[9.x] Add decimal validation rule

[RaazPuspa]

This PR adds decimal validation rule to check that an attribute is numeric and consists of decimal part digits between a set of values.

Usage

decimal:min,max: parameter(s) defines minimum and maximum (optional) number of acceptable digits (both inclusive) in decimal part

Examples

• decimal:1,3: requires values to have one, two or three digits in fractional part
• decimal:0,3: requires values to have at most three digits in fractional part
• decimal:3: requires values to have at least three digits in fractional part

[taylorotwell]

tweaked behavior of rule so that decimal:2 means exactly 2 decimal places.

[taylorotwell]

Added translation line and replacements.

[willrowe]

Does this allow negative numbers?

[RaazPuspa]

@willrowe, I tested the case and found that current implementation does not support validation of signed numbers.

I have updated the validation rule logic allowing signed numeric values. Look #45407 for the changes.

[vlakoff]

The dot is unescaped, I think it is unintentional:

preg_match('/^\d*.(\d*)$/', $value, $matches);

(example of value that can erroneously pass currently: 1e12)

However, by just escaping that dot, the case 42 (no decimal part) would not work anymore (it currently works by luck, because the dot makes a backtrack and consumes the last digit).

After some thoughts, I would suggest the following code (includes #45407):

preg_match('/^[+-]?\d*(\.\d*)?$/', $value, $matches);

$decimals = isset($matches[1]) ? (strlen($matches[1]) - 1) : 0;

Also note the following cases are supported by the underlying is_numeric(), and also by this validation rule currently: 42. and .43 (also -42. and -.43). Maybe tests should be added for these.

[vlakoff]

And for consistency with the Numeric rule, maybe also support leading/trailing spaces:

preg_match('/^\s*[+-]?\d*(\.\d*)?\s*$/', $value, $matches);

preg_match('/^[+-]?\d*(\.\d*)?$/', trim($value, " \t\n\r\v"), $matches); // variant

(the null-byte character \x0 is not supported by is_numeric(), hence excluding it from the trim() in the variant)

[vlakoff]

I could open a PR for this, but I would need your stance on the points I exposed above.

• Should .42 and 42. values pass?
• Should leading/trailing spaces pass?

I'd say yes for both. Not because they would arguably be better per se, but for consistency, because is_numeric(), thus the Numeric rule, accept them.