re.RegexFlag
Five flags cover almost everything: I ignores case, M makes ^ and $ work per line, S lets . match newlines, X allows whitespace and comments, A restricts \w \d \s to ASCII. Each has a one-letter alias and an inline form.
Common call
re.findall(r'^\w+', text, flags=re.I | re.M)
Returns
re.IGNORECASE|re.MULTILINE — flags combine with |
Replaces
lowercasing the text first, splitting into lines first
Watch out
Pass flags by keyword to sub/split: the 4th positional argument of re.sub is count, not flags
re.IGNORECASE | re.MULTILINE | re.DOTALL | re.VERBOSE | re.ASCII | re.UNICODE | re.LOCALE | re.NOFLAG
→ re.RegexFlag
Demo
Live evaluation
Without and with re.IGNORECASE.
Try:
Inputs
patternstra regular expression
textstrtext
Code
import re (re.findall('python', 'Python PYTHON python'), re.findall('python', 'Python PYTHON python', re.IGNORECASE))
Result
(['python'], ['Python', 'PYTHON', 'python'])
IGNORECASE uses Unicode case rules for str patterns, so straße matches STRAßE and ω matches Ω. MULTILINE only changes ^ and $; DOTALL only changes the dot — they are independent and are often used together. In VERBOSE mode a literal space must be escaped (\ ) or written as [ ] — and 'New\ York | Paris' shows alternation still works with the layout spaces removed.
Parameters
| Name | Type | Required | Description |
|---|---|---|---|
| re.I / re.IGNORECASE | RegexFlag = 2 | no | Case-insensitive matching (Unicode case folding for str patterns). Inline: (?i). |
| re.M / re.MULTILINE | RegexFlag = 8 | no | ^ and $ also match at the start/end of every line. Inline: (?m). |
| re.S / re.DOTALL | RegexFlag = 16 | no | . also matches a newline. Inline: (?s). |
| re.X / re.VERBOSE | RegexFlag = 64 | no | Whitespace in the pattern is ignored and # starts a comment (except inside [...] or when escaped). Inline: (?x). |
| re.A / re.ASCII | RegexFlag = 256 | no | \w \W \b \B \d \D \s \S and case-insensitivity use ASCII only. Inline: (?a). |
| re.U / re.UNICODE | RegexFlag = 32 | no | The default for str patterns; kept for compatibility. Incompatible with re.ASCII. |
| re.L / re.LOCALE | RegexFlag = 4 | no | Locale-dependent matching — bytes patterns only; a ValueError with str. |
| re.NOFLAG | RegexFlag = 0 | no | No flags (3.11+) — a readable default for code that adds flags conditionally. |
Return value
re.RegexFlag — An enum.IntFlag; combine members with | and pass them as flags= (or put (?imsxa) at the start of the pattern).
Common patterns
Combine flags
Flags are IntFlag members: | combines them.
import re pat = re.compile(r'^error:.*$', re.IGNORECASE | re.MULTILINE)
A documented pattern with VERBOSE
Comments and line breaks make long patterns maintainable.
import re DATE = re.compile(r''' (?P<year>\d{4}) - # year (?P<month>\d{2}) - # month (?P<day>\d{2}) # day ''', re.VERBOSE)
Inline flags and scoped flags
(?i) at the very start applies to the whole pattern; (?i:...) only to the group (3.6+).
import re re.compile(r'(?i)^subject:') re.compile(r'ID-(?i:[a-f0-9]+)')
Build flags conditionally
re.NOFLAG (3.11+) is a clear starting value.
import re flags = re.NOFLAG if ignore_case: flags |= re.IGNORECASE pat = re.compile(needle, flags)
Examples
1. Case-insensitive
import re
re.findall(r'python', 'Python PYTHON python', re.I)
Returns
['Python', 'PYTHON', 'python']2. ^ on every line
import re
re.findall(r'^\w+', 'one\ntwo\nthree', re.M)
Returns
['one', 'two', 'three']3. . across a newline
import re
re.search(r'a.b', 'a\nb', re.S)
Returns
<re.Match object; span=(0, 3), match='a\nb'>4. ASCII-only \w
import re
re.findall(r'\w+', 'café naïve', re.A)
Returns
['caf', 'na', 've']5. Inline flag
import re
re.findall(r'(?i)python', 'Python')
Returns
['Python']6. Combined flags repr
import re
re.I | re.M
Returns
re.IGNORECASE|re.MULTILINE7. Flags are ints
import re
(int(re.I), int(re.M), int(re.S), int(re.X), int(re.A))
Returns
(2, 8, 16, 64, 256)8. Unicode digits by default
import re
(re.findall(r'\d', '1٣'), re.findall(r'\d', '1٣', re.A))
Returns
(['1', '٣'], ['1'])Pitfalls
1. Flags passed positionally to re.sub
re.sub's fourth parameter is count. re.I is 2, so this means "replace at most 2" — and nothing is case-insensitive.
re.sub(p, r, s, re.I)
import re re.sub('a', '-', 'AaAa', re.I)
'A-A-'
flags=re.I
import re re.sub('a', '-', 'AaAa', flags=re.I)
'----'
2. Inline global flags in the middle
Since 3.11 a global (?i) must be at the start. Use a scoped group (?i:...) elsewhere.
a(?i)b
import re re.compile(r'a(?i)b')
re.PatternError: global flags not at the start of the expression at position 1
a(?i:b)
import re re.findall(r'a(?i:b)', 'ab aB Ab')
['ab', 'aB']
3. Expecting MULTILINE to make . cross lines
MULTILINE is about ^ and $. The dot needs DOTALL.
re.M
import re print(re.search(r'start.*end', 'start\nend', re.M))
None
re.S
import re re.search(r'start.*end', 'start\nend', re.S)
<re.Match object; span=(0, 9), match='start\nend'>
When to use
Use it
- IGNORECASE for user-facing search; MULTILINE for line-oriented text
- DOTALL when a match may span lines; VERBOSE for any pattern longer than one line
- ASCII when \w and \d must mean [a-zA-Z0-9_] and [0-9] (identifiers, protocols)
Reach for something else
- LOCALE with str patterns — not allowed
- UNICODE — it is already the default for str
- IGNORECASE as a substitute for proper normalization (casefold, unicodedata)
Notes
CPython impl
re.RegexFlag is an enum.IntFlag defined in Lib/re/__init__.py; the bit values come from Lib/re/_constants.py (SRE_FLAG_*)
Inline
(?aiLmsux) at the start of the pattern sets flags for the whole pattern; (?imsx-imsx:...) sets or clears them inside a group (a, L, u can only be set, not cleared)
Versions
Flags became RegexFlag enum members in 3.6; NOFLAG was added in 3.11; since 3.11 global inline flags must be at the start
Pattern.flags
A compiled str pattern reports re.UNICODE (32) as well, even when you did not pass it
DEBUG
re.DEBUG (128) prints the parsed pattern while compiling — a debugging aid, not part of __all__
FAQ
Pass re.IGNORECASE (or re.I): re.search(r'python', text, re.I). Or start the pattern with (?i). For part of a pattern use a scoped group: (?i:python).