Learning about regular expressions

malanon · February 11, 2019, 5:27am

Hi there, I am just getting started in this course and learning about deep learning. There are many links to video tutorials and posts about regular expressions in the Resources of Lesson 1. Should I invest a good amount of time learning about regular expressions now, or can it wait? Is the usage of regular expressions common in deep learning? I am trying to make the best use of my time and not get too into the weeds just yet.

Thanks,
Andrew

wyquek · February 11, 2019, 6:54am

short answer, it can wait

mejri · February 11, 2019, 2:58pm

Regular expressions are just a tool in certain tasks. You don’t need to invest much time on that.The best way to learn fast is by practice. Whenever you need to find a certain pattern in text try to do it using a regex and there is a good website https://regex101.com/ that can guide you and help you debug and understand each regex you type

MadeUpMasters · February 11, 2019, 6:15pm

I had this same question in my head about two weeks ago. It definitely can wait, but it’s not as hard as it looks.

There is a fantastic interactive tutorial at regexone.com. It will walk you through what everything means and give you problems to solve (that reflect what your regex pattern is matching in real time). Maybe takes an hour to an hour and a half. Hope that gives you more info to make a decision

MadeUpMasters · February 11, 2019, 6:23pm

On a separate note, I think everyone comes into this class from different backgrounds. Many of us haven’t had exposure to the large number of technologies that, while not essential, are pretty important for deep learning. My personal list includes…

Git/Github
Numpy
Pandas
MatPlotLib
Jupyter Notebook
Regular Expressions
A remote editor like vim/nano/sublime
…

I think it would be great to have a thread or wiki with the best resources and tutorials for these mini-topics, just like we do for the recommended python resources. It would help people who are new to a lot more than AI to not get overwhelmed knowing there’s a neat list they can bite off one chunk at a time.

Finally it should include a warning that these are not in any way prerequisites that should take precedence over the lessons/code, but instead are meant to be tackled gradually based on what your most pressing AI needs are.

Let me know if this already exists somewhere and if not let’s build it!

malanon · February 18, 2019, 7:10pm

I am definitely late in replying here but thank you all for the feedback! Very good input to help me prioritize my learning time.

Sree · May 28, 2019, 2:41pm

Hi. Im trying to write a regex for a filename of type “word-word-digits.jpg”.
My regex - ^\w+-\w+-\d+.(jpg|jpeg)$ . I checked it on https://regex101.com and it shows its a match with my string provided. But when i use it in the notebook i get an error " Failed to find “re.compile” " . Could someone help me with this. Thanks!

ste · May 28, 2019, 3:05pm

WARNING: OPINIONATED OPINION
I try to prefer functions/lambdas over regular expression; according to my experience they’re simpler to maintain.

So if you already know RegExp, use them, otherwise use functions and focus on more interesting topics. As far as I remember I’ve never use regexp in any fast.ai notebook that I made.

NOTE: It’s fun to notice that from_re actually wraps from_func

    def label_from_re(self, pat:str, full_path:bool=False, label_cls:Callable=None, **kwargs)->'LabelList':
        "Apply the re in `pat` to determine the label of every filename.  If `full_path`, search in the full name."
        pat = re.compile(pat)
        def _inner(o):
            s = str((os.path.join(self.path,o) if full_path else o).as_posix())
            res = pat.search(s)
            assert res,f'Failed to find "{pat}" in "{s}"'
            return res.group(1)
        return self.label_from_func(_inner, label_cls=label_cls, **kwargs)

github.com

fastai/fastai/blob/c0f9a774f3760232bb2633eeb0f09fb262d7ab18/fastai/data_block.py#L305




def label_from_func(self, func:Callable, label_cls:Callable=None, **kwargs)->'LabelList':
    "Apply `func` to every input to get its label."
    return self._label_from_list([func(o) for o in self.items], label_cls=label_cls, **kwargs)


def label_from_folder(self, label_cls:Callable=None, **kwargs)->'LabelList':
    "Give a label to each filename depending on its folder."
    return self.label_from_func(func=lambda o: (o.parts if isinstance(o, Path) else o.split(os.path.sep))[-2],
                                label_cls=label_cls, **kwargs)


def label_from_re(self, pat:str, full_path:bool=False, label_cls:Callable=None, **kwargs)->'LabelList':
    "Apply the re in `pat` to determine the label of every filename.  If `full_path`, search in the full name."
    pat = re.compile(pat)
    def _inner(o):
        s = str((os.path.join(self.path,o) if full_path else o).as_posix())
        res = pat.search(s)
        assert res,f'Failed to find "{pat}" in "{s}"'
        return res.group(1)
    return self.label_from_func(_inner, label_cls=label_cls, **kwargs)


def databunch(self, **kwargs):

RJSD3V · April 7, 2020, 4:21pm

Its very simple, if you simply learn the syntax of writing regular expressions and how to do it in python, you’ll be all set. Its not that hard at all. regular expressions are just tools used for looking for patterns in text. follow this simple chart apart from basic definitions and you’ll understand everything. 1944b14175b0136b578287c0169a5c69