The Python Journey - Regular Expressions

Useful Links

Regular Expression HOWTO Python RE module Python Regex module Regexr Pyrexp

Regular expressions (often shortened to regex) are a way to look for matching patterns in any text. We can use the pattern to determine where the pattern appears in the text, or to do more sophisticated things like replacing patterns in text.

Page Contents

Patterns
Matching Patterns in Python
Naming Groups
Iterating through Matches
Replacing
More Examples
- Asserting the Match, and Mapping
- Using findall()
Examples

Patterns

A pattern is something we want to match within a string. Patterns can be simple, or complex. Patterns include the text we want to look for, along with metacharacters which have special meanings.

Check out this tutorial for a guide on how to build patterns.

Then, make a note of the awesome regexr.com, which is a great place to test and build your regular expressions. It also includes some really useful cheat sheets and references.

Matching Patterns in Python

Python provides a built-in library for working with regular expressions, called re. This will generally be good enough for most of our regular expression needs in Python. However, there are some niche cases where you want to do stuff that re doesn’t offer. In this case, try out the third party Python regex module, which is basically re on steroids. E.g. finding overlapping pattern matches.

In general, the approach to regex in Python is to compile the pattern, and then use one of handful of methods to apply the pattern to a string or strings.

For example:

import re

# Assume we've loaded in some multiline text data into the variable `data`

# We want to match rows of data that looks like: 5-7 z: qhcgzzz
# We want to obtain 5, 7, z, and qhcgzzz as four separate variables
matcher = re.compile(r"(\d+)-(\d+) ([a-z]): ([a-z]+)")
for row in data:
    match = matcher.match(row)
    min_val, max_val, policy_char, token_str = match.groups()

Here:

We wrap the pattern with r"regex", to avoid any need for convoluted escape characters. The r prefix turns the str into Python’s raw string format. In short, it’s just a good idea to always pass patterns in this raw format.
The match(row) method looks for the pattern within the text called row. In particular, the match() method will look for the match from the beginning of the line of data.
If we want to look for the pattern at any position in the data, we should use search() instead of match().
If match() or search() are successful, they return a match object. If not, they return None.
The regex pattern itself is split into several groups, by wrapping each group within parentheses, i.e. (group). When we call the groups() method against any successful match object, this returns a tuple of the four groups in our regex pattern.

It can be useful to perform assignment at the same time as checking if match object was returned. For example, here we will only enter the if block if a match was found. If a match was found, then the match object will have been assigned to the variable called match:

    if match := matcher.match(row):
        # do stuff with match object

We don’t have to compile the pattern in advance. For example, we can do this:

# We're looking for data like... "25,50 -> 30,600"
for line in data:
    x1, y1, x2, y2 = map(int, re.match(r"(\d+),(\d+) -> (\d+),(\d+)", line).groups())
    lines.append(Line(x1, y1, x2, y2))

Here, we’re:

Calling match() against re, rather than against a precompiled pattern.
We’re returning the four groups in the pattern.
We’re mapping all four groups from str type to int type, since we expect the data to always be numeric.
We use the four numbers - which are two pairs of x,y coordinates - to create a Line object.

Naming Groups

We can actually name groups in the regex pattern itself. Then, instead of calling groups() on a match object, we can instead call groupdict(). This returns a dictionary, where the keys are the names of the groups, and the values are the string values from the match.

Compare these two approaches:

import re

test = "John Smith"

# First, just using groups() and then unpacking the tuple
name_pattern = r"(\w+) (\w+)"
if (match := re.match(name_pattern, test)):
    first_name, last_name = match.groups()
    print(f"Unpacking groups(): {first_name}, {last_name}")

# Now, using named groups and returning a dict
name_pattern = r"(?P<first>\w+) (?P<last>\w+)"
if (match := re.match(name_pattern, test)):
    name_dict = match.groupdict()
    print(f"Using groupdict(): {name_dict['first']}, {name_dict['last']}")

Output:

Unpacking groups(): John, Smith
Using groupdict(): John, Smith

Even more usefully, we can actually embed Python variables within the pattern string. You can see how this can be useful if using groupdict():

test = "John Smith"

first_name_grp = "first"
last_name_grp = "last"
name_pattern = rf"(?P<{first_name_grp}>\w+) (?P<{last_name_grp}>\w+)"
if (match := re.match(name_pattern, test)):
    name_dict = match.groupdict()
    print("Using groupdict() with variables in the pattern: "
         f"{name_dict[first_name_grp]}, {name_dict[last_name_grp]}")

Note how we’re prefixing the pattern string with both r to make it raw, and f in order to use f-string interpoloation, i.e. so that we can reference variables like {first_name_grp} within the string.

Iterating through Matches

The finditer() function is useful for iterating over non-overlapping matches of a regular expression pattern within a given string. Its first parameter is the pattern to search for, and the second is the string to search. It returns an iterator that produces match objects for each match found.

For example:

import re

text = "Hello, Mycroft! Mycroft is a hunky cat."

pattern = r"Mycroft"

matches = re.finditer(pattern, text)

for match in matches:
    print("Match found:", match.group())
    print("Start position:", match.start())
    print("End position:", match.end())
    print("Match span:", match.span())
    print()

Here is the output:

Match found: Mycroft
Start position: 7
End position: 14
Match span: (7, 14)

Match found: Mycroft
Start position: 16
End position: 23
Match span: (16, 23)

Replacing

Use the sub() method to replace occurrences of a match with a replacement string.

For example:

line = re.sub(r"(\d+)", r"RULE_\1", line)

Here, any number that we find is replaced by “RULE_” + number. E.g. 15 becomes RULE_15.

The trick to this is to use \n to reference the nth group in the preceeding pattern.

This example turns any number n into X(n). E.g. 456 becomes X(456):

re.sub(r"(\d+)", r"X(\1)", input)

Here’s a more sophisticated example. It takes a string like:
= x yz | ab c
and replaces it with:
= ((x yz) / (ab c))

line = re.sub(r"= (.*) \| (.*)$", r"= ((\1) / (\2))", line)

More Examples

Asserting the Match, and Mapping

INSTR_PATTERN = re.compile(r"(\d+),(\d+) through (\d+),(\d+)")

for line in data:
    match = INSTR_PATTERN.search(line)
    assert match, "All instruction lines are expeted to match"
    tl_x, tl_y, br_x, br_y = map(int, match.groups())

Here we’re processing multiple lines of data. We’re looking for lines that contain something like:

4,14 through 6,16

The four (\d+) groups are used to capture each of the four numbers in the line.
We’re using an assert to validate that we found a match to our pattern in the current line. By including this assertion, we’re explicitly stating that we believe the input data is good. If any line fails to match the pattern, then the program will terminate with an AssertionError.
We’re using the groups() method to return the four groups as a tuple. We’re then using tuple unpacking to unpack the tuple into four separate variables.
Finally, because we know the four variables are all numbers, we’re using the map() function, to convert each variable from a str to an int. Here, the map() function is applying the int() function each member of the tuple that was passed in.

Using findall()

Here is another way to retrieve matches and their groups. Note how each line expects to return a single match, which is why we always index the return value of findall() with [0]. This match contains our four groups, as a tuple.

pattern = re.compile(r"(\d+),(\d+) through (\d+),(\d+)")
print("\nUsing findall with existing pattern:")
for line in data:
    tl_x, tl_y, br_x, br_y = map(int, pattern.findall(line)[0])
    print(f"tl_x: {tl_x}, tl_y: {tl_y}, br_x: {br_x}, br_y: {br_y}")
    
print("\nUsing findall, pattern on the fly:")
for line in data:
    tl_x, tl_y, br_x, br_y = map(int, re.findall(r"(\d+),(\d+) through (\d+),(\d+)", line)[0])
    print(f"tl_x: {tl_x}, tl_y: {tl_y}, br_x: {br_x}, br_y: {br_y}")

The next two two blocks of code achieve the same outcome. The first obtains a match and then the groups:

    boxes = []

    p = re.compile(r"(\d+)x(\d+)x(\d+)")  # our regex returns a match containing three groups
    for line in data:
        if match := p.match(line):
            dims = list(map(int, match.groups())) # turn tuple of str into ints
            boxes.append(Box(dims))

And here we use findall(), which circumvents the need to first get the match.

    boxes = []

    p = re.compile(r"(\d+)x(\d+)x(\d+)")  # our regex returns a match containing three groups
    for line in data:
        dims = list(map(int, p.findall(line)[0]))
        boxes.append(Box(dims))

Examples

‹›