Redex is a Python library that aims to be a readable and scalable replacement of regular expressions (regex). It uses boolean operators and built-in functionalities to create expressions.
Redex can be installed through the Python package installer (PyPi). It doesn't have any dependencies outside of the standard Python library. It requires Python 3.8 as a minimum Python version.
pip3 install python-redexNext, importing Redex can be done as follows.
import redex as rdRedex has three main actions: has (returns boolean), find (returns list) and count (returns int). For example, a Redex query that finds all words that start with an uppercase character and contain an hyphen can be formulated as follows.
rd.find('startswith:*upper and contains:-', string)Next, a Redex query that checks for the occurence of characters on specific locations and max proximity of upper- and lowercase characters can be formulated as follows.
rd.find('(location:{o,1} or location:{a,1}) and proximity:{*upper,*lower}3', string)Finally, a (commonly used) Redex query that finds email addresses in text can be formulated as follows.
rd.find('sequence:{*alpha,@,.com} or sequence:{*alpha,@,.co.uk}', string)A complete tutorial/list of examples can be found in demo.ipynb (see files of this repository).
This section documents the different components of Redex. Namely, the splitter, the built-in actions, the wildcards and the multi-threaded searching.
The splitter creates subqueries (often words or sentences) to conduct the Redex queries on. By default, this is set to a space (' ') with a granularity of 1. Setting multiple splitting characters can be done through adding a list. For example, adding punctuation: split=['.',',',' '].
Next, the granularity refers to the amount of splitting characters needed to actually split the string in subqueries. By default, this is set to 1 (and it can only be increased). For example, findings substrings where two consecutive words start with an uppercase character can be formulated as follows rd.find('count:{*upper,2}', string, granularity=2)
Redex has a number of built-in functionalities that can be used in a boolean expression/Redex query. This subsection gives an overview of these functionalities.
| Name | Description |
|---|---|
| startswith:substring | Return True if string starts with substring |
| endswith:substring | Return True if string ends with substring |
| contains:substring | Return True if string contains substring |
| count:{substring,int} | Return True if string contains at least n occurences of the substring |
| proximity:{substring1, substring2,...}int | Return True if max proximity between substring is less than the threshold/int |
| sequence:{substring1,substring2,...} | Return True if sequence of substrings exists in string |
| location:{substring,location} | Return True if substring occurs on location |
By default, the substrings can be implemented using the following wildcards. Note, you can add wildcards yourself. For example, say you want to add a wildcard for countries that you can do that as follows: rd.wildcard['*country'] = ['United States','Canada','United Kingdom'].
| Name | Description |
|---|---|
| * | Any thing |
| *num | Any number |
| *alpha | Any alphabetic character |
| *upper | Any uppercase character |
| *lower | Any lowercase character |
| *special | Any special character |
| *space | Any whitespace |
| *punt | Any punctuation |
The search operations can be executed conurrently. This allows Redex to scale better than regular expressions. However, finding the optimal number of threads can be a tedious process, since it requires knowledge of your CPU (cores/threads). Hence, as a general rule of thumb, don't touch this unless you really know what you're doing.
# increases the search operations executed concurrently to 4
rd.find('count:{*upper,2}', string, granularity=2, threads=4)This library is maintained by Timo Kats as a side project. If you have any suggestions feel free to reach out or fork the repository.
Future updates
- Error handling
- Multiple wildcards