is_numeric() limitation

[8ctopus]

Description

is_numeric() is of great help to check if a string contains a number but in some cases, it creates problems:

is_numeric('2E822890835372110'); // true
is_numeric('64091761E47298917'); // true

If I understand correctly, is_numeric detects the E within the string and considers it as exponent, so in the first case 2 * 10^822890835372110.

In order, to avoid this I would like to suggest the following change to the function:

is_numeric(mixed $value): bool

to

is_numeric(mixed $value, int $flags): bool

$flags being a combination of:

• NUMERIC_EXPONENT
• and maybe some other flags...

[cmb69]

Hmm, there is already FILTER_VALIDATE_FLOAT (and FILTER_VALIDATE_INT) which appears to be more suitable for these kinds of checks (e.g. https://3v4l.org/b67Kl).

Anyhow, this feature would need an RFC.

[8ctopus]

@cmb69 ah didn't know about filter_var because it's not in the is_numeric see also doc. I'll update it.

[cmb69]

Are you still interested in the suggested extension of is_numeric()?

[8ctopus]

@cmb69 You've probably solved my issue, I just need to confirm it works. If that's the case, I'll close this feature request.

[8ctopus]

still have an issue in some cases (see 3rd one).

var_dump(filter_var('2E822890835372110', FILTER_VALIDATE_FLOAT)); // false
var_dump(filter_var('64091761E47298917', FILTER_VALIDATE_FLOAT)); // false
var_dump(filter_var('1337e0', FILTER_VALIDATE_FLOAT)); // float(1337)

what do you think?

[damianwadley]

Is it possible you don't want to validate the string as a float but instead as an integer?

In addition to the INT flag cmb69 mentioned earlier, there's also the ctype_digit function.

[8ctopus]

@damianwadley I would like to have a function that validates '12.92', '11', but not '2e0', so FILTER_VALIDATE_INT won't do it.

ctype_digit has the same issue with floats, because . is not supported.

[cmb69]

Actually, I would go with a regexp, but I think that there should be a way to validate floats without allowing scientic notation. There is already FILTER_FLAG_ALLOW_SCIENTIFIC, but that is so far only supported by FILTER_SANITIZE_NUMBER_FLOAT (which is not useful here, since it just skips the e). It looks reasonable to also support this flag for FILTER_VALIDATE_FLOAT. PoC:

 ext/filter/logical_filters.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ext/filter/logical_filters.c b/ext/filter/logical_filters.c
index 3f58b2a3c4..b578cdd503 100644
--- a/ext/filter/logical_filters.c
+++ b/ext/filter/logical_filters.c
@@ -416,7 +416,7 @@ void php_filter_float(PHP_INPUT_FILTER_PARAM_DECL) /* {{{ */
 					*p++ = *str++;
 				}
 			}
-			if (*str == 'e' || *str == 'E') {
+			if ((flags & FILTER_FLAG_ALLOW_SCIENTIFIC) && (*str == 'e' || *str == 'E')) {
 				*p++ = *str++;
 				if (str < end && (*str == '+' || *str == '-')) {
 					*p++ = *str++;

The "only" problem is that this would be a BC break, and we would need to set the flag by default.