Function to validate regular expression

[timint]

Description

I have been missing a core function in PHP to validate a regular expression. I am therefore proposing something like preg_validate().

There are some workarounds mentioned in this StackOverflow thread. But from my understanding they will all produce a PHP warning that needs to be suppressed.
https://stackoverflow.com/questions/4440626/how-can-i-validate-regex

What are your thoughts and opinions on implementing a regex syntax validator?

[cmb69]

That had already been requested in the old bug tracker; I'm closing that as duplicate.

[timint]

pcre_get_compiled_regex_cache_ex() indeed looks like most of the work done already. 👍

[KapitanOczywisty]

@timint Lets say it would be: function preg_validate(string $pattern): bool, I expect that no warning should be emitted, so should preg_validate return somehow what error caused validation to fail? I thought that preg_last_error_msg can be used, but there is only Internal error returned. Errors like: No ending delimiter '/' found are reported as E_WARNING (php_error_docref(NULL,E_WARNING, "No ending delimiter '%c' found", delimiter);).

Bad options:
a) do not return errors/warnings at all, only validate
b) emit warning like in preg_match
c) return error as reference (references are bad)
d) change preg_last_error_msg to return detailed errors (BC)

Less bad options:
e) add preg_validate_error_msg to return specific error
f) return error message as string => function preg_validate(string $pattern): string|true
g) something else ??

WIP patch (without error handling solved): https://gist.github.com/KapitanOczywisty/5f82d8f033be1deebd9f916a7d3a80e3

[damianwadley]

Like was mentioned with the recent idea of validating JSON strings, does this need to be another function or should it be a new capability with filter_var?

[KapitanOczywisty]

This is also a good idea, however to have complete solution there should be a way to get error message.
FILTER_VALIDATE_REGEXP already exists so it would be probably FILTER_VALIDATE_PATTERN / FILTER_VALIDATE_REGEXP_PATTERN.

[timint]

@KapitanOczywisty I agree it would be useful to be able to determine what the actual syntax error is. That can be done by returning false for no error, and return a specific error flag upon error. Or it could populate optional variable references $error_code and $error_message like preg_validate($pattern, &$error_code=null, &$error_message=null) while the function itself just returns true or false. It should not generate a warning as that is what preg_match() already does with an invalid pattern. (Someone is probably gonna say, hey just throw an exception.)

filter_var() by FILTER_VALIDATE_REGEXP_PATTERN is also an option if we don't care what the specific syntax error is. 👍

[KapitanOczywisty]

Or it could populate optional variable references $error_code and $error_message like preg_validate($pattern, &$error_code=null, &$error_message=null)

References are used in some places to pass errors, but I don't think that it is recommended for new functions.

filter_var() by FILTER_VALIDATE_REGEXP_PATTERN is also an option if we don't care what the specific syntax error is. 👍

@timint You proposed that, you should know if we care :)

I'll try to make FILTER_VALIDATE_REGEXP_PATTERN variant then.

[KapitanOczywisty]

POC FILTER_VALIDATE_REGEXP_PATTERN: KapitanOczywisty@3ed8b79

@cmb69 Does this require RFC? And is there better way to suppress errors in pcre_get_compiled_regex_cache_ex?

Also it could be worth to add option for checking number of capture groups min_captures + max_captures or only captures. For the future scope maybe array of required named capture groups.

[cmb69]

Does this require RFC?

I tend to say yes. At the very least, this needs to be discussed on the internals mailing list.

And is there better way to suppress errors in pcre_get_compiled_regex_cache_ex?

I think it is okay as is, but that can be worked out in a PR.

[KapitanOczywisty]

@timint Can you pitch FILTER_VALIDATE_REGEXP_PATTERN on the internals mailing list?