RegExp testing fails when string is validated in RegExp tools

Question

I have a small regexp that should verify if a commits subject adheres to the ReactJS commit message format. Since the expression works with my test strings the code left me baffled.

This small example should reproduce the behaviour:

#!/bin/bash

function test_subject {
  local subject="$1"
  local pattern="^(feat|fix|docs|style|refactor|test|chore)\([a-zA-Z0-9._-]+\): [^\n]+$"

  if ! [[ $subject =~ $pattern ]]; then
    echo "Invalid subject: $subject"
  else
    echo "  Valid subject: $subject"
  fi
}

test_subject "chore(gh-actions): add script for commit check"
test_subject "chore(gh-actions): add script for commit checking"
test_subject "feat(ABC-123): add new feature"
test_subject "fix(ABC123): add new feature"
test_subject "fix(ABC123): fix previously added feature"
test_subject "fix(scope): fix bug"

This leads to the following output:

  Valid subject: chore(gh-actions): add script for commit check
Invalid subject: chore(gh-actions): add script for commit checking
Invalid subject: feat(ABC-123): add new feature
Invalid subject: fix(ABC123): add new feature
  Valid subject: fix(ABC123): fix previously added feature
  Valid subject: fix(scope): fix bug

Change your regex to: '^(feat|fix|docs|style|refactor|test|chore)$[a-zA-Z0-9._-]+$: .+$' — anubhava
– anubhava, Commented Dec 18, 2024 at 18:42

anubhava · Accepted Answer · 2024-12-24 13:36:59Z

6

You will need to use . instead of [^\n] in your shell regex to match any character.

rx="[^\n]" doesn't mean any character except newline. It is seeing \ and n as two separate characters hence [^\n] is seen as any character other than \ and n.

Note that your example strings number 2, 3, and 4 have the letter n somewhere after matching : hence it is not matching till end of line and $ assertion in the end fails the match.

This should work for you:

test_subject() {
  local subject="$1"
  local pattern="^(feat|fix|docs|style|refactor|test|chore)\([a-zA-Z0-9._-]+\): .+$"

  if ! [[ $subject =~ $pattern ]]; then
    echo "Invalid subject: $subject"
  else
    echo "  Valid subject: $subject"
  fi
}

test_subject "chore(gh-actions): add script for commit check"
test_subject "chore(gh-actions): add script for commit checking"
test_subject "feat(ABC-123): add new feature"
test_subject "fix(ABC123): add new feature"
test_subject "fix(ABC123): fix previously added feature"
test_subject "fix(scope): fix bug"

Output:

Valid subject: chore(gh-actions): add script for commit check
Valid subject: chore(gh-actions): add script for commit checking
Valid subject: feat(ABC-123): add new feature
Valid subject: fix(ABC123): add new feature
Valid subject: fix(ABC123): fix previously added feature
Valid subject: fix(scope): fix bug

edited Dec 24, 2024 at 13:36

answered Dec 18, 2024 at 18:46

anubhava

790k67 gold badges603 silver badges671 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

M. Nejat Aydin Dec 20, 2024 at 14:44

"[^\n] is being evaluated as [^n]" may not be correct: pat="[^\n]"; [[ '\' =~ $pat ]] && echo matches. The backslash is a literal character here. Compare with pat="[^n]"; [[ '\' =~ $pat ]] && echo matches

anubhava Dec 23, 2024 at 17:30

Please compare output of [[ 'renew' =~ [^\n]+ ]] && declare -p BASH_REMATCH and [[ 'renew' =~ [^n]+ ]] && declare -p BASH_REMATCH in your bash. For me it shows declare -ar BASH_REMATCH='([0]="re")' both the times.

M. Nejat Aydin Dec 23, 2024 at 18:47

That is different than my example. In my and OP's example, the pattern is embedded to a variable and the backslash is a literal character in this context. You may want to compare output of pat="[^\n]+"; [[ 'r\enew' =~ $pat ]] && declare -p BASH_REMATCH and pat="[^n]+"; [[ 'r\enew' =~ $pat ]] && declare -p BASH_REMATCH

anubhava Dec 24, 2024 at 13:37

Ah ok, I get your point. Thanks for pointing it out. I changed explanation in my answer to make it clear.

White Owl · Accepted Answer · 2024-12-18 19:11:54Z

4

Bash regexp does not know \n to be a new line character. The [^\n] is just a [^n] so the lines which your script marks as invalid - have "n" in them ("checking", "new").

The new line bash regexp do recognize is actually $ - end of the line.

Another point - your test strings do not have \n characters in them, so there is no real need to check for its absence - hence a simple <complex_pattern>: .+ is enough, which would mean that complex pattern ends with colon, space and you have something after that. No real need to end the pattern with $ since it already will go up to the end of a string.

Said that... if your just hit Enter inside the pattern, you will get a string like:

  local pattern="^(feat|fix|docs|style|refactor|test|chore)\([a-zA-Z0-9._-]+\): [^
]+$"

But bush (or at least some versions of it) will see it as a single string with a \n and do a real "no \n inside the string".

It is possible - yes. Is it better than simple .+ - definitely not. But as a funny feature and possible way to confuse people - yes.

answered Dec 18, 2024 at 19:11

White Owl

1,4977 silver badges20 bronze badges

1 Comment

user1934428 Dec 19, 2024 at 11:20

Perhaps worth noting: \n is not only not special to bash, but does not have a specific meaning in POSIX extended regex in general.

Collectives™ on Stack Overflow

RegExp testing fails when string is validated in RegExp tools

2 Answers 2

4 Comments

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related