Does dill have the register_pickle_by_value cloudpickle equivalent ?

[shashwath94]

I'm trying to register modules before pickling a class instance object. I used to achieve that with the register_pickle_by_value function, is there an equivalent in dill that I can use?

cloudpickle example

import cloudpickle
import module1
import module2
from module3 import Class

cloudpickle.register_pickle_by_value(module1)
cloudpickle.register_pickle_by_value(module2)

object = Class(**params)

with open("class.pkl", "wb") as f:
    cloudpickle.dump(object, f)

Can I do the same thing but with dill instead of cloudpickle?

[mmckerns]

So, yes. Several methods, actually.
Given a module (here module.py), you can use dill.dump -- the result includes search locations and picked objects, or you can use dill.source.importable (or dill.source.getsource) to get the source of the module.

>>> import dill
>>> import module
>>> dill.dumps(module)
b'\x80\x04\x95E\x02\x00\x00\x00\x00\x00\x00\x8c\ndill._dill\x94\x8c\x0e_import_module\x94\x93\x94\x8c\x06module\x94\x85\x94R\x94}\x94(\x8c\x08__name__\x94h\x03\x8c\x07__doc__\x94\x8c\x0f\nsimple module\n\x94\x8c\x0b__package__\x94\x8c\x00\x94\x8c\x08__spec__\x94\x8c\x11_frozen_importlib\x94\x8c\nModuleSpec\x94\x93\x94)\x81\x94}\x94(\x8c\x04name\x94h\x03\x8c\x06loader\x94\x8c\x1a_frozen_importlib_external\x94\x8c\x10SourceFileLoader\x94\x93\x94)\x81\x94}\x94(h\x12h\x03\x8c\x04path\x94\x8c8/Users/mmckerns/tmp/module.py\x94ub\x8c\x06origin\x94h\x1a\x8c\x0cloader_state\x94N\x8c\x1asubmodule_search_locations\x94N\x8c\r_set_fileattr\x94\x88\x8c\x07_cached\x94\x8cP/Users/mmckerns/tmp/__pycache__/module.cpython-38.pyc\x94\x8c\r_initializing\x94\x89ub\x8c\x08__file__\x94h\x1a\x8c\n__cached__\x94h \x8c\x05Power\x94h\x03h$\x93\x94\x8c\x07squared\x94h%)\x81\x94}\x94\x8c\x01n\x94K\x02sb\x8c\x05cubed\x94h%)\x81\x94}\x94h)K\x03sbub.'
>>> dill.source.importable(module)
'import module\n'
>>> dill.source.importable(module, source=True)
"'''\nsimple module\n'''\n\nclass Power(object):\n  def __init__(self, n):\n    self.n = n\n  def __call__(self, x):\n    _type = type(x)\n    import numpy as np\n    return np.power(x, self.n).tolist()\n\nsquared = Power(2)\ncubed = Power(3)\n"

It's also possible to use dill.dump_module for more complex options.

Please reopen if you have further questions with regard to the above.