Show the shape of numpy array & tensorflow Tensor

[QAQ-v]

Is there any solution to show the variable shape directly?

[cool-RR]

Maybe you can make your own function show and then pass watch='show(my_array)' to PySnooper. Other than that I don't want to add special cases.

[alexmojaki]

watch='arr.shape' should work.

[jh-yi]

I'm also looking for the answer. And here might be one possible solution: (at least works in my case)

1. Go to the declaration of function pysnooper.snoop() (in Pycharm, just ctrl+B+single click this function)

2. Go to the declaration of class Tracer (line 69 in pysnooper.py)
   Or you can directly find the file in path_to_env_where_you_install_pysnooper_/env_name/lib/python2.7/site-packages/pysnooper/tracer,py (python2.7 for example)

3. Modifiy following sentences:

 
# Line: 273
self.write('{indent}{newish_string}{name} = {value_repr}'.format(**locals()))

# replace with
try:
    self.write('{indent}{newish_string}{name} = {}, {value_repr}'.format(locals()['frame'].f_locals[name].shape, **locals()))
except AttributeError:   # in case built-in objects don't have .shape, e.g. int, list
    self.write('{indent}{newish_string}{name} = {value_repr}'.format(**locals()))

and

 
# Line: 276
self.write('{indent}Modified var:.. {name} = {value_repr}'.format(**locals()))

# replace with
try:
    self.write('{indent}Modified var:.. {name} = {}, {value_repr}'.format(locals()['frame'].f_locals[name].shape, **locals()))
except AttributeError:  # in case built-in objects don't have .shape, e.g. int, list
    self.write('{indent}Modified var:.. {name} = {value_repr}'.format(**locals()))

hope this helps.

[cool-RR]

I think you'll be better off monkeypatching get_shortish_repr. (Ensure you're on the latest PySnooper version.)