Add a setup.py file

[timvieira]

Thanks for the great package. In an effort to make the package easier to use, I'd strongly recommend making it a python package by simply adding a setup.py file, which should (minimally) look like:

from setuptools import setup
setup(name='berkeley_parse_analyser',
      version='1.0',
      author='Jonathan Kummerfeld',
      description='A tool for classifying mistakes in the output of parsers',
      packages=['berkeley_parse_analyser'])

Note: It's conventional rename the src directory to the name of the project, e.g., berkeley_parse_analyser.

It would also be a good idea to run this code through pylint and the pep8 coding style checker (which pylint will do for you).

[jkkummerfeld]

Thanks for the suggestion! I followed your suggestions to make this change, and started making fixes as recommended by pylint. There are many more to address (and in an ideal world I would also make this python 3 compatible), but that will have to wait for another day.

[timvieira]

Awesome! Thanks for replying so quickly :-D I'm glad you like the suggestions. Another one I'd recommend is using python classes instead of dicts for storing info, e.g. here

def gen_missing_successor(ctree, error):
    # ...
    info = {
        'type': 'add',
        'label': get_label(nnode),
        'span': nnode.span,
        'subtrees': [get_label(subtree) for subtree in nnode.subtrees],
        'parent': nnode.parent.label,
        # etc...
    }
    # ...

Becomes:

class Transform:
    def __init__(self, nnode, ttype)
        self.type = ttype
        self.label = get_label(nnode)
        self.span = nnode.span
        self.subtrees = [get_label(subtree) for subtree in nnode.subtrees]
        self.parent = nnode.parent.label
        # etc...

def gen_missing_successor(ctree, error):
    # ...
    info = Transform(nnode, ttype='add')
    # ...

[jkkummerfeld]

Yeah... that was definitely a case of a quick and dirty solution that grew and grew.

[timvieira]

Ha! I know that phenomena all too well. Using dictionaries that way is very "Perl style" (as is quick 'n dirty). I thought maybe you were be a Perl hacker writing Python.

[jkkummerfeld]

I've written practically no Perl actually, but with sufficiently little effort poor programming style can be achieved in any language :)