Add data access through BIDSFile

[tyarkoni]

For BIDSFile objects associated with files containing text data (e.g., .tsv, tsv.gz, .json, etc.), users should be able to directly retrieve the contents (as a DataFrame or dict) via a .get_data() method.

[effigies]

Is the polymorphism based on file extension a good idea? Suppose I do a query that can get multiple file types and don't think to check. If I always (or at least in some mistaken cases) get something useful with .get_data(), it might be a few steps before I try to do something with it that raises an error.

For example, if I thought I was reading a JSON file but instead read a TSV file, I might try to set a value in the dict, which actually creates a column in the DataFrame.

We could instead try something like:

def get_df(self):
    if self.ext not in ('.tsv', '.tsv.gz'):
        raise ValueError
    # return pandas.DataFrame

def get_text(self):
    # return contents as str
def get_dict(self):
    # return dict interpretation
def get_image(self):
    # return nibabel image

Some of these could still apply to multiple image types, but the output type would be predictable.

[tyarkoni]

I'm not sure I see much of a use case for get_dict or get_text (they're a single line of code to read in, whereas updating the DF columns etc. based on metadata can involve a bit more work), and having a whole family of methods seems to me to unnecessarily clutter the API. But I'd be fine saying that you can only get back a pandas DF from get_data(), and anything else will raise an error (this is actually how I just started to implement it). I'd also be fine renaming it to get_df() to reduce ambiguity. We can always add other methods later.

[effigies]

Sure, makes sense. I would prefer get_df() to be explicit. I think we do provide access to nibabel images already.

[tyarkoni]

Yeah, there are get_image() and get_metadata() methods.

[tyarkoni]

Closed in #432.