Hi, your project is still young so it's the best time to choose and use a naming convention. By there I mean that you should not, for example, call your function Where but where (you can see here that quite all frameworks work this way) (The only one I found to use PascalCase is ... Apache, whose style guide link is dead

Check here for more informations and great stuff : http://www.phptherightway.com/

Greetings, PunKeel

This provides the tools necessary to analyse a given expression tree.

This branch currently supports:

• Getting type and reflection information from almost all types of expressions.
• Emulating PHP's native type system for all operators as of PHP 5.6
• Extendable type system implementation with type data for date for date time classes, native interfaces (eg ArrayAccess), string / array / math internal functions and the ITraversable query API.

Unsupported:

• Dynamic language features such as variable function / method calls, variable class types...
• Proper handling of references for all use cases (this is almost impossible in a language such as PHP)
• Only a small subset of internal classes or functions type data is implemented.
• User defined classes type data. Possible through utilising docblocks via phpDocumentor/ReflectionDocBlock

Hey.

I am working with Yii's 1.1.x version and I would like to extend CActiveRecord to implement Pinq.

So I looked into the documentation but all I find is either a very long class list, or examples where the original classes with their implementation is not shown - or at least a pseudo implementation. There also seems to be no detail how Pinq is actually used by the developer.

I am aware of this one: http://elliotswebsite.com/Pinq/api.html ...but it is just a method listing and an explanation of what is for what, but i can't make up an implementation from this.

Could you maybe give a short example for a possible implementation? Thanks!

Kind regards, Ingwie

Different data-sources, such as RDBMSs or MongoDB, use different DSLs (SQL/JS queries, etc).

What I found out so far is this page in the docs, which seems to simulate usage of a visitor on different types of filtering/aggregation requests.

It is really hard for newcomers to understand how to use the visitors to lazily evaluate Pinq expressions to convert them into, for example, SQL.

A simplified SQL that appends conditionals to a given SQL string would probably be useful in those examples.

Hi there,

I am trying to do some grouping on a dataset by "author" and would return a new dataset with "author" and its relevant count.

The code is something like this:

$filter1 = $data
        ->where(function($row)
        {
            return $row['price'] > 90 && $row['price'] < 99;
        })
        ->orderByDescending(function($row)
{
    return $row['id'];
});

$filter1 returns the results where its price is between 90 and 99, perfect.

Next:

$filter2 = $filter1
        ->groupBy(function($row)
        {
            return $row['author'];
        })
        ->select(
        function(ITraversable $filter1)
{
    return ['Author' => $filter1['author'], 'Count' => $filter1->count()];
}
);

I would like to group the filtered dataset ($filter1) by "author" and get the count for each "author".

I have some problem to set the select function body:

return ['Author' => $filter1['author'], 'Count' => $filter1->count()];

I am expecting to return an array with two keys ('Author' and 'Count') with respective author name and count, like:

author1 12 author2 15 ...

How can I achieve this?

Hi,

I am trying to use Pinq to mimic a facet search and created a repo here: https://github.com/taylorren/pinq.

The issue is that after creating a facet result, I am not able to display it as in my "demo1" route.

Error message:

Catchable fatal error: Argument 1 passed to pinqDemo{closure}() must be an instance of pinqDemo\ITraversable, instance of Pinq\Traversable given, called in F:\www\pinq\vendor\timetoogo\pinq\Source\Iterators\ProjectionIterator.php on line 40 and defined in F:\www\pinq\web\pinqDemo.php on line 17

The function in question:

function facetauthor($data)
    {
        $filter = $data
                ->groupBy(function($row)
                {
                    return $row['author'];
                })
                ->select(
                        function(ITraversable $data)
                {
                    return ['author' => $data->first()['author'], 'count' => $data->count()];
                }
                );

        return $filter;
    }

Your competitor here. :grinning:

Pinq queries containing join and groupJoin can be thousands of times slower than native implementation, or any other LINQ library, when long arrays are joined (thousands of items). I have no idea what is going on, as I have trouble navigating your immense code base, but you should probably look into it.

I've run the tests, your code is often an order of magnitude slower than YaLinqo, sometimes less, sometimes much more (like in the cases above). See gist with benchmark results (master branch). I've significantly improved performance of sorting thanks to your implementation. I didn't know array_multisort is so much faster than usort — the difference caused Pinq to be the fastest of the major LINQ libraries in the sorting test (well, no longer so, but it was).

Considering your library is the only one providing LINQ to databases, I think it's critical to avoid performance traps in the basic functionality, even if the library is the slowest of the bunch.

Technically, we aren't even competitors, considering YaLinqo is designed to be fast and expressive LINQ to objects and Pinq is designed to be LINQ to both objects and databases. Functionality and performance is so different that there isn't much to compete over. I'm still surprised by the lack of functionality in the basic methods of Pinq though, but I guess limiting expressiveness is required to make implementing query providers actually possible.

What is the current state of MySQL query provider? Your website and readme say that it's a "demo", so I'm curious what plans you have on implementing production-ready query providers.

By the way, how did you manage to receive more than 200 stars on GitHub? Did SitePoint cause so much traffic or is it something else?

Anyway, thanks for writing your library. Ginq and Pinq forced me to update my own library: add missing methods, improve peformance, add integartion with CI and code quality tools. :grinning:

Forgive me if I've missed something big, but it seems that IQueryable has a lot of methods that let you pass functions, and provides a nice eloquent interface. I would like to implement it in such a way that I can guarantee that no function parsing happens in userland php at runtime. Is there a reason why this would not be possible or why the lazy evaluation must use parsing at runtime? If not, is there a set of classes from which I should extend, or do I just satisfy the interfaces without inheriting?

From what I can tell, the requests/filters/etc all need to be passed a function of a given spec, and then they should be able to just call the darn thing without needing to parse it and make a decision about what it means.

I want to use own Traversable class that adds a custom method, but it's impossible. It would be useful to have a environment object that can be passed to each iterator:

$pinq = new Environment(['Traversable' => 'Vendor\Traversable', 'GroupedTraversable' => ...]);
$pinq->traversable($data);

Has there been a way/method to generate SQL using PINQ? If not, I will have to go and code one. But I would like to know beforehand so I do not re-invent the wheel. :)

I get en error in IteratorGenerator, method &getIterator: Only variable references should be returned by reference. This should fix the problem. The error first occurred when I updated my server from php 5.6.7 to 5.6.8...

$generator = $this->iteratorGenerator($this->iterator);
return $generator;

I'm not really sure what exactly is the problem here and if this is a good fix. Otherwise I could create a pullrequest.

For instance:

$traversable = Traversable::from([
    ['date' => new \DateTime('1/1/2000'), 'message' => '...'],
    ['date' => new \DateTime('1/1/2000'), 'message' => '...'],
    ['date' => new \DateTime('2/1/2000'), 'message' => '...'],
]);

$results = $traversable
    ->groupBy(function ($row) { return $row['date']; })
    ->select(function (ITraversable $group, \DateTime $key) {
        return $key->format('d/m/Y') . ':' . $group->count(); 
    })
    ->asArray();

Due to strict comparisons on the group by this would yield some rather unexpected results:

['01/01/2000:1', '01/01/2000:1', '02/01/2000:1']

When what is actually wanted is:

['01/01/2000:2', '02/01/2000:1']

This PR changes the Identity::hash behaviour such that DateTimes are treated as a value type based on their timestamps (and class name) which would solve this problem.

The documentation is currently only in the gh-pages branch, which makes it quite easy for problems to popup if there are API breakages and the documentation isn't updated in sync.

Instead, the doc should probably be put into the master branch (with the repo itself) and deployed into the gh-pages branch via a script.