Fix #133 regex dot in character class

[krsriq]

What is the reason for this PR?

regexify produces a random character instead of a literal dot when called within a character class (e.g. [.]).

• Fixed an issue (resolve Regexify dot issue #133)

Author's checklist

• Follow the Contribution Guide
• New features and changes are documented

Summary of changes

Fixes the issue, adds test to verify.

Review checklist

• All checks have passed
• Changes are approved by maintainer

[IonBazan]

Looks good! I tested with random seed and complex regex dataset and it works perfectly!

[IonBazan]

Maybe it's worth to add a test asserting that calling regexify('[.]') will always return "."?

[krsriq]

Yeah good idea, added it to the already existing test cases here:
https://github.com/FakerPHP/Faker/pull/135/files#diff-c648f7b997049e16add4fc6034d0c6f41f5c6b0c6a52a14b7fe4dd797a2a873cR313

[pimjansen]

But a literal dot in regex always need an escape in the input? With this change we actually limit ourself in normal regex use with a catchall dot itself?

[krsriq]

But a literal dot in regex always need an escape in the input? With this change we actually limit ourself in normal regex use with a catchall dot itself?

In a bracketed character class a dot does not have to be escaped to be interpreted as a literal dot: https://perldoc.perl.org/perlrecharclass#Bracketed-Character-Classes

The problem here was, that the character class replacement logic did not take that into account. When called with [.] the character class character selector returns .

Faker/src/Faker/Provider/Base.php

Lines 496 to 499 in afa2aab

	// All [ABC] become B (or A or C)
	$regex = preg_replace_callback('/\[([^\]]+)\]/', function ($matches) {
	return Base::randomElement(str_split($matches[1]));
	}, $regex);

which is then erroneously replaced with a random ascii character here

Faker/src/Faker/Provider/Base.php

Line 503 in afa2aab

$regex = preg_replace_callback('/(?<!\\\)\./', 'static::randomAscii', $regex);

When the character class character selector returns \. instead of . (as is done with this PR), the dot is not replaced by a random ascii character. The additional escape \ is then removed here:

Faker/src/Faker/Provider/Base.php

Line 505 in afa2aab

$regex = str_replace('\\', '', $regex);

with the expected end result of a literal dot.

To verify that calling regexify with a normal dot works as expected, I've added ['.', 'catch-all dot'] to the testRegexifySupportedRegexSyntax dataProvider.

Btw thinking about this I realized that we probably have other related issues here, for example when regexify is called with /[*]/..

[pimjansen]

@krsriq shall we merge this one and raise a new issue for the other issue?

This way we already close this bug right away

[krsriq]

@pimjansen - yes, good idea!

[krsriq]

Thanks!