php.net |  support |  documentation |  report a bug |  advanced search |  search howto |  statistics |  random bug |  login
Request #43304 Casting During Comparison
Submitted: 2007-11-15 16:27 UTC Modified: -
Votes:19
Avg. Score:4.5 ± 0.7
Reproduced:17 of 17 (100.0%)
Same Version:9 (52.9%)
Same OS:10 (58.8%)
From: ken at smallboxsoftware dot net Assigned:
Status: Open Package: Feature/Change Request
PHP Version: 5.2.5 OS: ALL
Private report: No CVE-ID:
Have you experienced this issue?
Rate the importance of this bug to you:

 [2007-11-15 16:27 UTC] ken at smallboxsoftware dot net
Description:
------------
Casting during comparison should only happen if values of different types are being compared. 

For example: "42" == "+42" should not return true. 

This result is not very intuitive and probably leads to a number of subtle errors in various scripts. 

Reproduce code:
---------------
<?php

echo "Should Return False:".("400" == "+400")."<BR>";
echo "Should Return False:".((string) "400" == (string) "+400")."<BR>";
echo "Does Return False:".("400" === "+400")."<BR>";

?>


Patches

Add a Patch

Pull Requests

Add a Pull Request

History

AllCommentsChangesGit/SVN commitsRelated reports
 [2008-09-06 23:49 UTC] drm at melp dot nl
Though documented (http://www.php.net/language.operators, quote "If you compare two numerical strings, they are compared as integers."), this is indeed unexpected behaviour (and thus a bug)

This bug will probably be closed and denoted as "sorry, documented feature", but I think this is really a candidate for revision in any next major PHP version.

Comparison should always feel intuitive, and PHP has made the move towards more "secure" defaults. This falls (imo) under the same category. Please witness following example.

client side code:
<form ...>
   <input type="radio" value="0" name="selection" /> No selection
   <input type="radio" value="whatever" name="selection" /> whatever selection
   (...)
</form>

server side code:
if ( $_GET['selection'] == 0 ) {
   echo '<p>Please make a selection!</p>';
} else {
   // process request
}

Though the example does not imply any security issues, they are not unimaginable. Please reconsider the behaviour of the comparison operator.

My suggestion:
Implement (and document!) the following:
Any left side of the comparison is only cast to number if the right side doesn't fail to. Casting to number is considered successful, if and only if the string would syntactically be a number in the PHP
Therefore, the string should match the following:

http://php.net/language.types.integer 
http://php.net/language.types.float

[At both pages, see the "formal" definition of both types.]

In any OTHER case, the comparison should act as follows:
If ANY of the sides is of a string type (and of course fails to cast to number), the other side is cast to string, causing the types to follow these rules:

integer: the string representation of the integer (e.g. "0", "-4" or "123")
float: the string representation of the floating point number, represented in one and only one particular format, preferably just the %.f format.
boolean: cast to integer, cast to string (so true becomes 1, becomes "1"; false becomes 0 becomes "0")
array: "array" causing a notice
object: "object" causing a notice, or calling the __toString method if defined.

The oppposing problem is a non-generic way of handling any other comparison. As a rule, we could state that any other comparison -if both sides of different type- is handled the way it is in the current PHP version. If they are of the same type, the == operator should act the same as the === operator.

This is, as i am very sure, easily realized. 

Please do not hesitate to ask me any questions through mail.
 [2011-02-17 08:58 UTC] kissifrot at gmail dot com
I agree with that, for example if I have
$ric1 = '00001000010000000198358';
$ric2 = '00001000010000000198455';
var_dump($ric1 == $ric2)

should return false, not true, just because it's logical.
But the current behavior makes PHP somewhat unreliable :(
 [2011-03-11 09:43 UTC] delphists at apollo dot lv
Sorry for digging up this bug report, but it's still opened, so I don't think that digging it up should be a problem.

I agree with kissifrot about specific case - both values convert to "(int)2147483647", at least on 32bit system, which is nowhere near to "00001000010000000198358".

Though, I can't agree with drm. There is a reason for having two types of comparisons - if you are sure that you don't need values to be converted, just use strict comparison.

I agree that there are cases when numerical strings need to be compared as strings, but there are also lots of cases when numerical strings have to be compared as integers. For example, some functions that return data from DB, return integers as strings, which may be compared to data from browser, which by default is also string, but comparison needs to work like if they were numbers.
 [2012-01-16 14:29 UTC] michael dot kluge at wundermedia dot de
No comments from any PHP devs yet?

Well I strongly disagree with the current implementation.

From PHP-Docs:
$a === $b Identical TRUE if $a is equal to $b, and they are of the same type.

This is the correct definition of the === operator.
An so it should work like that definition:

So
$a = "001234";
$b = "1234";
$a == $b should be equal to $a === $b (both should be false)
as both variables hold the same type
and from the above definition
those expressions should be equal:
$a === $b
($a == $b && typeof($a)==typeof($b))

So if (typeof($a)==typeof($b)) is true, $a===$b should be identical to $a==$b.

If comparing two variables of the same type in NO case any type-conversion should be applied!

If you need conversions in those cases these should be explicitly noted as e.g.:
intval($a) == intval($b)

Adding special exceptions to the general definition doesn't make much sense in my eyes; it just makes the base definition wrong in certain circumstances and leads to unexpected behaviour in certain cases (see comment from kissifrot).

We stumbled upon this when comparing certain product ids in use by one of our customers, just being different in the leading zeros. (So id "001" is another product than "01" ==> therefore using strings; please don't ask why, as I don't know what reasons might have caused these strange ids)

delphists said that "there are also lots of cases when numerical strings have to be compared as integers". I agree with that, but in those cases the developer should convert the types manually.
Also the example with the database should, in my opinion, be implemented differently:
The model should return integers for integer-typed database fields and not strings.
On the other hand if you expect input parameters containing integer values these could also be converted to integers, which can be easily done in the validity checks of the parameters.

IMHO, despite of PHP trying to make type-conversions transparent to the developer, the developer should in any case keep in mind, that PHP is not typeless and respect the types accordingly. And "0123" and "123" are definitely NOT logically equal in any case! I think the weird concept of "numerical strings" should be dropped.
 [2012-01-16 15:53 UTC] delphists at apollo dot lv
From docs:
[quote]
If you compare a number with a string or the comparison involves numerical strings, then each string is converted to a number and the comparison performed numerically.
[/quote]

Want to be strict, when comparing? Then use strict operators...
 [2012-01-16 16:49 UTC] michael dot kluge at wundermedia dot de
Well, this is what we do.
But this is in my opinion nothing but a workaround for that.

I would regard the current behaviour as a bug or, sadly, rather as an example of bad design in any language. It's just not logical.

Javascript e.g. implements this correctly (from Firebug; Mozilla JS-implementation):
>>> var x1='001',x2='01';
undefined
>>> x1==x2
false
>>> x1===x2
false
>>> x2=1
1
>>> x1==x2
true
>>> x1===x2
false

Therefore I added my last comment to (re-)start a discussion on that, hoping that this might be changed or at least be made configurable. 
At least this request is marked as "Feature/Change Request". ;-)

My idea is to change this:
- Drop the concept of "numerical strings"
- Equal types do never convert on any comparison

First this should be made configurable (default is to stop the weird behaviour), to give grace time to the developers to adapt their scripts
In later versions this should be removed completely.
Maybe a similiar way the magic-quotes-crap is going...

Maybe one more thing about the docs:
It states:
"If you compare a number with a string or the comparison involves numerical strings, then each string is converted to a number and the comparison performed numerically."
This also constricts the simple declarations of the operators in the case of both operands being "numerical strings".
This is exactly the behaviour being very bad in my eyes, and should be changed for creating unexpected results in certain circumstances.
I see no valid reason for the operators to change behaviour if both operands are of the same type.
(Except maybe lazy devs not willing to obey types) ;-)

Maybe I've overseen something... Does anybody know of any good reason why the actual behaviour might be regarded better than the one proposed by others and me?
 [2015-01-09 00:58 UTC] ajf@php.net
By the way: the behaviour kissifrot mentions was actually fixed a while ago. I'm not sure in which version, I think it was 5.4 or 5.5.
 
PHP Copyright © 2001-2017 The PHP Group
All rights reserved.
Last updated: Thu Aug 17 11:01:43 2017 UTC