_GLOBAL: argval, argrepr = _get_name_info(arg//2, get_name) if (arg & 1) and argrepr: argrepr = f"{argrepr} + NULL" elif deop == LOAD_ATTR: argval, argrepr = _get_name_info(arg//2, get_name) if (arg & 1) and argrepr: argrepr = f"{argrepr} + NULL|self" elif deop == LOAD_SUPER_ATTR: argval, argrepr = _get_name_info(arg//4, get_name) if (arg & 1) and argrepr: argrepr = f"{argrepr} + NULL|self" else: argval, argrepr = _get_name_info(arg, get_name) elif deop in hasjabs: argval = arg*2 argrepr = f"to L{self.labels_map[argval]}" elif deop in hasjrel: signed_arg = -arg if _is_backward_jump(deop) else arg argval = offset + 2 + signed_arg*2 caches = _get_cache_size(_all_opname[deop]) argval += 2 * caches if deop == ENTER_EXECUTOR: argval += 2 argrepr = f"to L{self.labels_map[argval]}" elif deop in (LOAD_FAST_LOAD_FAST, STORE_FAST_LOAD_FAST, STORE_FAST_STORE_FAST): arg1 = arg >> 4 arg2 = arg & 15 val1, argrepr1 = _get_name_info(arg1, self.varname_from_oparg) val2, argrepr2 = _get_name_info(arg2, self.varname_from_oparg) argrepr = argrepr1 + ", " + argrepr2 argval = val1, val2 elif deop in haslocal or deop in hasfree: argval, argrepr = _get_name_info(arg, self.varname_from_oparg) elif deop in hascompare: argval = cmp_op[arg >> 5] argrepr = argval if arg & 16: argrepr = f"bool({argrepr})" elif deop == CONVERT_VALUE: argval = (None, str, repr, ascii)[arg] argrepr = ('', 'str', 'repr', 'ascii')[arg] elif deop == SET_FUNCTION_ATTRIBUTE: argrepr = ', '.join(s for i, s in enumerate(FUNCTION_ATTR_FLAGS) if arg & (1<> 1 lasti = bool(dl&1) entries.append(_ExceptionTableEntry(start, end, target, depth, lasti)) except StopIteration: return entries def _is_backward_jump(op): return opname[op] in ('JUMP_BACKWARD', 'JUMP_BACKWARD_NO_INTERRUPT', 'ENTER_EXECUTOR') def _get_instructions_bytes(code, linestarts=None, line_offset=0, co_positions=None, original_code=None, labels_map=None, arg_resolver=None): """Iterate over the instructions in a bytecode string. Generates a sequence of Instruction namedtuples giving the details of each opcode. """ # Use the basic, unadaptive code for finding labels and actually walking the # bytecode, since replacements like ENTER_EXECUTOR and INSTRUMENTED_* can # mess that logic up pretty badly: original_code = original_code or code co_positions = co_positions or iter(()) labels_map = labels_map or _make_labels_map(original_code) starts_line = False local_line_number = None line_number = None for offset, start_offset, op, arg in _unpack_opargs(original_code): if linestarts is not None: starts_line = offset in linestarts if starts_line: local_line_number = linestarts[offset] if local_line_number is not None: line_number = local_line_number + line_offset else: line_number = None positions = Positions(*next(co_positions, ())) deop = _deoptop(op) op = code[offset] if arg_resolver: argval, argrepr = arg_resolver.get_argval_argrepr(op, arg, offset) else: argval, argrepr = arg, repr(arg) instr = Instruction(_all_opname[op], op, arg, argval, argrepr, offset, start_offset, starts_line, line_number, labels_map.get(offset, None), positions) caches = _get_cache_size(_all_opname[deop]) # Advance the co_positions iterator: for _ in range(caches): next(co_positions, ()) if caches: cache_info = [] for name, size in _cache_format[opname[deop]].items(): data = code[offset + 2: offset + 2 + 2 * size] cache_info.append((name, size, data)) else: cache_info = None yield Instruction(_all_opname[op], op, arg, argval, argrepr, offset, start_offset, starts_line, line_number, labels_map.get(offset, None), positions, cache_info) def disassemble(co, lasti=-1, *, file=None, show_caches=False, adaptive=False, show_offsets=False): """Disassemble a code object.""" linestarts = dict(findlinestarts(co)) exception_entries = _parse_exception_table(co) _disassemble_bytes(_get_code_array(co, adaptive), lasti, co._varname_from_oparg, co.co_names, co.co_consts, linestarts, file=file, exception_entries=exception_entries, co_positions=co.co_positions(), show_caches=show_caches, original_code=co.co_code, show_offsets=show_offsets) def _disassemble_recursive(co, *, file=None, depth=None, show_caches=False, adaptive=False, show_offsets=False): disassemble(co, file=file, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets) if depth is None or depth > 0: if depth is not None: depth = depth - 1 for x in co.co_consts: if hasattr(x, 'co_code'): print(file=file) print("Disassembly of %r:" % (x,), file=file) _disassemble_recursive( x, file=file, depth=depth, show_caches=show_caches, adaptive=adaptive, show_offsets=show_offsets ) def _make_labels_map(original_code, exception_entries=()): jump_targets = set(findlabels(original_code)) labels = set(jump_targets) for start, end, target, _, _ in exception_entries: labels.add(start) labels.add(end) labels.add(target) labels = sorted(labels) labels_map = {offset: i+1 for (i, offset) in enumerate(sorted(labels))} for e in exception_entries: e.start_label = labels_map[e.start] e.end_label = labels_map[e.end] e.target_label = labels_map[e.target] return labels_map _NO_LINENO = ' --' def _get_lineno_width(linestarts): if linestarts is None: return 0 maxlineno = max(filter(None, linestarts.values()), default=-1) if maxlineno == -1: # Omit the line number column entirely if we have no line number info return 0 lineno_width = max(3, len(str(maxlineno))) if lineno_width < len(_NO_LINENO) and None in linestarts.values(): lineno_width = len(_NO_LINENO) return lineno_width def _disassemble_bytes(code, lasti=-1, varname_from_oparg=None, names=None, co_consts=None, linestarts=None, *, file=None, line_offset=0, exception_entries=(), co_positions=None, show_caches=False, original_code=None, show_offsets=False): offset_width = len(str(max(len(code) - 2, 9999))) if show_offsets else 0 labels_map = _make_labels_map(original_code or code, exception_entries) label_width = 4 + len(str(len(labels_map))) formatter = Formatter(file=file, lineno_width=_get_lineno_width(linestarts), offset_width=offset_width, label_width=label_width, line_offset=line_offset) arg_resolver = ArgResolver(co_consts, names, varname_from_oparg, labels_map) instrs = _get_instructions_bytes(code, linestarts=linestarts, line_offset=line_offset, co_positions=co_positions, original_code=original_code, labels_map=labels_map, arg_resolver=arg_resolver) print_instructions(instrs, exception_entries, formatter, show_caches=show_caches, lasti=lasti) def print_instructions(instrs, exception_entries, formatter, show_caches=False, lasti=-1): for instr in instrs: if show_caches: is_current_instr = instr.offset == lasti else: # Each CACHE takes 2 bytes is_current_instr = instr.offset <= lasti \ <= instr.offset + 2 * _get_cache_size(_all_opname[_deoptop(instr.opcode)]) formatter.print_instruction(instr, is_current_instr) deop = _deoptop(instr.opcode) if show_caches and instr.cache_info: offset = instr.offset for name, size, data in instr.cache_info: for i in range(size): offset += 2 # Only show the fancy argrepr for a CACHE instruction when it's # the first entry for a particular cache value: if i == 0: argrepr = f"{name}: {int.from_bytes(data, sys.byteorder)}" else: argrepr = "" formatter.print_instruction( Instruction("CACHE", CACHE, 0, None, argrepr, offset, offset, False, None, None, instr.positions), is_current_instr) formatter.print_exception_table(exception_entries) def _disassemble_str(source, **kwargs): """Compile the source string, then disassemble the code object.""" _disassemble_recursive(_try_compile(source, ''), **kwargs) disco = disassemble # XXX For backwards compatibility # Rely on C `int` being 32 bits for oparg _INT_BITS = 32 # Value for c int when it overflows _INT_OVERFLOW = 2 ** (_INT_BITS - 1) def _unpack_opargs(code): extended_arg = 0 extended_args_offset = 0 # Number of EXTENDED_ARG instructions preceding the current instruction caches = 0 for i in range(0, len(code), 2): # Skip inline CACHE entries: if caches: caches -= 1 continue op = code[i] deop = _deoptop(op) caches = _get_cache_size(_all_opname[deop]) if deop in hasarg: arg = code[i+1] | extended_arg extended_arg = (arg << 8) if deop == EXTENDED_ARG else 0 # The oparg is stored as a signed integer # If the value exceeds its upper limit, it will overflow and wrap # to a negative integer if extended_arg >= _INT_OVERFLOW: extended_arg -= 2 * _INT_OVERFLOW else: arg = None extended_arg = 0 if deop == EXTENDED_ARG: extended_args_offset += 1 yield (i, i, op, arg) else: start_offset = i - extended_args_offset*2 yield (i, start_offset, op, arg) extended_args_offset = 0 def findlabels(code): """Detect all offsets in a byte code which are jump targets. Return the list of offsets. """ labels = [] for offset, _, op, arg in _unpack_opargs(code): if arg is not None: label = _get_jump_target(op, arg, offset) if label is None: continue if label not in labels: labels.append(label) return labels def findlinestarts(code): """Find the offsets in a byte code which are start of lines in the source. Generate pairs (offset, lineno) lineno will be an integer or None the offset does not have a source line. """ lastline = False # None is a valid line number for start, end, line in code.co_lines(): if line is not lastline: lastline = line yield start, line return def _find_imports(co): """Find import statements in the code Generate triplets (name, level, fromlist) where name is the imported module and level, fromlist are the corresponding args to __import__. """ IMPORT_NAME = opmap['IMPORT_NAME'] consts = co.co_consts names = co.co_names opargs = [(op, arg) for _, _, op, arg in _unpack_opargs(co.co_code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op == IMPORT_NAME and i >= 2: from_op = opargs[i-1] level_op = opargs[i-2] if (from_op[0] in hasconst and level_op[0] in hasconst): level = _get_const_value(level_op[0], level_op[1], consts) fromlist = _get_const_value(from_op[0], from_op[1], consts) yield (names[oparg], level, fromlist) def _find_store_names(co): """Find names of variables which are written in the code Generate sequence of strings """ STORE_OPS = { opmap['STORE_NAME'], opmap['STORE_GLOBAL'] } names = co.co_names for _, _, op, arg in _unpack_opargs(co.co_code): if op in STORE_OPS: yield names[arg] class Bytecode: """The bytecode operations of a piece of code Instantiate this with a function, method, other compiled object, string of code, or a code object (as returned by compile()). Iterating over this yields the bytecode operations as Instruction instances. """ def __init__(self, x, *, first_line=None, current_offset=None, show_caches=False, adaptive=False, show_offsets=False): self.codeobj = co = _get_code_object(x) if first_line is None: self.first_line = co.co_firstlineno self._line_offset = 0 else: self.first_line = first_line self._line_offset = first_line - co.co_firstlineno self._linestarts = dict(findlinestarts(co)) self._original_object = x self.current_offset = current_offset self.exception_entries = _parse_exception_table(co) self.show_caches = show_caches self.adaptive = adaptive self.show_offsets = show_offsets def __iter__(self): co = self.codeobj original_code = co.co_code labels_map = _make_labels_map(original_code, self.exception_entries) arg_resolver = ArgResolver(co.co_consts, co.co_names, co._varname_from_oparg, labels_map) return _get_instructions_bytes(_get_code_array(co, self.adaptive), linestarts=self._linestarts, line_offset=self._line_offset, co_positions=co.co_positions(), original_code=original_code, labels_map=labels_map, arg_resolver=arg_resolver) def __repr__(self): return "{}({!r})".format(self.__class__.__name__, self._original_object) @classmethod def from_traceback(cls, tb, *, show_caches=False, adaptive=False): """ Construct a Bytecode from the given traceback """ while tb.tb_next: tb = tb.tb_next return cls( tb.tb_frame.f_code, current_offset=tb.tb_lasti, show_caches=show_caches, adaptive=adaptive ) def info(self): """Return formatted information about the code object.""" return _format_code_info(self.codeobj) def dis(self): """Return a formatted view of the bytecode operations.""" co = self.codeobj if self.current_offset is not None: offset = self.current_offset else: offset = -1 with io.StringIO() as output: _disassemble_bytes(_get_code_array(co, self.adaptive), varname_from_oparg=co._varname_from_oparg, names=co.co_names, co_consts=co.co_consts, linestarts=self._linestarts, line_offset=self._line_offset, file=output, lasti=offset, exception_entries=self.exception_entries, co_positions=co.co_positions(), show_caches=self.show_caches, original_code=co.co_code, show_offsets=self.show_offsets) return output.getvalue() def main(): import argparse parser = argparse.ArgumentParser() parser.add_argument('-C', '--show-caches', action='store_true', help='show inline caches') parser.add_argument('-O', '--show-offsets', action='store_true', help='show instruction offsets') parser.add_argument('infile', type=argparse.FileType('rb'), nargs='?', default='-') args = parser.parse_args() with args.infile as infile: source = infile.read() code = compile(source, args.infile.name, "exec") dis(code, show_caches=args.show_caches, show_offsets=args.show_offsets) if __name__ == "__main__": main()