PHP Object Model Manager for Postgresql

This class should be split in order to be able to create Map instances on non table objects (rows ... like a SELECT by example or a VIEW). This would mean separation on everything related to tables (UPDATE, INSERT, DELETE) and the way fields are retreived.

Hi,

I have added a branch with a new feature: a Pomm\Connection\Service layer. This class has the ability to perform transactions in place of Connection in the previous versions. Because complex transaction involving lot of tables should not lie in the controllers and cannot be in a Map class either there was no real place to do such tasks with Pomm.

This is particularly useful when doing data consolidation, let's take an example. Imagine probes sending data. Once in while, those data are aggregated in other tables. Service class could be like the following:

<?php

use Pomm\Connection\Service;

class MyService extends Service
{
    protected function getMapFor($class)
    {
        return $this->connection->getMapFor(sprintf("\My\App\%s", $class));
    }
    public function computeData($to_timestamp)
    {
        try
        {
            $this->begin();
            $this->getMapFor('UsefulData')
                ->fetchFrom($to_timestamp);
            $this->getMapFor('ProbeData')
                ->deleteFrom($to_timestamp);
            $this->commit();
        }
        catch(ConnectionException $e)
        {
            $this->rollback();

            throw $e;
        }

        return $this;
    }
}

Since transactions are not widely available from Connection anymore, I had to refactor the tests to make them pass. But it made them clearer and more understandable (still some work to do here btw) but this also means it breaks backward compatibilty and it makes difficult to migrate big apps from 1.2 to 1.3.

In order not to do something harmful, I have pushed this feature on a servicelayer branch. Would some of you test it and post some feedback ? Thx

I have tried to load Pomm but it gave me error. http://stackoverflow.com/q/20481613/233286

Is it possible to provide it without need for Composer? (with its own autoloader)

For portability reasons and there are people who cannot install Composer.

The initial problem, when I create a second connections:

Warning: pg_close() expects parameter 1 to be resource, null given in Pomm/Connection/Connection.php line 147

This probably silenced the real problem, but I don’t understand what happens.

I was not sure to have to touch stored json as it could be decoded in an entity getter but it appears Json API just export entities as array and any json content is then escaped by a json_encode php call.

Do you think the json converter should use json_decode and json_encode or just treat them as text and let the developers do what they want with it ?

If I have a collection from a query, something like

$positions = $this->get('pomm')
    ->getDatabase()
    ->createConnection()
    ->getMapFor('Survos\GeoBundle\Pomm\Position')
    ->findWhere('true', array(), 'LIMIT 3');

and I pass it to a Twig template and try this

{% for i, d in dateCounts %}
    {{ dump(i, d) }}
{% endfor %}

then instead of entities I get boolean false:

int 0

boolean false

int 1

boolean false

int 2

boolean false

I don't see what I might be doing wrong.

ScanSchemaTool save files in $prefix_dir/Dbname/Schema/ with namespace $namespace\Dbname\Schema

But i my environment, dev and prod have different Dbname, so i have to re-generate models and do a replacement of \Models\DevDbName\DevSchema by \Models\ProdDbname\ProdSchema

But i think it will be more powerfull to set prefix using :

array(
"prefix_dir" => "Foo/Bar/%dbname%/Baz/%schema%",
"namespace" => 'Models\\Foo\\Bar\\%dbname%\\Baz\\%schema%',
);

for my environment i use :

array(
"prefix_dir" => "ProjectName/Models/%schema%",
"namespace" => 'ProcjectName\\Models\\%schema%',
);

Hi,

I know that you can use a custom collection class with your Pomm project, but i suggest to add theses __sleep and __descruct method in the Collection class in order to make them serializable.

In my environment i have many databases connections, one model is related to one databases connections. So in my application i have to do :

$entity1 = $database1
  ->createConnection()
  ->getMapFor('Database1\Schema\MyFirstEntity')
  ->createObject();
$entity2 = $database2
  ->createConnection()
  ->getMapFor('Database2\Schema\MySecondEntity')
  ->createObject();

So we have to remember on which connection the entity is related to. Does the ObjectMap know his connection ?

Is there something like ?

<?php
$map = new \Database\Schema\MyFirstEntityMap();
$e = $map->createObject();
[...]
$map->save($e);

$cnx = $map->getDefaultDatabase();
$map->setDefaultDatabase($cnx);

// also

$map2 = $database2
  ->createConnection()
  ->getMapFor('Database2\Schema\MySecondEntity');
$e2 = $map->createObject();
$map2->save($e2);

I suggest to implement whenever its possible a fluent interface to most Pomm methods - it's cheap, and can be very useful :)

For instance, while I'm writing a UT with PHPUnit, I have to feed a test with a DataProvider. I'd like to keep it compact, and thus would like to write something like that:

return
                [
                    [(new Clients())->hydrate(['activated_on' => new \DateTime('2010-2-5')]), true]
                ];

Unfortunately, hydrate() method returns void, so I have to break my code in multiple parts, doing that:

$client = new Clients();
            $client->hydrate(['activated_on' => new \DateTime('2010-2-5')]);
            $dataSets[] = [$client, true]; 

            return $dataSets;

If hydrate() would return $this; the first sample would be working fine and save me loads of lines of code :)

Hi, is there a way to wrap specific table fields with functions when reading and writing the database?

Specifically, we have spacial database using PostGIS extensions where the "geometry" type is encoded internally. PostGIS provides functions to extract them as geoJSON or other formats. To use these you perform a "select ST_AsGeoJSON(field_name)..." statement.

I see you have converters that can process data retrieved from the database but at that point it's too late. We need the above sql run when executing ->findWhere(), ->findAll(), etc calls on the Entities so that the data returned to PHP from postgres is JSON output from the ST_AsGeoJSON call in the select statement.

Any thoughts on how to accomplish this in POMM?