Smaller fix for the ext/filter boolean filter limitation (was Re: cvs: php-src /ext/filter long subject) View as Markdown

Hi,

First my apologies for not being able to go a real reply to the real thread,
but I just subscribed to the list a couple days ago (and the original thread
subject was way too long anyway ;-)).

Boolean values are the only case where it is not possible to
distinguish FALSE from invalid as they can be FALSE, but it is really
not an issue.
If it wasn't an issue I wouldn't have fixed it.

IMHO there's a more simple way to fix that without changing all the functions
behavior. I think the distinction between NULL (non-existent data) and FALSE
(invalid) should be kept and that having both cases return NULL, thus forcing
users to use filter_has_var() to determine what this NULL means is a bit
counter-intuitive.

A quite smaller modification of the boolean filter behavior could allow this
(I'm sorry I can't send you a patch, but I'm a "PHP user" who hasn't touched
a C source file for years... I'll try and do better next time if I can).

Make php_filter_boolean() return:

• (int)1 when the value is 1/on/true/yes
• (int)0 when the value is 0/off/false/no
• (bool)false when the value is something else

This is mostly backward compatible, as PHP considers that 0 == false and 1 ==
true, and would still allow people like Derick who'd need to make a
difference between a "valid" boolean and an invalid value the possibility to
do so by testing the returned value using ===.

Also, it would just require a small change in one single function, which I
think would be more safe considering 5.2.0 is already an RC5.

Hoping I could be helpful,

skalpa.