Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Implementing goMongoDB-like Query expression object evaluation

I've been looking for a MongoDb-like ( http://docs.mongodb.org/manual/applications/read/#find, docs.mongodb.org/manual/reference/operators/ ) query expression object evaluation function implementation or a class. It may cover not all the advanced features, and should have extensible architecture.

MongoDB-like query expression objects are easy for understanding and usage, providing ability to write clean, self-explaining code, because both query and objects to search in, are associative arrays.

Basically talking its a convenient function to extract information from php arrays. Knowing the array structure(the arrayPath), it will allow to perform operations on multidimensional arrays data, without the need for multiple nested loops.

If you are not familiar with MongoDb, take a look at a given expression object and array to search in.

I wrote it as JSON string for simplicity. The object contents makes no sense, just showng the MongoDb query syntax.

MongoDb-like query expression object

{
    "name": "Mongo",
    "type": "db",
    "arch": {
        "$in": [
            "x86",
            "x64"
        ]
    },
    "version": {
        "$gte": 22
    },
    "released": {
        "$or": {
            "$lt": 2013,
            "$gt": 2012
        }
    }
}

The array to search in

[
    {
        "name": "Mongo",
        "type": "db",
        "release": {
            "arch": "x86",
            "version": 22,
            "year": 2012
        }
    },
    {
        "name": "Mongo",
        "type": "db",
        "release": {
            "arch": "x64",
            "version": 21,
            "year": 2012
        }
    },
    {
        "name": "Mongo",
        "type": "db",
        "release": {
            "arch": "x86",
            "version": 23,
            "year": 2013
        }
    }
]

Find using Mongo-like query expressions

So, with the help of the function, we should be able to issue the following query to the target array.

$found=findLikeMongo($array, $queryExpr); //resulting in a $array[0] value;
//@return found array

Get array path using Mongo-like query expressions

$arrayPath=getPathFromMongo($array, $queryExpr);// resulting in array("0")
//@return array path, represented as an array where entries are consecutive keys.

Homework

  • I found that goessner.net/articles/JsonPath/ could possibly cover my needs(not being an exact match because it uses Xpath-like expressions), the caveat is, that it heavily relies on regular expressions and string parsing, what will definitely slow it down compared to array only(JSON like) implementation.

  • Also I've found a similar question here, @stackoverflow Evaluating MongoDB-like JSON Queries in PHP. The resulting answer was to use some SPL functions, which I am used to avoid most of the time.
    Wonder if the author had came up with function, he had been trying to develop.

  • The possible arrayPath implementation was found on thereisamoduleforthat.com/content/dealing-deep-arrays-php, thus the lack of this implementation, is that it relies on pointers.

I know its not a trivial question with a oneliner answer, that's why I'm asking it before starting the actual development of my own class.

I appreciate architecture tips, related or similar code, which may be a good practice example for building php "if..else" expressions on the fly.emphasized text

How to write a non-SPL version?

@Baba provided an excellent class, which is written with the use of SPL. I wonder how to rewrite this code without SPL.

There are two reasons for this

  • calling the class multiple times will give function overhead, that can be avoided rewriting it in raw PHP.
  • it would be easily portable to raw Javascript where SPL is not available, leading to easier code maintenance on both platforms.

Results

The created ArrayQuery class is published on Github, consider checking-out the repository for updates.

SPL, raw PHP version and Chequer2 FORP profiler output

In brief-

  1. the raw PHP version performs 10x faster than the SPL one, consuming 20% less memory.
  2. Chequer2 class performs 40% slower than PHP SPL class, and almost 20x slower than raw PHP version.
  3. MongoDb is the fastest(10x faster than raw PHP implementation and consumes 5x less memory), do not use these classes unless you are sure you want to avoid interaction with MongoDb.

MongoDb version

MongoDb reference profiling results

SPL version

PHP with SPL class profiling results

Raw PHP(latest ArrayQuery class) version

raw PHP ArrayQuery class profiling results

Chequer2 version

Chequer2 PHP class profiling results

MongoDb reference test profiling code

$m = new MongoClient(); // connect
$db = $m->testmongo; // select a database
$collection = $db->data;
$loops=100;
for ($i=0; $i<$loops; $i++) {
    $d = $collection->find(array("release.year" => 2013));
}
print_r( iterator_to_array($d) );

PHP with SPL class profiling code

include('data.php');
include('phpmongo-spl.php');
$s = new ArrayCollection($array, array("release.year" => 2013),false);
$loops=100;
for ($i=0; $i<$loops; $i++) {
    $d = $s->parse();
}
print_r( $d );

The SPL class parse() function has been slightly modified to return the value after execution, it could be also be modified to accept expression, but it's not essential for profiling purposes as the expression is being reevaluated every time.

raw PHP(latest ArrayQuery class) profiling code

include('data.php');
include('phpmongo-raw.php');
$s = new ArrayStandard($array);
$loops=100;
for ($i=0; $i<$loops; $i++) {
    $d = $s->find(array("release.year" => 2013));
}
print_r( $d );

chequer2 PHP profiling code

<?php
include('data.php');
include('../chequer2/Chequer.php');
$query=array("release.year" => 2013);

$loops=100;
for ($i=0; $i<$loops; $i++) {
    $result=Chequer::shorthand('(.release.year > 2012) ? (.) : NULL')
        ->walk($array);

}
print_r($result);
?>

data used(same as @baba provided in his answer)

$json = '[{
    "name":"Mongo",
    "type":"db",
    "release":{
        "arch":"x86",
        "version":22,
        "year":2012
    }
},
{
    "name":"Mongo",
    "type":"db",
    "release":{
        "arch":"x64",
        "version":21,
        "year":2012
    }
},
{
    "name":"Mongo",
    "type":"db",
    "release":{
        "arch":"x86",
        "version":23,
        "year":2013
    }
},      
{
    "key":"Diffrent",
    "value":"cool",
    "children":{
        "tech":"json",
        "lang":"php",
        "year":2013
    }
}
]';

$array = json_decode($json, true);

the forp-ui slightly modified sample ui loader(to be called with ?profile=FILE_TO_PROFILE)

<!doctype html>
<html>
    <head>
        <style>
            body {margin : 0px}
        </style>
    </head>
    <body>
        <div class="forp"></div>
<?php
register_shutdown_function(
    function() {
        // next code can be append to PHP scripts in dev mode
        ?>
        <script src="../forp-ui/js/forp.min.js"></script>
        <script>
        (function(f) {
            f.find(".forp")
             .each(
                function(el) {
                    el.css('margin:50px;height:300px;border:1px solid #333');
                }
             )
             .forp({
                stack : <?php echo json_encode(forp_dump()); ?>,
                //mode : "fixed"
             })
        })(forp);
        </script>
        <?php
    }
);

// start forp
forp_start();

// our PHP script to profile
include($_GET['profile']);

// stop forp
forp_end();
?>
</body>
</html>
like image 631
fitheflow Avatar asked Feb 20 '13 03:02

fitheflow


1 Answers

Latest Update

@baba has given a great raw PHP version of a class implementing MongoDB-like query expression object evaluation, but the output structure differs a bit, I mean the dot notation in the nested array output( [release.arch] => x86 ), instead of regular arrays( [release] => Array([arch] => x86) ). I would appreciate your tip how to make the class fully compatible with mongoDB in this order, as it seems its strictly tied to the raw PHP class implementation.

=======================================================================

Answer:

What you want is very easy, All you need is 2 corrections in the current code input and output loop and you would get your new format.

What do i mean ?

A. Changed

  foreach ( $array as $part ) {
        $this->flatten[] = $this->convert($part);
    }

To

    foreach ( $array as $k => $part ) {
        $this->flatten[$k] = $this->convert($part);
    }

B. Changed

    foreach ( $this->flatten as $data ) {
        $this->check($find, $data, $type) and $f[] = $data;
    }

To:

    foreach ( $this->flatten as $k => $data ) {
        $this->check($find, $data, $type) and $f[] = $this->array[$k];
    }

New Array for resting 

$json = '[
  {
    "name": "Mongo",
    "release": {
      "arch": "x86",
      "version": 22,
      "year": 2012
    },
    "type": "db"
  },
  {
    "name": "Mongo",
    "release": {
      "arch": "x64",
      "version": 21,
      "year": 2012
    },
    "type": "db"
  },
  {
    "name": "Mongo",
    "release": {
      "arch": "x86",
      "version": 23,
      "year": 2013
    },
    "type": "db"
  },
  {
    "name": "MongoBuster",
    "release": {
      "arch": [
        "x86",
        "x64"
      ],
      "version": 23,
      "year": 2013
    },
    "type": "db"
  },
  {
    "children": {
      "dance": [
        "one",
        "two",
        {
          "three": {
            "a": "apple",
            "b": 700000,
            "c": 8.8
          }
        }
      ],
      "lang": "php",
      "tech": "json",
      "year": 2013
    },
    "key": "Diffrent",
    "value": "cool"
  }
]';

$array = json_decode($json, true);

Simple Test

$s = new ArrayStandard($array);
print_r($s->find(array("release.arch"=>"x86")));

Output

Array
(
    [0] => Array
        (
            [name] => Mongo
            [type] => db
            [release] => Array
                (
                    [arch] => x86
                    [version] => 22
                    [year] => 2012
                )

        )

    [1] => Array
        (
            [name] => Mongo
            [type] => db
            [release] => Array
                (
                    [arch] => x86
                    [version] => 23
                    [year] => 2013
                )

        )

)

If you also want to retain original array key position you can have

    foreach ( $this->flatten as $k => $data ) {
        $this->check($find, $data, $type) and $f[$k] = $this->array[$k];
    }

Just for Fun Part

A. Support for regex

Just for fun i added support for $regex with alias $preg or $match which means you can have

print_r($s->find(array("release.arch" => array('$regex' => "/4$/"))));

Or

print_r($s->find(array("release.arch" => array('$regex' => "/4$/"))));

Output

Array
(
    [1] => Array
        (
            [name] => Mongo
            [type] => db
            [release] => Array
                (
                    [arch] => x64
                    [version] => 21
                    [year] => 2012
                )

        )

)

B. Use Simple array like queries

$queryArray = array(
        "release" => array(
                "arch" => "x86"
        )
);
$d = $s->find($s->convert($queryArray));

$s->convert($queryArray) has converted

Array
(
    [release] => Array
        (
            [arch] => x86
        )

)

To

Array
(
    [release.arch] => x86
)

C. Modulus $mod

print_r($s->find(array(
        "release.version" => array(
                '$mod' => array(
                        23 => 0
                )
        )
)));

 //Checks release.version % 23 == 0 ;

D. Count elements with $size

print_r($s->find(array(
        "release.arch" => array(
                '$size' => 2
        )
)));

// returns count(release.arch) == 2;

E. Check if it matches all element in array $all

print_r($s->find(array(
        "release.arch" => array(
                '$all' => array(
                        "x86",
                        "x64"
                )
        )
)));

Output

Array
(
    [3] => Array
        (
            [name] => MongoBuster
            [release] => Array
                (
                    [arch] => Array
                        (
                            [0] => x86
                            [1] => x64
                        )

                    [version] => 23
                    [year] => 2013
                )

            [type] => db
        )

)

F. If you are not sure of the element key name then you ca use $has its like the opposite of $in

print_r($s->find(array(
        "release" => array(
                '$has' => "x86"
        )
)));

=======================================================================

Old Update

@Baba provided an excellent class, which is written with the use of SPL. I wonder how to rewrite this code without SPL. The reason is that calling this class multiple times will give function overhead, that can be avoided rewriting it in raw PHP, and maybe using goto statement in final version, to avoid recursive function calls.

=======================================================================

Since you don't want SPL and functions .. it took a while but i was able to come up with alternative class that is also flexible and easy to use

To avoid loading the array multiple times you declare it once :

$array = json_decode($json, true);
$s = new ArrayStandard($array);

A. Find where release.year is 2013

$d = $s->find(array(
        "release.year" => "2013"
));
print_r($d);

Output

Array
(
    [0] => Array
        (
            [name] => Mongo
            [type] => db
            [release.arch] => x86
            [release.version] => 23
            [release.year] => 2013
        )

)

B. For the first time you can run complex $and or $or statement like find where release.arch = x86 and release.year = 2012

$d = $s->find(array(
        "release.arch" => "x86",
        "release.year" => "2012"
), ArrayStandard::COMPLEX_AND);

print_r($d);

Output

Array
(
    [0] => Array
        (
            [name] => Mongo
            [type] => db
            [release.arch] => x86
            [release.version] => 22
            [release.year] => 2012
        )

)

C. Imagine a much more complex query

$d = $s->find(array(
        "release.year" => array(
                '$in' => array(
                        "2012",
                        "2013"
                )
        ),
        "release.version" => array(
                '$gt' => 22
        ),
        "release.arch" => array(
                '$func' => function ($a) {
                    return $a == "x86";
                }
        )
), ArrayStandard::COMPLEX_AND);

print_r($d);

Output

Array
(
    [0] => Array
        (
            [name] => Mongo
            [type] => db
            [release.arch] => x86
            [release.version] => 23
            [release.year] => 2013
        )

)

The new Modified class

class ArrayStandard {
    const COMPLEX_OR = 1;
    const COMPLEX_AND = 2;
    private $array;
    private $tokens;
    private $found;

    function __construct(array $array) {
        $this->array = $array;
        foreach ( $array as $k => $item ) {
            $this->tokens[$k] = $this->tokenize($item);
        }   
    }

    public function getTokens() {
        return $this->tokens;
    }

    public function convert($part) {
        return $this->tokenize($part, null, false);
    }

    public function find(array $find, $type = 1) {
        $f = array();
        foreach ( $this->tokens as $k => $data ) {
            $this->check($find, $data, $type) and $f[$k] = $this->array[$k];
        }
        return $f;
    }

    private function check($find, $data, $type) {
        $o = $r = 0; // Obigation & Requirement
        foreach ( $data as $key => $value ) {
            if (isset($find[$key])) {
                $r ++;
                $options = $find[$key];
                if (is_array($options)) {
                    reset($options);
                    $eK = key($options);
                    $eValue = current($options);
                    if (strpos($eK, '$') === 0) {
                        $this->evaluate($eK, $value, $eValue) and $o ++;
                    } else {
                        throw new InvalidArgumentException('Missing "$" in expession key');
                    }
                } else {
                    $this->evaluate('$eq', $value, $options) and $o ++;
                }
            }
        }

        if ($o === 0)
            return false;

        if ($type == self::COMPLEX_AND and $o !== $r)
            return false;

        return true;
    }

    private function getValue(array $path) {
        return count($path) > 1 ? $this->getValue(array_slice($path, 1), $this->array[$path[0]]) : $this->array[$path[0]];
    }

    private function tokenize($array, $prefix = '', $addParent = true) {
        $paths = array();
        $px = empty($prefix) ? null : $prefix . ".";
        foreach ( $array as $key => $items ) {
            if (is_array($items)) {
                $addParent && $paths[$px . $key] = json_encode($items);
                foreach ( $this->tokenize($items, $px . $key) as $k => $path ) {
                    $paths[$k] = $path;
                }
            } else {
                $paths[$px . $key] = $items;
            }
        }
        return $paths;
    }

    private function evaluate($func, $a, $b) {
        $r = false;

        switch ($func) {
            case '$eq' :
                $r = $a == $b;
                break;
            case '$not' :
                $r = $a != $b;
                break;
            case '$gte' :
            case '$gt' :
                if ($this->checkType($a, $b)) {
                    $r = $a > $b;
                }
                break;

            case '$lte' :
            case '$lt' :
                if ($this->checkType($a, $b)) {
                    $r = $a < $b;
                }
                break;
            case '$in' :
                if (! is_array($b))
                    throw new InvalidArgumentException('Invalid argument for $in option must be array');
                $r = in_array($a, $b);
                break;

            case '$has' :
                if (is_array($b))
                    throw new InvalidArgumentException('Invalid argument for $has array not supported');
                $a = @json_decode($a, true) ?  : array();
                $r = in_array($b, $a);
                break;

            case '$all' :
                $a = @json_decode($a, true) ?  : array();
                if (! is_array($b))
                    throw new InvalidArgumentException('Invalid argument for $all option must be array');
                $r = count(array_intersect_key($a, $b)) == count($b);
                break;

            case '$regex' :
            case '$preg' :
            case '$match' :

                $r = (boolean) preg_match($b, $a, $match);
                break;

            case '$size' :
                $a = @json_decode($a, true) ?  : array();
                $r = (int) $b == count($a);
                break;

            case '$mod' :
                if (! is_array($b))
                    throw new InvalidArgumentException('Invalid argument for $mod option must be array');
                list($x, $y) = each($b);
                $r = $a % $x == 0;
                break;

            case '$func' :
            case '$fn' :
            case '$f' :
                if (! is_callable($b))
                    throw new InvalidArgumentException('Function should be callable');
                $r = $b($a);
                break;

            default :
                throw new ErrorException("Condition not valid ... Use \$fn for custom operations");
                break;
        }

        return $r;
    }

    private function checkType($a, $b) {
        if (is_numeric($a) && is_numeric($b)) {
            $a = filter_var($a, FILTER_SANITIZE_NUMBER_FLOAT);
            $b = filter_var($b, FILTER_SANITIZE_NUMBER_FLOAT);
        }

        if (gettype($a) != gettype($b)) {
            return false;
        }
        return true;
    }
}
like image 110
Baba Avatar answered Oct 21 '22 22:10

Baba