new function proposal: find and replace about pyjanitor

[szuckerman]

So they’re kind-of different things. I think this answer gives a better description of the differences than I could write here:

https://askubuntu.com/questions/714503/regular-expressions-vs-filename-globbing

In general, when searching for things I find globbing more useful (cause you just need an asterisk) but for validation regex is better because one can be more precise with what is expected.

[szuckerman]

The above looks good, except “exact” is unnecessary. Regex or glob without wildcards will be equivalent.

[ericmjl]

Needs a column_name / col_name (?), right?

Thank you. Updating...

[zbarry]

Also think that maybe col_name should come before find and replace in the args list... From the at-a-glance understanding perspective, maybe df.do_this(to_this, with_this_particularly) is more in line with what we have done with other functions.

[ericmjl]

@zbarry good point! Updated the function signature.

[zbarry]

Should it support regular expressions?
🤔

[szuckerman]

Regex is ok.... but globbing is nicer:

https://docs.python.org/3.6/library/fnmatch.html

[ericmjl]

@zbarry I see you're saturated with relatives!

I think it's possible to support multiple ways of find and replace:

def find_replace(df, find, replace, kind='exact'):
    kinds=['exact', 'regex', 'glob']
    if kind not in kinds:
        raise ValueError(f'kind kwarg has to be one of {kinds}')
    if kind == 'exact':
        # implementation goes here
    if kind == 'regex':
        # implementation goes here
    if kind == 'glob':
        # implementation goes here
    return df

What do you all think?

[ericmjl]

Ok, got it! Pardon my ignorance here, but is globbing a "subset" of regexing, or there mutually exclusive parts to them?

[zbarry]

Yeah 90% of the time I just use globbing. I only use regex when I absolutely have to. No idea how often you would need them in pandas-style applications.

[ericmjl]

Ok! Synthesizing the discussion thus far:

def find_replace(df, col_name, find, replace, kind='glob'):
    """
    # docstring opener above
    Whether or not you use kind="glob" or "regex", if you do not use any wildcards,
    it will default to an exact match. We expect an 80/20 split between exact/wildcard
    usage, and with globbing being more easily understood than regexes, it is hence
    the default choice.
    # continue docstring below
    """
    kinds=['regex', 'glob']
    if kind not in kinds:
        raise ValueError(f'kind kwarg has to be one of {kinds}')
    if kind == 'regex':
        # implementation goes here
    else:
        # implementation goes here
    return df

Comments?

[zbarry]

Needs a column_name / col_name (?), right?