rfc:inconsistent-behaviors

PHP RFC: Inconsistent behaviors to discuss/document

This RFC is to discuss comparison and conversion inconsistencies in PHP.

Introduction

There are number of in comparison and conversion inconsistencies.

For example,

There are number issues of like this.

Purpose of this RFC is fix inconsistency where it's feasible, otherwise document then fully if it's not documented already.

Inconsistency

Conversion/comparison

Type juggling only works for INTEGER or HEX like strings.

Most problematic is HEX like strings being auto-coerced during comparison, but using different rules from manual casting. That is, ( 0x0A == “0x0A” ) is not treated as ( 0x0A == (int)“0x0A” ), although “0x0A” is translated to a number.

This despite http://us2.php.net/manual/en/language.operators.comparison.php, which states clearly that for number-string comparison, we “Translate strings and resources to numbers.” While it is feasible that some string patterns cannot be “translated” (OCTAL and BINARY) at all, once a “translation” is attempted, it should follow the same rules as (int) casting for the same string. It is hard to view it is anything but a bug that it does not.

HEX

Code

var_dump(0x0A);
var_dump("0x0A");
var_dump((int)"0x0A");
var_dump((float)"0x0A");
var_dump(intval("0x0A"));
var_dump(floatval("0x0A"));

Output

int(10)
string(4) "0x0A"
int(0)
float(0)
int(0)
float(0)

Code

if (0x0A == '0x0A') {
  echo "0x0A == '0x0A'".PHP_EOL;
}
if (0x0A == "0x0A") {
  echo '0x0A == "0x0A"'.PHP_EOL;
}

Output

0x0A == '0x0A'
0x0A == "0x0A"

Octal

Code

var_dump(010);
var_dump("010");
var_dump((int)"010");
var_dump((float)"010");
var_dump(intval("010"));
var_dump(floatval("010"));

Output

int(8)
string(3) "010"
int(10)
float(10)
int(10)
float(10)

CODE

if (010 == '010') {
  echo "010 == '010'".PHP_EOL;
}
if (010 == "010") {
  echo '010 == "010"'.PHP_EOL;
}

OUTPUT

(NONE)

BINARY

Code

var_dump(0b110);
var_dump("0b110");
var_dump((int)"0b110");
var_dump((float)"0b110");
var_dump(intval("0b110"));
var_dump(floatval("0b110"));

Output

int(6)
string(5) "0b110"
int(0)
float(0)
int(0)
float(0)

CODE

if (0b010 == '0b010') {
  echo "0b010 == '0b010'".PHP_EOL;
}
if (0b010 == "0b010") {
  echo '010 == "010"'.PHP_EOL;
}

OUTPUT

(NONE)

Array of Chars

Null string is not handled as ARRAY.

https://github.com/php/php-src/pull/463

Test script:

$a = ''; // empty string
$a[10] = 'a';
echo $a; // "Array"

$b = ' '; // non empty string
$b[10] = 'b';
echo $b; // "          b"

Expected result:

"          a"
"          b"

Actual result:

"Array"
"          b"

String Integer conversion

PHP converts “integer like string to integer”.

<?php

// this is the problem, which we'd expect
// to return false, but which returns true:
echo (2 == '2b').'<br />';

// this is probably what's happening:
echo (2 == intval('2b')).'<br />';

// this is what probably should happen:
echo (strval(2) != '2b').'<br />';

?>

https://bugs.php.net/bug.php?id=66211

Not only is this not a bug, it isn't even exceptional behavior on the modern web. Users who find this behavior surprising are likely inexperienced with MySQL -- clearly PHP's partner in server-side ubiquity as part of the dominant *AMP stack -- which has the exact same rules for auto-coercion of “numeroalphabetic” strings in a comparison context.

In MySQL (all supported versions):

SELECT CASE WHEN '-45herearesomeletters' = -45 THEN 'true' ELSE 'false' END

prints 'true'

There are other popular languages that follow the same casting/coercion rule, though they do not automatically perform the coercion during comparison. For example, JavaScript parseInt('-45herearesomeletters') results in the integer -45. In SQLite, the ubiquitous embedded SQL database, also CAST( '-45herearesomeletters' AS SIGNED ) produces the integer -45.

The SQLite documentation explains the logic well:

When casting a TEXT value to INTEGER, the longest possible prefix of
the value that can be interpreted as an integer number is extracted
from the TEXT value and the remainder ignored. Any leading spaces in
the TEXT value when converting from TEXT to INTEGER are ignored. If
there is no prefix that can be interpreted as an integer number, the
result of the conversion is 0. (http://www.sqlite.org/lang_expr.html#castexpr)

And this behavior is not considered particularly “distinctive”. (http://sqlite.org/different.html)

Since the ubiquity of MySQL has been used to support the expectations users should have of PHP, it's fair to note Oracle, SQL Server, and PostgreSQL will not allow the above comparison to be performed: the statement produces a fatal error. It's a runtime casting error: these languages do not prohibit comparing values of different datatypes, as long as the engine can cast the runtime contents of the value. Yet such implementations, arguably, violate the “least astonishment” concept, since a errant letter modifier like '1A' will cause a fatal error where the expectation might be to either have a '1A' compare equal to 1 (as in MySQL) or fail gracefully (as in SQLite). In this respect, the SQLite behavior is more balanced than that of Oracle/MSSQL/PGSQL, and PHP and MySQL's behavior is graceful, generous, and reasonable.

With PHP and MySQL agreeing on this behavior, it is clear that automatically coercing a “numeroalphabetic” string (for want of a better term) to a number via truncation is common practice on the web, even if it is news to the inexperienced user.

String decrements

String decrements is inconsistent

https://wiki.php.net/rfc/alpanumeric_decrement

NAN/INF of float

NAN/INF issue.

$f = NAN;
var_dump(++$f);                 // float NAN
var_dump((float) NAN);   // float NAN
var_dump((int) NAN);       // int -2147483648 -> what?
var_dump((bool) NAN);   // bool true -> makes sense

$f = INF;
var_dump(++$f);                         // float INF
var_dump((float) INF);             // float INF
var_dump((int) INF);                 // int 0 -> what?
var_dump((bool) INF);             // bool true -> so why int 0?
var_dump((int) (bool) INF);   // int 1

E_WARNING for these invalid/unreliable operations might be better.

This could be mitigated by GMP float support.

Object Array conversion of numeric property/index

Object/Array cast looses accessibility of numeric property/element. https://bugs.php.net/bug.php?id=66173

$ php -v
PHP 5.5.7 (cli) (built: Dec 11 2013 07:51:06) 
Copyright (c) 1997-2013 The PHP Group
Zend Engine v2.5.0, Copyright (c) 1998-2013 Zend Technologies

$ php -r '$obj = new StdClass; $obj->{12} = 234; ${1} = 567; var_dump($obj, ${1}); $ary = (array)$obj; var_dump($ary, $ary[12]);'
object(stdClass)#1 (1) {
  ["12"]=>
  int(234)
}
int(567)
Notice: Undefined offset: 12 in Command line code on line 1
array(1) {
  ["12"]=>
  int(234)
}
NULL <= SHOULD BE int(234)

Function/Method

assert

assert() does not accept closure while it accepts functions.

php > function f() {return FALSE;}
php > assert(f());

Warning: assert(): Assertion failed in php shell code on line 1
php > assert(function() {return FALSE;});

https://wiki.php.net/rfc/expectations

base_convert

filter_var

https://bugs.php.net/bug.php?id=66682

var_dump(filter_var('01', FILTER_VALIDATE_INT));
var_dump(filter_var('01', FILTER_VALIDATE_FLOAT));
bool(false)
double(1)

is_numeric

min

https://bugs.php.net/bug.php?id=53104

This is not a bug. If one of operand is BOOL(or NULL), both operands are converted to BOOL and evaluated as BOOL. It may be good idea that document this behavior in min() manual.

Status Documented.

http://jp2.php.net/min

Return value of wrong internal function/method parameters

If not all, almost all functions return NULL when required function parameter is missing or wrong type. However, almost all functions return FALSE when they have errors.

The manual has document for this behavior http://www.php.net/manual/en/functions.internal.php

Note: If the parameters given to a function are not what it expects, such as passing an array 
where a string is expected, the return value of the function is undefined. In this case it will 
likely return NULL but this is just a convention, and cannot be relied upon.

This behavior could be cause of bug in scripts. For instance,

if (FALSE === some_func($wrong_parameter)) {
   // Error happend!
} else {
   // OK to go
}

Users should not rely on return value as it may return NULL for wrong parameters. Users should rely on error/exception handler for such case as internal functions raise E_WARNING in this case. (If there are function that does not raise error, it is subject to be fixed.)

It may be good to add use of error/exception handler as best practice in the manual. http://www.php.net/manual/en/functions.internal.php

There are bug reports that complain return value inconsistency. The document could be improved with more explanations.

Related Bug Reports

Bug reports are not verified carefully. Removing wrong one, adding proper one is appreciated.

Developer Guideline

  • Internal function/method should raise error(or exception) for invalid parameters. (parse parameters function does this)
  • Internal function/method is better to return NULL for invalid parameters as most functions do.
  • Internal function/method is better to return FALSE for other errors.

User Guideline

  • User should not rely return value only for failure condition, but should rely error/exception handler for failure also.

Proposal

Not yet.

Backward Incompatible Changes

Not yet.

Proposed PHP Version(s)

PHP 6.0 probably.

Impact to Existing Extensions

Not yet.

New Constants

Not yet.

php.ini Defaults

If there are any php.ini settings then list:

  • hardcoded default values
  • php.ini-development values
  • php.ini-production values

Not yet.

Open Issues

Make sure there are no open issues when the vote starts!

Unaffected PHP Functionality

List existing areas/features of PHP that will not be changed by the RFC.

This helps avoid any ambiguity, shows that you have thought deeply about the RFC's impact, and helps reduces mail list noise.

Future Scope

This sections details areas where the feature might be improved in future, but that are not currently proposed in this RFC.

Proposed Voting Choices

Include these so readers know where you are heading and can discuss the proposed voting options.

Patches and Tests

Links to any external patches and tests go here.

If there is no patch, make it clear who will create a patch, or whether a volunteer to help with implementation is needed.

Make it clear if the patch is intended to be the final patch, or is just a prototype.

Implementation

After the project is implemented, this section should contain

  1. the version(s) it was merged to
  2. a link to the git commit(s)
  3. a link to the PHP manual entry for the feature

References

Links to external references, discussions or RFCs

Rejected Features

Keep this updated with features that were discussed on the mail lists.

ChangeLog

  • 2014/02/05 - Renamed to “inconsistent-behaviors”
  • 2013/10/31 - Initial version.
rfc/inconsistent-behaviors.txt · Last modified: 2021/03/27 14:31 by ilutov