whos enhancement to pyttb utils

[DeepBlockDeepak]

Though there isn't a direct Python analogue to MATLAB's whos, this is a first stab at how to deal with it in Python. This specifically regards the whos use in the ttb docs.

At the moment it looks like:

def whos(*args):
    print(f"{'Name': <10}{'Size': <10}{'Bytes': <10}{'Class': <10}{'Num. Attributes': <5}")
    print("-"*55)
    for name in args:
        var = globals()[name]
        size = 'x'.join(map(str, var.shape)) if hasattr(var, 'shape') else 'N/A'
        attributes = [attr for attr in dir(var) if not attr.startswith('__')]
        print(f"{name: <10}{size: <10}{str(sys.getsizeof(var)): <10}{str(type(var).__name__): <10}{len(attributes): <5}")

• whos() can take any number of arguments, like in the ttb doc where two tensor class objects are passed.
• The global symbol table allows us to fetch the corresponding variables in args.
• The size of each variable is obtained, so long as it has the tensor-class' shape attribute. Since shape is a tuple, its elements are mapped to string and then joined to give us the MATLAB-styled tensor dimensions.
• Since I couldn't seem to find a direct analogue to the MATLAB-whos Attribute category, I decided to count the number of attributes belonging to the object, filtering out the dunder methods.
• The print statement aligns the data to the column names. The Class field uses only the name of the type class for brevity.

[dmdunla]

The %whos Jupyter notebook command provides sufficient functionality for now:

import pyttb as ttb
X = ttb.tenrand((2,2,2))

%whos

ariable   Type      Data/Info
------------------------------
X          tensor    tensor of shape 2 x 2 x 2<...>0.39772558 0.41173942]]\n
ttb        module    <module 'pyttb' from 'C:\<...>ttb\\pyttb\\__init__.py'>

[dmdunla]

See PR #135 for an draft implementation.