Improve feedback/information with "broken repr"

If the __repr__ of a local object fails, it gets reported as "<broken repr>":

Lines 416 to 451 in d7cf16c

    
           if PY2: 
        
               def safe_repr(value): 
        
                   # type: (Any) -> str 
        
                   try: 
        
                       rv = repr(value).decode("utf-8", "replace") 
        
                       # At this point `rv` contains a bunch of literal escape codes, like 
        
                       # this (exaggerated example): 
        
                       # 
        
                       # u"\\x2f" 
        
                       # 
        
                       # But we want to show this string as: 
        
                       # 
        
                       # u"/" 
        
                       try: 
        
                           # unicode-escape does this job, but can only decode latin1. So we 
        
                           # attempt to encode in latin1. 
        
                           return rv.encode("latin1").decode("unicode-escape") 
        
                       except Exception: 
        
                           # Since usually strings aren't latin1 this can break. In those 
        
                           # cases we just give up. 
        
                           return rv 
        
                   except Exception: 
        
                       # If e.g. the call to `repr` already fails 
        
                       return u"<broken repr>" 
        
           else: 
        
               def safe_repr(value): 
        
                   # type: (Any) -> str 
        
                   try: 
        
                       return repr(value) 
        
                   except Exception: 
        
                       return "<broken repr>"

It would be more helpful if it would include more information there, e.g. at least the class name of the exception.
It could also be something more sophisticated like:

def _try_repr_or_str(obj):
    try:
        return repr(obj)
    except (KeyboardInterrupt, SystemExit):
        raise
    except BaseException:
        return '{}("{}")'.format(type(obj).__name__, obj)


def _format_repr_exception(exc: BaseException, obj: Any) -> str:
    try:
        exc_info = _try_repr_or_str(exc)
    except (KeyboardInterrupt, SystemExit):
        raise
    except BaseException as exc:
        exc_info = "unpresentable exception ({})".format(_try_repr_or_str(exc))
    return "<[{} raised in repr()] {} object at 0x{:x}>".format(
        exc_info, type(obj).__name__, id(obj)
    )

(via https://github.com/blueyed/pytest/blob/8eb380425395332ffdf2cd42b8d7cf8ffab2ea2e/src/_pytest/_io/saferepr.py#L6-L24)

This would help with investigating why a repr is broken actually, e.g. with celery/py-amqp#361.

(slightly related: there was some related (still open) PR for raven-python about this already: getsentry/raven-python#1294)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	if PY2:

	def safe_repr(value):
	# type: (Any) -> str
	try:
	rv = repr(value).decode("utf-8", "replace")

	# At this point `rv` contains a bunch of literal escape codes, like
	# this (exaggerated example):
	#
	# u"\\x2f"
	#
	# But we want to show this string as:
	#
	# u"/"
	try:
	# unicode-escape does this job, but can only decode latin1. So we
	# attempt to encode in latin1.
	return rv.encode("latin1").decode("unicode-escape")
	except Exception:
	# Since usually strings aren't latin1 this can break. In those
	# cases we just give up.
	return rv
	except Exception:
	# If e.g. the call to `repr` already fails
	return u"<broken repr>"


	else:

	def safe_repr(value):
	# type: (Any) -> str
	try:
	return repr(value)
	except Exception:
	return "<broken repr>"

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions