tblib 3.0.0

Traceback serialization library.

Serialization library for Exceptions and Tracebacks.

• Free software: BSD license

It allows you to:

• Pickle tracebacks and raise exceptions with pickled tracebacks in different processes. This allows better error handling when running code over multiple processes (imagine multiprocessing, billiard, futures, celery etc).

• Create traceback objects from strings (the from_string method). No pickling is used.

• Serialize tracebacks to/from plain dicts (the from_dict and to_dict methods). No pickling is used.

• Raise the tracebacks created from the aforementioned sources.

• Pickle an Exception together with its traceback and exception chain (raise ... from ...) (Python 3 only)

Again, note that using the pickle support is completely optional. You are solely responsible for security problems should you decide to use the pickle support.

Installation

pip install tblib

Documentation

• Pickling tracebacks

• Unpickling tracebacks

• Raising

• Pickling Exceptions together with their traceback and chain (Python 3 only)

  • What if we have a local stack, does it show correctly ?

  • It also supports more contrived scenarios

• Reference

  • tblib.Traceback

    • tblib.Traceback.to_dict

    • tblib.Traceback.from_dict

    • tblib.Traceback.from_string

  • tblib.decorators.return_error

    • What if we have a local call stack ?

    • Other weird stuff

Pickling tracebacks

Note: The traceback objects that come out are stripped of some attributes (like variables). But you’ll be able to raise exceptions with those tracebacks or print them - that should cover 99% of the usecases.

>>> from tblib import pickling_support
>>> pickling_support.install()
>>> import pickle, sys
>>> def inner_0():
...     raise Exception('fail')
...
>>> def inner_1():
...     inner_0()
...
>>> def inner_2():
...     inner_1()
...
>>> try:
...     inner_2()
... except:
...     s1 = pickle.dumps(sys.exc_info())
...
>>> len(s1) > 1
True
>>> try:
...     inner_2()
... except:
...     s2 = pickle.dumps(sys.exc_info(), protocol=pickle.HIGHEST_PROTOCOL)
...
>>> len(s2) > 1
True

>>> try:
...     import cPickle
... except ImportError:
...     import pickle as cPickle
>>> try:
...     inner_2()
... except:
...     s3 = cPickle.dumps(sys.exc_info(), protocol=pickle.HIGHEST_PROTOCOL)
...
>>> len(s3) > 1
True

Unpickling tracebacks

>>> pickle.loads(s1)
(<...Exception'>, Exception('fail'...), <traceback object at ...>)

>>> pickle.loads(s2)
(<...Exception'>, Exception('fail'...), <traceback object at ...>)

>>> pickle.loads(s3)
(<...Exception'>, Exception('fail'...), <traceback object at ...>)

Raising

>>> from six import reraise
>>> reraise(*pickle.loads(s1))
Traceback (most recent call last):
  ...
  File "<doctest README.rst[14]>", line 1, in <module>
    reraise(*pickle.loads(s2))
  File "<doctest README.rst[8]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail
>>> reraise(*pickle.loads(s2))
Traceback (most recent call last):
  ...
  File "<doctest README.rst[14]>", line 1, in <module>
    reraise(*pickle.loads(s2))
  File "<doctest README.rst[8]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail
>>> reraise(*pickle.loads(s3))
Traceback (most recent call last):
  ...
  File "<doctest README.rst[14]>", line 1, in <module>
    reraise(*pickle.loads(s2))
  File "<doctest README.rst[8]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

Pickling Exceptions together with their traceback and chain (Python 3 only)

>>> try:  # doctest: +SKIP
...     try:
...         1 / 0
...     except Exception as e:
...         raise Exception("foo") from e
... except Exception as e:
...     s = pickle.dumps(e)
>>> raise pickle.loads(s)  # doctest: +SKIP
Traceback (most recent call last):
  File "<doctest README.rst[16]>", line 3, in <module>
    1 / 0
ZeroDivisionError: division by zero

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<doctest README.rst[17]>", line 1, in <module>
    raise pickle.loads(s)
  File "<doctest README.rst[16]>", line 5, in <module>
    raise Exception("foo") from e
Exception: foo

BaseException subclasses defined after calling pickling_support.install() will not retain their traceback and exception chain pickling. To cover custom Exceptions, there are three options:

1. Use @pickling_support.install as a decorator for each custom Exception

   >>> from tblib import pickling_support
   >>> # Declare all imports of your package's dependencies
   >>> import numpy  # doctest: +SKIP
   
   >>> pickling_support.install()  # install for all modules imported so far
   
   >>> @pickling_support.install
   ... class CustomError(Exception):
   ...     pass

   Eventual subclasses of CustomError will need to be decorated again.

2. Invoke pickling_support.install() after all modules have been imported and all Exception subclasses have been declared

   >>> # Declare all imports of your package's dependencies
   >>> import numpy  # doctest: +SKIP
   >>> from tblib import pickling_support
   
   >>> # Declare your own custom Exceptions
   >>> class CustomError(Exception):
   ...     pass
   
   >>> # Finally, install tblib
   >>> pickling_support.install()

3. Selectively install tblib for Exception instances just before they are pickled

   pickling_support.install(<Exception instance>, [Exception instance], ...)

   The above will install tblib pickling for all listed exceptions as well as any other exceptions in their exception chains.

   For example, one could write a wrapper to be used with ProcessPoolExecutor, Dask.distributed, or similar libraries:

>>> from tblib import pickling_support
>>> def wrapper(func, *args, **kwargs):
...     try:
...         return func(*args, **kwargs)
...     except Exception as e:
...         pickling_support.install(e)
...         raise

What if we have a local stack, does it show correctly ?

Yes it does:

>>> exc_info = pickle.loads(s3)
>>> def local_0():
...     reraise(*exc_info)
...
>>> def local_1():
...     local_0()
...
>>> def local_2():
...     local_1()
...
>>> local_2()
Traceback (most recent call last):
  File "...doctest.py", line ..., in __run
    compileflags, 1) in test.globs
  File "<doctest README.rst[24]>", line 1, in <module>
    local_2()
  File "<doctest README.rst[23]>", line 2, in local_2
    local_1()
  File "<doctest README.rst[22]>", line 2, in local_1
    local_0()
  File "<doctest README.rst[21]>", line 2, in local_0
    reraise(*exc_info)
  File "<doctest README.rst[11]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

It also supports more contrived scenarios

Like tracebacks with syntax errors:

>>> from tblib import Traceback
>>> from examples import bad_syntax
>>> try:
...     bad_syntax()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[58]>", line 1, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[57]>", line 2, in <module>
    bad_syntax()
  File "...tests...examples.py", line 18, in bad_syntax
    import badsyntax
  File "...tests...badsyntax.py", line 5
    is very bad
     ^
SyntaxError: invalid syntax

Or other import failures:

>>> from examples import bad_module
>>> try:
...     bad_module()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[61]>", line 1, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[60]>", line 2, in <module>
    bad_module()
  File "...tests...examples.py", line 23, in bad_module
    import badmodule
  File "...tests...badmodule.py", line 3, in <module>
    raise Exception("boom!")
Exception: boom!

Or a traceback that’s caused by exceeding the recursion limit (here we’re forcing the type and value to have consistency across platforms):

>>> def f(): f()
>>> try:
...    f()
... except RuntimeError:
...    et, ev, tb = sys.exc_info()
...    tb = Traceback(tb)
...
>>> reraise(RuntimeError, RuntimeError("maximum recursion depth exceeded"), tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[32]>", line 1, in f
    def f(): f()
  File "<doctest README.rst[32]>", line 1, in f
    def f(): f()
  File "<doctest README.rst[32]>", line 1, in f
    def f(): f()
  ...
RuntimeError: maximum recursion depth exceeded

Reference

tblib.Traceback

It is used by the pickling_support. You can use it too if you want more flexibility:

>>> from tblib import Traceback
>>> try:
...     inner_2()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[21]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[21]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

tblib.Traceback.to_dict

You can use the to_dict method and the from_dict classmethod to convert a Traceback into and from a dictionary serializable by the stdlib json.JSONDecoder:

>>> import json
>>> from pprint import pprint
>>> try:
...     inner_2()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...     tb_dict = tb.to_dict()
...     pprint(tb_dict)
{'tb_frame': {'f_code': {'co_filename': '<doctest README.rst[...]>',
                         'co_name': '<module>'},
              'f_globals': {'__name__': '__main__'},
              'f_lineno': 5},
 'tb_lineno': 2,
 'tb_next': {'tb_frame': {'f_code': {'co_filename': ...,
                                     'co_name': 'inner_2'},
                          'f_globals': {'__name__': '__main__'},
                          'f_lineno': 2},
             'tb_lineno': 2,
             'tb_next': {'tb_frame': {'f_code': {'co_filename': ...,
                                                 'co_name': 'inner_1'},
                                      'f_globals': {'__name__': '__main__'},
                                      'f_lineno': 2},
                         'tb_lineno': 2,
                         'tb_next': {'tb_frame': {'f_code': {'co_filename': ...,
                                                             'co_name': 'inner_0'},
                                                  'f_globals': {'__name__': '__main__'},
                                                  'f_lineno': 2},
                                     'tb_lineno': 2,
                                     'tb_next': None}}}}

tblib.Traceback.from_dict

Building on the previous example:

>>> tb_json = json.dumps(tb_dict)
>>> tb = Traceback.from_dict(json.loads(tb_json))
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[21]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[21]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

tblib.Traceback.from_string

>>> tb = Traceback.from_string("""
... File "skipped.py", line 123, in func_123
... Traceback (most recent call last):
...   File "tests/examples.py", line 2, in func_a
...     func_b()
...   File "tests/examples.py", line 6, in func_b
...     func_c()
...   File "tests/examples.py", line 10, in func_c
...     func_d()
...   File "tests/examples.py", line 14, in func_d
... Doesn't: matter
... """)
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[42]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "...examples.py", line 2, in func_a
    func_b()
  File "...examples.py", line 6, in func_b
    func_c()
  File "...examples.py", line 10, in func_c
    func_d()
  File "...examples.py", line 14, in func_d
    raise Exception("Guessing time !")
Exception: fail

If you use the strict=False option then parsing is a bit more lax:

>>> tb = Traceback.from_string("""
... File "bogus.py", line 123, in bogus
... Traceback (most recent call last):
...  File "tests/examples.py", line 2, in func_a
...   func_b()
...    File "tests/examples.py", line 6, in func_b
...     func_c()
...    File "tests/examples.py", line 10, in func_c
...   func_d()
...  File "tests/examples.py", line 14, in func_d
... Doesn't: matter
... """, strict=False)
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[42]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "bogus.py", line 123, in bogus
  File "...examples.py", line 2, in func_a
    func_b()
  File "...examples.py", line 6, in func_b
    func_c()
  File "...examples.py", line 10, in func_c
    func_d()
  File "...examples.py", line 14, in func_d
    raise Exception("Guessing time !")
Exception: fail

tblib.decorators.return_error

>>> from tblib.decorators import return_error
>>> inner_2r = return_error(inner_2)
>>> e = inner_2r()
>>> e
<tblib.decorators.Error object at ...>
>>> e.reraise()
Traceback (most recent call last):
  ...
  File "<doctest README.rst[26]>", line 1, in <module>
    e.reraise()
  File "...tblib...decorators.py", line 19, in reraise
    reraise(self.exc_type, self.exc_value, self.traceback)
  File "...tblib...decorators.py", line 25, in return_exceptions_wrapper
    return func(*args, **kwargs)
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

How’s this useful? Imagine you’re using multiprocessing like this:

# Note that Python 3.4 and later will show the remote traceback (but as a string sadly) so we skip testing this.
>>> import traceback
>>> from multiprocessing import Pool
>>> from examples import func_a
>>> pool = Pool()  # doctest: +SKIP
>>> try:  # doctest: +SKIP
...     for i in pool.map(func_a, range(5)):
...         print(i)
... except:
...     print(traceback.format_exc())
...
Traceback (most recent call last):
  File "<doctest README.rst[...]>", line 2, in <module>
    for i in pool.map(func_a, range(5)):
  File "...multiprocessing...pool.py", line ..., in map
    ...
  File "...multiprocessing...pool.py", line ..., in get
    ...
Exception: Guessing time !
<BLANKLINE>
>>> pool.terminate()  # doctest: +SKIP

Not very useful is it? Let’s sort this out:

>>> from tblib.decorators import apply_with_return_error, Error
>>> from itertools import repeat
>>> pool = Pool()
>>> try:
...     for i in pool.map(apply_with_return_error, zip(repeat(func_a), range(5))):
...         if isinstance(i, Error):
...             i.reraise()
...         else:
...             print(i)
... except:
...     print(traceback.format_exc())
...
Traceback (most recent call last):
  File "<doctest README.rst[...]>", line 4, in <module>
    i.reraise()
  File "...tblib...decorators.py", line ..., in reraise
    reraise(self.exc_type, self.exc_value, self.traceback)
  File "...tblib...decorators.py", line ..., in return_exceptions_wrapper
    return func(*args, **kwargs)
  File "...tblib...decorators.py", line ..., in apply_with_return_error
    return args[0](*args[1:])
  File "...examples.py", line 2, in func_a
    func_b()
  File "...examples.py", line 6, in func_b
    func_c()
  File "...examples.py", line 10, in func_c
    func_d()
  File "...examples.py", line 14, in func_d
    raise Exception("Guessing time !")
Exception: Guessing time !
<BLANKLINE>
>>> pool.terminate()

Much better !

What if we have a local call stack ?

>>> def local_0():
...     pool = Pool()
...     try:
...         for i in pool.map(apply_with_return_error, zip(repeat(func_a), range(5))):
...             if isinstance(i, Error):
...                 i.reraise()
...             else:
...                 print(i)
...     finally:
...         pool.close()
...
>>> def local_1():
...     local_0()
...
>>> def local_2():
...     local_1()
...
>>> try:
...     local_2()
... except:
...     print(traceback.format_exc())
Traceback (most recent call last):
  File "<doctest README.rst[...]>", line 2, in <module>
    local_2()
  File "<doctest README.rst[...]>", line 2, in local_2
    local_1()
  File "<doctest README.rst[...]>", line 2, in local_1
    local_0()
  File "<doctest README.rst[...]>", line 6, in local_0
    i.reraise()
  File "...tblib...decorators.py", line 20, in reraise
    reraise(self.exc_type, self.exc_value, self.traceback)
  File "...tblib...decorators.py", line 27, in return_exceptions_wrapper
    return func(*args, **kwargs)
  File "...tblib...decorators.py", line 47, in apply_with_return_error
    return args[0](*args[1:])
  File "...tests...examples.py", line 2, in func_a
    func_b()
  File "...tests...examples.py", line 6, in func_b
    func_c()
  File "...tests...examples.py", line 10, in func_c
    func_d()
  File "...tests...examples.py", line 14, in func_d
    raise Exception("Guessing time !")
Exception: Guessing time !
<BLANKLINE>

Other weird stuff

Clearing traceback works (Python 3.4 and up):

>>> tb = Traceback.from_string("""
... File "skipped.py", line 123, in func_123
... Traceback (most recent call last):
...   File "tests/examples.py", line 2, in func_a
...     func_b()
...   File "tests/examples.py", line 6, in func_b
...     func_c()
...   File "tests/examples.py", line 10, in func_c
...     func_d()
...   File "tests/examples.py", line 14, in func_d
... Doesn't: matter
... """)
>>> import traceback, sys
>>> if sys.version_info > (3, 4):
...     traceback.clear_frames(tb)

Credits

• mitsuhiko/jinja2 for figuring a way to create traceback objects.

Changelog

3.0.0 (2023-10-22)

• Added support for __context__, __suppress_context__ and __notes__. Contributed by Tim Maxwell in #72.

• Added the get_locals argument to tblib.pickling_support.install(), tblib.Traceback and tblib.Frame. Fixes #41.

• Dropped support for now-EOL Python 3.7 and added 3.12 in the test grid.

2.0.0 (2023-06-22)

• Removed support for legacy Pythons (2.7 and 3.6) and added Python 3.11 in the test grid.

• Some cleanups and refactors (mostly from ruff).

1.7.0 (2020-07-24)

• Add more attributes to Frame and Code objects for pytest compatibility. Contributed by Ivanq in #58.

1.6.0 (2019-12-07)

• When pickling an Exception, also pickle its traceback and the Exception chain (raise ... from ...). Contributed by Guido Imperiale in #53.

1.5.0 (2019-10-23)

• Added support for Python 3.8. Contributed by Victor Stinner in #42.

• Removed support for end of life Python 3.4.

• Few CI improvements and fixes.

1.4.0 (2019-05-02)

• Removed support for end of life Python 3.3.

• Fixed tests for Python 3.7. Contributed by Elliott Sales de Andrade in #36.

• Fixed compatibility issue with Twised (twisted.python.failure.Failure expected a co_code attribute).

1.3.2 (2017-04-09)

• Add support for PyPy3.5-5.7.1-beta. Previously AttributeError: 'Frame' object has no attribute 'clear' could be raised. See PyPy issue #2532.

1.3.1 (2017-03-27)

• Fixed handling for tracebacks due to exceeding the recursion limit. Fixes #15.

1.3.0 (2016-03-08)

• Added Traceback.from_string.

1.2.0 (2015-12-18)

• Fixed handling for tracebacks from generators and other internal improvements and optimizations. Contributed by DRayX in #10 and #11.

1.1.0 (2015-07-27)

• Added support for Python 2.6. Contributed by Arcadiy Ivanov in #8.

1.0.0 (2015-03-30)

• Added to_dict method and from_dict classmethod on Tracebacks. Contributed by beckjake in #5.