Implement new array function array_column()

[ramsey]

array_column() returns the values of the specified column from a multi-dimensional array.

I've actually had this patch sitting around for several years, and I decided to take a look, clean it up, and finally submit it. Inspired by database methods like PDOStatement::fetchColumn(), I wrote this to scratch an itch, since I've often had the need to take a recordset array and pull a single column of values from it, which normally involves looping over the array and creating a new array of the values for the column I want to pull. I decided to implement this in the PHP core to save that step.

You can take a look at the test included to see how it works. The second parameter is the key for the column to return, which can be a number for numeric indexes or a string for associative arrays. If it can't find the column specified, it simply returns an empty array.

[smalyshev]

Maybe RETURN_FALSE ? Though then I'd move the array init after the check.

[ramsey]

I've updated this to RETURN_FALSE. I've also cleaned it up a bit by removing some of the other cases.

[pierrejoye]

IS_OBJECT should be there too, calling convert_to_string if necessary (will deal with the existence of __toString or not.

[pierrejoye]

convert_to_string_ex afair :)

[ramsey]

I've added IS_OBJECT.

[cataphract]

I'd rather have a more general mechanism for extracting subsets of arrays, like in Mathematica:

In[6]:= {{1, 2}, {3, 4}, {5, 6, 7}}[[All, 2;;]]
Out[6]= {{2}, {4}, {6, 7}}

In[7]:= {{1, 2}, {3, 4}, {5, 6, 7}}[[All, 2]]
Out[7]= {2, 4, 6}

In[8]:= {{1, 2}, {3, 4}, {5, 6, 7}}[[{1,3}, 2]]
Out[8]= {2, 6}

but I guess this is better than nothing.

[ramsey]

@cataphract, can you suggest how that might look like PHP? I don't think my implementation is too far off. With some slight modifications, it might be possible to achieve something similar to what you're asking.

[pierrejoye]

@ramsey

btw, pls put a RFC together and send a notification mail to internals once you are ready, referring to this discussion too to avoid to have to discuss everything again from scratch :)

[cataphract]

@ramsey You could for instance use null mean both All and to specify a range. This can be sorted out later. For instance, for my examples above:

array_part($foo, [null, [2, null]]);
array_part($foo, [null, 2]);
array_part($foo, [[1, 3], 2]);

[smalyshev]

I think we shouldn't try to boil the ocean here. array_column is very useful on its own. If we will need super-method to do everything and make fresh coffee, we could add it later, but that IMHO should not hold adding array_column.

[dsp]

Yes I agree. I think array_column is good enough to be added. I think we are still waiting for an RFC.

[ramsey]

I finally got around to creating an RFC and submitting this to internals:

• https://wiki.php.net/rfc/array_column
• http://news.php.net/php.internals/60930

[ccampbell]

Hey @ramsey, I think think this proposal is a great idea (I do stuff like this all the time), but how do you feel about extending it slightly.

For example if you start with the following array:

$data = array(
    array('id' => 1, 'name' => 'bob', 'valid' => 1, 'email' => '[email protected]'),
    array('id' => 2, 'name' => 'joe', 'valid' => 1, 'email' => '[email protected]'),
    array('id' => 3, 'name' => 'allison', 'valid' => 1, 'email' => '[email protected]')
);

Then you do array_column($data, 'name'); it would return

array(
    'bob',
    'joe',
    'allison
)

But what if you could do array_column($data, 'id', 'name', 'email') or array_column($data, ['id', 'name', 'email']) and have that return

array(
    'id' => array(
        1,
        2,
        3
    ),
    'name' => array(
        'bob',
        'joe',
        'allison'
    ),
    'email' => array(
        '[email protected]',
        '[email protected]',
        '[email protected]'
    )
)

This allows you to basically achieve the same thing but instead of having to make multiple calls you could get a bunch of arrays back in one iteration.

Not sure if this would really hurt performance in which case maybe it could be a separate method called array_columns, but I think it could be useful.

We actually do this at Vimeo quite a bit. Functions like array_multisort require the data in this format.

[Abeja]

Hello @ramsey, in my work I often use a similar function, perhaps it will be interesting to you:

function array_collect( $Items , $IndexKey , $ValueKey ) {
    if ( is_null( $IndexKey ) && is_null( $ValueKey ) || !count( $Items ) ) return $Items;
    $Result = array();
    if ( !is_null( $IndexKey ) ) {
        $Keys = is_array( $IndexKey ) ? array_values( $IndexKey ) : array( $IndexKey );
        $Key = ( is_null( end( $Keys ) ) && !array_pop( $Keys ) ) ? '[]' : '';
        $Code = '$Result[ $Item[ $Keys[ ' . implode( '] ] ][ $Item[ $Keys[ ' , array_keys( $Keys ) ) . '] ] ]' . $Key . ' = is_null( $ValueKey ) ? $Item : $Item[ $ValueKey ];';
        foreach ( $Items as $Item ) eval( $Code );
    } else
        foreach ( $Items as $Item ) 
            $Result[] = is_null( $ValueKey ) ? $Item : $Item[ $ValueKey ]; 
    return $Result;                             
}

Examples:

$records = array(
    array( 'id' => 1 , 'parent-id' => 7 , 'name' => 'Doe'),
    array( 'id' => 2 , 'parent-id' => 3 , 'name' => 'Smith'),
    array( 'id' => 4 , 'parent-id' => 7 , 'name' => 'Jones'),
);

array_collect( $records , array( 'parent-id' , 'id' ) , 'name' )
// => array( 7 => array( 1 => 'Doe' , 4 => 'Jones' ), 3 => array( 2 => 'Smith' ) )

array_collect( $records , array( 'parent-id' , null ) , 'name' )
// => array( 7 => array( 'Doe' , 'Jones' ), 3 => array ('Smith' ) )

array_collect( $records , null , 'name' )
// => array( 'Doe' , 'Smith' , 'Jones' )

array_collect( $records , 'id' , null )
/* => 
    array(
        1 => array( 'id' => 1 , 'parent-id' => 7 , 'name' => 'Doe'),
        2 => array( 'id' => 2 , 'parent-id' => 3 , 'name' => 'Smith'),
        4 => array( 'id' => 4 , 'parent-id' => 7 , 'name' => 'Jones'),
    );
*/

[pierrejoye]

What's the status of this PR? Are we ready to vote on the RFC?

[ramsey]

I need to make some changes, based on feedback.

[lstrojny]

@ramsey could you call a vote on internals on this function?

[ramsey]

Closing this pull request. It is superseded by pull request #257.

[Majkl578]

@ramsey: Next time, there is no need for new issue, just rebase your branch and force-push it, history will be updated here automatically.