[DEP] NumPy required in base install

[mmaybeno]

Describe the bug
The base installation requires numpy. Specifically importing the main package calls __init__.py which imports benchmarks which imports numpy. This was noticed in a downstream package that has PyThaiNLP as a requirement.

To Reproduce

# Install with no extras
pip install pythainlp

>>> python
Python 3.7.6 (default, Dec 30 2019, 19:38:26) 
[Clang 11.0.0 (clang-1100.0.33.16)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pythainlp
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/maybeno/workspace/Envs/newspaper-nvz7rtlC/lib/python3.7/site-packages/pythainlp/__init__.py", line 26, in <module>
    from pythainlp.benchmarks import benchmark
  File "/Users/maybeno/workspace/Envs/newspaper-nvz7rtlC/lib/python3.7/site-packages/pythainlp/benchmarks/__init__.py", line 1, in <module>
    from .word_tokenization import benchmark
  File "/Users/maybeno/workspace/Envs/newspaper-nvz7rtlC/lib/python3.7/site-packages/pythainlp/benchmarks/word_tokenization.py", line 6, in <module>
    import numpy as np
ModuleNotFoundError: No module named 'numpy'

Expected behavior
Run import pythainlp without error.

Source Code
https://github.com/PyThaiNLP/pythainlp/blob/dev/pythainlp/__init__.py#L26

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: MacOS
• Python Version 3.7
• Version 2.1.3

Additional context
Add any other context about the problem here.

[wannaphong]

Thanks for your report. We will fix the issue in the next release.

[bact]

Or I think we can consider as well to remove from pythainlp.benchmarks import benchmark from pythainlp/__init__.py.

When install with pip install pythainlp[benchmarks], ["numpy>=1.16", "pandas>=0.24"] will be auotmatically installed.

[mmaybeno]

Ya I think having a the default package without numpy dependencies would be ideal. Numpy is great but kind of overkill if you're not using the benchmarks. Some added messaging if you try to import benchmark without numpy would be great. Thanks again!

[bact]

Thank you. Good idea for the message. Will try to do that.
I think we can use ModuleNotFoundError here.

[PNNutkung]

There is another problem.

The base installation requires pandas. Specifically importing the main package calls init.py which imports benchmarks which import pandas as pd. This was noticed in a downstream package that has PyThaiNLP as a requirement.

ModuleNotFoundError: No module named 'pandas'

[bact]

The base installation requires pandas. Specifically importing the main package calls init.py which imports benchmarks which import pandas as pd. This was noticed in a downstream package that has PyThaiNLP as a requirement.

ModuleNotFoundError: No module named 'pandas'

Thanks. Do you mean this file? https://github.com/PyThaiNLP/pythainlp/blob/dev/pythainlp/__init__.py If that the case, the from pythainlp.benchmarks import benchmark has been removed.

[PNNutkung]

The base installation requires pandas. Specifically importing the main package calls init.py which imports benchmarks which import pandas as pd. This was noticed in a downstream package that has PyThaiNLP as a requirement.

ModuleNotFoundError: No module named 'pandas'

Thanks. Do you mean this file? https://github.com/PyThaiNLP/pythainlp/blob/dev/pythainlp/__init__.py If that the case, the from pythainlp.benchmarks import benchmark has been removed.

Yes, it is. Did it release on PyPI?