Everything

php

Seeking 'coalesce' php internal function

php.internals

D. Dante Lorenso

20 years ago

All,

I'm sure this has been asked somewhere, but since I see requests for
features for 5.2 or 6.0, I'd like to add a "simple" item to the list
which would be quite useful to me and would simplify and clean up a lot
of code out there:

function coalesce(...)

This works much like in the SQL version of the same.  In SQL, the
function returns the first non-null argument from an arbitrary list.  In
our use, it should return the first non-empty value from the list:

Example:

$x = "dante";
$y = "";
$z = "";

$value = coalesce($z, $y, $x); // $value = "dante"

This function would ideally be built into the language and bypass
warnings about undefined variables like 'empty' does.  It might be nice
to have several varieties of this function:

* return first parameter where empty() is FALSE
* return first parameter where isset() is TRUE

I don't think something like this can NOT be written in userspace
because the 'isset' and 'empty' checks need to be run before arguments
can be passed to a user function or warnings will start flying.  A
function like this simplifies code which used to look like this:

if (!empty($_POST["postkey"])) {
$value = $_POST["postkey"];
}
elseif (!empty($_GET["getkey"])) {
$value = $_POST["getkey"];
}
elseif (!empty($default_value)) {
$value = $default_value;
}
else {
$value = "hard coded value";
}

Into this:

$value = coalesce($_POST["postkey"], $_GET["getkey"],
$default_value, "hard coded value");

Can this be built and included?

Dante

D. Dante Lorenso

20 years ago

D. Dante Lorenso wrote:

> I don't think something like this can NOT be written in userspace
> because the 'isset' and 'empty' checks need to be run before arguments
> can be passed to a user function or warnings will start flying.  A
> function like this simplifies code which used to look like this:
>
>    if (!empty($_POST["postkey"])) {
>        $value = $_POST["postkey"];
>    }
>    elseif (!empty($_GET["getkey"])) {
>        $value = $_POST["getkey"];
>    }
>    elseif (!empty($default_value)) {
>        $value = $default_value;
>    }
>    else {
>        $value = "hard coded value";
>    }
>
> Into this:
>
>    $value = coalesce($_POST["postkey"], $_GET["getkey"],
> $default_value, "hard coded value");

I mean "I DON'T think like this CAN be written in userspace".  And, yes,
there is a bug in that code up above (around $_GET testing), but the
'coalesce' function removes the test/set with identical values so
removes those kinds of careless bugs.

Dante

Steph

20 years ago

D.

Please read the mailing list archives. And don't mention it again ;)

- Steph


----- Original Message -----
From: "D. Dante Lorenso" <dante@vocalspace.com>
To: <dante@lorenso.com>
Cc: <internals@lists.php.net>
Sent: Wednesday, May 03, 2006 8:05 AM
Subject: Re: [PHP-DEV] Seeking 'coalesce' php internal function

Johannes Schlueter

20 years ago

Hi,

please search the archives for "ifsetor".

johannes

On Wednesday 03 May 2006 07:56, D. Dante Lorenso wrote:

D. Dante Lorenso

20 years ago

Johannes Schlueter wrote:

> please search the archives for "ifsetor".
>

I have completed this search and find:

http://marc.theaimsgroup.com/?r=1&w=2&q=b&l=php-dev&s=coalesce
http://marc.theaimsgroup.com/?l=php-dev&w=2&r=1&s=ifsetor&q=b

I am using PHP 5.1.2 currently and thought using pass-by-reference was
frowned upon.  Assuming a mistaken assumption I move forward and see the
following code which was presented to the list:

<?php
function ifsetor(&$var, $value) {
return (isset($var)) ? $var : $value;
}
print ifsetor($x, "dante");
?>

This works as I'd like in my dev environment, however, this does not
solve the problem of allowing variable number of parameters (variables
and values).  So, I've whipped up this 10 variable version:

<?php
function ifsetor(&$var1, &$var2, &$var3, &$var4, &$var5, &$var6, &$var7,
&$var8, &$var9, &$var10) {
if (isset($var1)) return $var1;
if (isset($var2)) return $var2;
if (isset($var3)) return $var3;
if (isset($var4)) return $var4;
if (isset($var5)) return $var5;
if (isset($var6)) return $var6;
if (isset($var7)) return $var7;
if (isset($var8)) return $var8;
if (isset($var9)) return $var9;
if (isset($var10)) return $var10;
return false;
}
print ifsetor($x["notset"]->notset, $y["notset"]->notset, $z->notset,
"dante");
?>

PHP Fatal error:  Only variables can be passed by reference in
.../ifsetor.php on line 15
PHP Warning:  Missing argument 5 for ifsetor(), called in
.../ifsetor.php on line 15 and defined in .../ifsetor.php on line 2
PHP Warning:  Missing argument 6 for ifsetor(), called in
.../ifsetor.php on line 15 and defined in .../ifsetor.php on line 2
PHP Warning:  Missing argument 7 for ifsetor(), called in
.../ifsetor.php on line 15 and defined in .../ifsetor.php on line 2
PHP Warning:  Missing argument 8 for ifsetor(), called in
.../ifsetor.php on line 15 and defined in .../ifsetor.php on line 2
PHP Warning:  Missing argument 9 for ifsetor(), called in
.../ifsetor.php on line 15 and defined in .../ifsetor.php on line 2
PHP Warning:  Missing argument 10 for ifsetor(), called in
.../ifsetor.php on line 15 and defined in .../ifsetor.php on line 2

I can work around passing values by reference by calling the function
like this:

print ifsetor($x["notset"]->notset, $y["notset"]->notset,
$z->notset, $temp = "dante");

But you have to admit that's getting ugly and still, there are undefined
parameters declared ... so then I tried it by assigning default values:

<?php
function ifsetor(&$var1=null, &$var2=null, &$var3=null, &$var4=null,
&$var5=null, &$var6=null, &$var7=null, &$var8=null, &$var9=null,
&$var10=null) {
if (isset($var1)) return $var1;
if (isset($var2)) return $var2;
if (isset($var3)) return $var3;
if (isset($var4)) return $var4;
if (isset($var5)) return $var5;
if (isset($var6)) return $var6;
if (isset($var7)) return $var7;
if (isset($var8)) return $var8;
if (isset($var9)) return $var9;
if (isset($var10)) return $var10;
return false;
}
print ifsetor($x["notset"]->notset, $y["notset"]->notset, $z->notset,
$temp = "dante")."\n";
print_r($x);
?>

Now, $x, $y, and $z ARE set ...

dante
Array ( [notset] => stdClass Object ( [notset] => ) )

This is a nasty side-effect and can not be accepted.  I did not want to
set these variables, just check their existance.

I am willing to go away and concede that this should be done in
userspace IF you can show me that it is in fact possible.  The 2
argument ifsetor is trivial in userspace, yes, but the variable case is
more useful in a variety of situations and provides the most benefit to
simplify code.  I did not find the answer in the archives.  Can you
point me in the right direction?  I need to be able to do all of the
following:

* variable number of parameters
* test 'isset' or 'empty'
* testing variables and non-variable values (pass-by-reference won't
work on values)
* not have side-effect of defining values which are not already set
* not trigger notices or warnings
* returns value of first proper match

Dante

Eric Coleman

20 years ago

http://us3.php.net/func_get_args
http://us3.php.net/array

Enjoy.

Eric Coleman

Eric Coleman

http://aplosmedia.com
home: 412 399 1024
cell: 412 779 5176

On May 3, 2006, at 3:13 AM, D. Dante Lorenso wrote:

D. Dante Lorenso

20 years ago

Eric,

This reply is too basic and is not the answer.  The problem is more
complex then you have grasped.

The only way to remove the notice and warning errors is by using
pass-by-reference in the userspace function.  Pass-by-reference can not
be done for literal values and will only work on variables.  All notices
and warnings have been thrown long before func_get_args can be run.
Besides, func_get_args will return an array and then isset() and empty()
no longer apply to the original variables.

Dante

Eric Coleman wrote:

Rick Widmer

20 years ago

D. Dante Lorenso wrote:

> Eric,
>
> This reply is too basic and is not the answer.  The problem is more
> complex then you have grasped.

function ifsetor()  {

$args = func_get_args();
$count = count( $args );

for( $i=0; $i<$count; $i++ ) {
if isset( $args[ $i ] )) {
return $args[ $i ];
}

return false;
}

D. Dante Lorenso

20 years ago

Rick Widmer wrote:

> D. Dante Lorenso wrote:
>> Eric,
>> This reply is too basic and is not the answer.  The problem is more
>> complex then you have grasped.
> function ifsetor()  {
> $args = func_get_args();
> $count = count( $args );
>
> for( $i=0; $i<$count; $i++ ) {
>    if isset( $args[ $i ] )) {
>       return $args[ $i ];
>    }
>
> return false;
> }
>

No, that doesn't address the problem.  See this:

---------- 8< -------------------- 8< -------------------- 8< ----------
<pre>
<?php
function ifsetor()  {
$args = func_get_args();
$count = count($args);
for ($i=0; $i<$count; $i++) {
if (isset($args[$i])) {
return $args[$i];
}
}
return false;
}
print ifsetor($x, $y, $z, "dante")."\n";
?>
</pre>

PHP Notice:  Undefined variable: x in .../ifsetor.php on line 13
PHP Notice:  Undefined variable: y in .../ifsetor.php on line 13
PHP Notice:  Undefined variable: z in .../ifsetor.php on line 13
---------- 8< -------------------- 8< -------------------- 8< ----------

See my post regarding a specification for 'filled()' as a function.  An
important feature is that warnings about unset variables should not
occur.  Your suggestion does not qualify.  This problem can not be
solved in userspace.

Dante

Rick Widmer

20 years ago

D. Dante Lorenso wrote:

> No, that doesn't address the problem.  See this:

print @ifsetor($x, $y, $z, "dante")."\n";

Jochem Maas

20 years ago

Rick Widmer wrote:

> D. Dante Lorenso wrote:
>
>> No, that doesn't address the problem.  See this:
>
>
> print @ifsetor($x, $y, $z, "dante")."\n";

point 1: regardless of how people think there should be an ifsetor()
in the engine - the devs don't so there wont be one.

point 2: anything trying to implement this functionality in userspace
is a hack. (to stregethen that stance: Rasmus reiterated the point a few
days ago [can't remember which mailinglist] that people should be trying to
write *error free* code - ergo the error suppression is not a proper solution)

Dante, live with the hack. :-)

D. Dante Lorenso

20 years ago

Jochem Maas wrote:

> point 1: regardless of how people think there should be an ifsetor()
> in the engine - the devs don't so there wont be one.

Arrrrvvhggghh!  Don't you just want to scream!

> point 2: anything trying to implement this functionality in userspace
> is a hack. (to stregethen that stance: Rasmus reiterated the point a few
> days ago [can't remember which mailinglist] that people should be
> trying to
> write *error free* code - ergo the error suppression is not a proper
> solution)

Ok.  You CAN NOT implement 'ifsetor/coalesce/filled' in userspace.  And
the devs don't want it to exist ... so, WTF?

> Dante, live with the hack. :-)

Translate to: Dante, you don't need PHP to be better, adequate is good
enough.

<soap box>
Come on devs, I'm pleading with you.  Say it isn't so.  You force me to
write so much extra code unless this functionality exists in the core.

Explain to me how every possible thing you might want to do to an array
gets it's own function in the language, but something as trivial as this
gets refused so strongly!

I mean really: array_change_key_case, array_chunk, array_combine,
array_count_values, array_diff_assoc, array_diff_key, array_diff_uassoc, ...
http://www.php.net/manual/en/ref.array.php

You can't tell me that most if not all of these functions could not have
been implemented in userspace?  I might use 10% of all the array
functions, but I'd use 'coalesce' and 'filled' daily.
</soap box>

Tell me at least this much.  Can these two functions be implemented as
an extension? From what you know about PHP internals, does what WE want
have to be a language construct?  If I can write coalesce() and filled()
as extensions, I'll submit a new extension/patch.  Maybe it could join
PECL and with popularity make it's way begrudgingly into the core?

Dante

D. Dante Lorenso

20 years ago

Rick Widmer wrote:

> D. Dante Lorenso wrote:
>> No, that doesn't address the problem.  See this:
> print @ifsetor($x, $y, $z, "dante")."\n";

Usage of the shutup operator, @, in this context is a shameful coding
habit and hides too much:

print @ifsetor($x, fakefunction($y), $z, "dante")."\n";

Does not show any warnings or errors, but REALLY hides this:

PHP Notice:  Undefined variable: x in .../ifsetor.php on line 13
PHP Fatal error:  Call to undefined function fakefunction() in
.../ifsetor.php on line 13

Ignoring the notice on $x is fine, but that fatal error shouldn't be
ignored.

Dante

D. Dante Lorenso

20 years ago

Dear Internals,

I'd like a white bikeshed, but until you can decide on a color, how
about we just build the bikeshed already.  Please consider the following
specification:

========== 8< ==================== 8< ==================== 8< ==========

filled

(PHP 5.2.0)

filled - Find first non-empty value in a variable list where empty()
evaluates as FALSE.

Description
------------------------------

mixed filled ( mixed varname [, mixed ...] )

filled() takes a variable number of parameters.  For each of these,
*filled()* looks for a variable where !empty( varname ) evaluates as
TRUE.  It returns the first non-empty value or NULL if all values are
empty().

Note: filled() only checks variables as anything else will result in a
parse error.  In other words, the following will not work:
filled(trim($name)).
filled() is the opposite of empty(), and no warning is generated
when any one of the variables is not set.

Return Values
------------------------------

Returns first non-empty() value.

If all values are empty(), returns NULL.

Example
------------------------------
<?php
echo filled($x["somekey"], $_GET["getkey"], $default, "example");
echo filled("yes"); // prints yes
filled(false); // returns NULL
echo filled($x["apple"], $y, "banana"); // prints banana
$y = "cat";
echo filled($x["apple"], $y, "banana"); // prints cat
$x["apple"] = "pear";
echo filled($x["apple"], $y, "banana"); // prints pear
unset($x["apple"]);
echo filled($x["apple"], $y, "banana"); // prints cat
?>

See also empty()

========== 8< ==================== 8< ==================== 8< ==========

Dante

D. Dante Lorenso

20 years ago

The default value for 'filled()' if all values test TRUE for empty()
should be FALSE, not NULL.  This is more consistent with the return
values of empty().

Since filled() returns the first non-empty() value in the list of
parameters provided, it works well when cast into a boolean context
because the non-empty() value should evaluate as TRUE.  Example:

filled("testing") // returns testing
if (filled("testing")) {} // true
filled(87) // returns 87
(boolean) filled(87) // true

$x["apple"] = "42";
filled($x["apple"]) // returns 42
filled($x["pear"]) // returns false
if (filled($x["apple"]) !== false) {} // true
if (filled($x["pear"]) !== false) {} // false

Just wanted to chime in with an update in case 'filled()' can actually
be considered.  I'd like everyone to consider this one function to be
evaluated on it's merits alone and apart from any other functions which
also might be wanted.  filled() is the opposite of empty() and is not
dealing with tests for 'isset()'.  Another function should be proposed
to solve that problem if it would still be needed after filled() is
implemented.

Do I have any power to call a vote on this?

Dante


D. Dante Lorenso wrote: