Skip to content

Instantly share code, notes, and snippets.

@carlsmith
Created April 5, 2017 02:08
Show Gist options
  • Star 25 You must be signed in to star a gist
  • Fork 4 You must be signed in to fork a gist
  • Save carlsmith/b2e6ba538ca6f58689b4c18f46fef11c to your computer and use it in GitHub Desktop.
Save carlsmith/b2e6ba538ca6f58689b4c18f46fef11c to your computer and use it in GitHub Desktop.
A Python function that does multiple string replace ops in a single pass.
import re
def replace(string, substitutions):
substrings = sorted(substitutions, key=len, reverse=True)
regex = re.compile('|'.join(map(re.escape, substrings)))
return regex.sub(lambda match: substitutions[match.group(0)], string)
@carlsmith
Copy link
Author

carlsmith commented Apr 5, 2017

If you ran the following code, output would be "spam FOO BAR FOO BAR spam".

string = "spam foo bar foo bar spam"
substitutions = {"foo": "FOO", "bar": "BAR"}
output = replace(string, substitutions)

Copy link

ghost commented Sep 12, 2018

Thanks! Nice Solution!!!!

@thirumalailucifer
Copy link

One of the best Method hats off to you !

@feldi-to-go
Copy link

Hi all!
I am a newly in programming Python, and found your code regarding easy replacing text, using Python.
Now I am facing a problem to use your code, where the substitutions are not pure text like "foo": "Foo", rather using regex to find special pattern.
For example: I have a string with a double tab's, double blank's, certain numbers or text, and those need to replace each of them different. The double tab with a single tab, the double blank with a single blank, all "006" with "007" and so on.
Is it possible, and how is it possible to modify you code to have a dictionary (or list) with regex-pattern and on the other hand, the related replacements?
I would appreciate to get some help on this topic.
Thank you in advance!

@ridhitatineni1
Copy link

This is such a cool solution! Thank you @carlsmith !

@lvrfrc87
Copy link

+1

@FritzPeleke
Copy link

hey @carl i have a question. I have a string say 'AGCGTGCGCGGACGCGTCGCGGCGCGCGCGCAC'. I want o substistute the regex 'AC' and 'GT'. However, i need GT and AC to be closer to one another. My string has several AT and GC. But i want to only replace the two occurance of these regex that are closer.

@dgr113
Copy link

dgr113 commented Sep 17, 2020

Thank you much!

This is a slightly edited solution that is a bit more suitable for production:

import re
from dataclasses import dataclass, field
from typing import Any, Mapping, Match, Pattern


@dataclass
class Parser:
    substitutions: Mapping[str, Any]
    patt: 'Pattern' = field(init=False)

    def __post_init__(self):
        self.patt = self.__patt_build()

    def __patt_build(self) -> Pattern:
        substrings = sorted(self.substitutions, key=len, reverse=True)
        regex = re.compile('|'.join(map(re.escape, substrings)))
        return regex

    def _repl_match(self, match: Match) -> str:
        match_group = match.group(0)
        result = self.substitutions[match_group]
        return str(result)

    def replace(self, string: str) -> str:
        return self.patt.sub(self._repl_match, string)



def main():
    context = {'@A': 111, '@B': {'B1': 333}}
    s = "{ 'A': @A, 'B': @B }"

    parser = Parser(context)
    result = parser.replace(s)
    print(result)


if __name__ == '__main__':
    main()

It also supports any (supports casting to string) types in substitutions :)

@carlsmith
Copy link
Author

@dgr113 - Being totally honest, I haven't studied your code properly, and it uses some stuff I'm not familiar with, so can't offer much in the way of feedback, but it's always nice to have different solutions from different perspectives, in the same language.

Personally, I prefer minimalist solutions that tend to rely on the same, few dozen standard libraries, but that's mainly to do with being lazy, and wanting to support likeminded people.

If you could explain a bit about the advantages of your solution, I'd be interested, and expect other readers would find it useful too. As it stands, it's hard to see the advantage of rewriting a three-line function as a rather complex looking class (where no two method names use the same combination of (presumably meaningful) underscores). The code is well written, so you seem to know what you're doing. I just don't get what it improves really.

@dgr113
Copy link

dgr113 commented Sep 19, 2020

@carlsmith
Firstly
Your code causes an error if there are non-string values in "substitutions". Numbers for example.
Secondly
In edited version, regex pattern (re.compile) is created once after object is created (and can be cached).
In your case - every time this function is called.
This can be quite an expensive operation. As far as I know, this was in an incorrect comparison of Rust and Go based on regular expressions example. In this "comparison" with permanent recompilation of regex pattern object in Rust code, Go managed to slightly overtake Rust.
In corrected testing (with an object created once), situation was reversed.

@carlsmith
Copy link
Author

carlsmith commented Sep 21, 2020

Your code causes an error if there are non-string values in "substitutions". Numbers for example.

Non-string values should be an error. I could just call str to convert the args to strings, but it makes more sense to leave that to the user, if that's their intention. I don't want to coerce the wrong types, as that's not part of the function's purpose.

In edited version, regex pattern (re.compile) is created once after object is created (and can be cached).

It would be trivial to wrap my function in a function that takes a regex, and returns my function (now only accepting the substitutions, and enclosing the regex):

def f(regex):
    def myfunc(subs): etc(regex, subs)
    return myfunc

It's two extra lines. I'm just not seeing the need for even a simple class, or the purpose of the wacky method names??

@dgr113
Copy link

dgr113 commented Sep 22, 2020

You have your point of view, and I have my...
The principles of code readability are important to me.
I try to use lambda functions as little as possible, because they degrade performance and do not support typing.
In addition, I prefer to immediately make the functionality more flexible for adding new features in the future. It seems to me that it is more convenient to add a new methods or protocols than to rewrite everything.
I myself don't really like to use OOP unnecessarily, but I think the principle of encapsulation (in methods for example) is the most important.

PS: I only suggested my version, including your solution. With hope that this will help someone, according to OpenSource spirit.
I'm not attacking your programming principles =)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment