Currently I am using a DynamoDB instance for my social media application. While designing the schema I sticked to the "one table" rule. So I am putting every data in the same table like posts, users, comments etc. Now I want to make flexible queries for my data. Here I found out that I could use the #searchable annotation to create an Elastic Search instance for a table which is annotated with #model
In my GraphQL schema I only have one #model, since I only have one table. My problem now is that I don't want to make everything in the table searchable, since that would be most likely very expensive. There are some data which don't have to be added to the Elastic Search instance (For example comment related data). How could I handle it ? Do I really have to split my schema down into multiple tables to be able to manage the #searchable annotation ? Couldn't I decide If the row should be stored to the Elastic Search with help of the Partitionkey / Primarykey, acting like a filter ?
The current implementation of the amplify-cli uses a predefined python Lambda that are added once we add the #searchable directive to one of our models.
The Lambda code can not be edited and currently, there is no option to define a custom Lambda, you read about it
https://github.com/aws-amplify/amplify-cli/issues/1113
https://github.com/aws-amplify/amplify-cli/issues/1022
If you want a custom Lambda where you can filter what goes to the Elasticsearch Instance, you can follow the steps described here https://github.com/aws-amplify/amplify-cli/issues/1113#issuecomment-476193632
The closest you can get is by creating a template in amplify\backend\api\myapiname\stacks\ where you can manage all the resources related to Elasticsearch. A good start point is to
Add #searchable to one of your model in the schema.grapql
Run amplify api gql-compile
Copy the generated template in the build folder, \amplify\backend\api\myapiname\build\stacks\SearchableStack.json to amplify\backend\api\myapiname\stacks\
Remove the #searchable directive from the model added in step 1
Start editing your new template copied in step 3
Add a Lambda and use it in the template as the resolver for the DynamoDB Stream
Using this approach will give you total control of the resources related to the Elasticsearch service, but, will also require to do it all by your own.
Or, just go by creating a table for each model.
Hope it helps
It is now possible to override the generated streaming function code as well.
thanks to the AWS Support for the information provided
leaved a message on the related github issue as well https://github.com/aws-amplify/amplify-category-api/issues/437#issuecomment-1351556948
All you need is to run
amplify override api
edit the corresponding overrode.ts
change the code with the resources.opensearch.OpenSearchStreamingLambdaFunction.code
resources.opensearch.OpenSearchStreamingLambdaFunction.functionName = 'python_streaming_function';
resources.opensearch.OpenSearchStreamingLambdaFunction.handler = 'index.lambda_handler';
resources.opensearch.OpenSearchStreamingLambdaFunction.code = {
zipFile: `
# python streaming function customized code goes here
`
}
Resources:
[1] https://docs.amplify.aws/cli/graphql/override/#customize-amplify-generated-resources-for-searchable-opensearch-directive
[2]AWS::Lambda::Function Code - Properties - https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-properties-lambda-function-code.html#aws-properties-lambda-function-code-properties
In dev environment there are 16 database queries (using doctrine 2.6 & only dql queries in my repositories). Somehow in production the exact same page shows 101 queries - because all the relations are queried, even most of them are not used. I have tried different fetch modes like EAGER but it didn't change anything.
Here is a sample annotation which gets queried but is never used on the page:
/**
* #var Doctrine\Common\Collections\Collection
*
* #Assert\Valid(traverse=true)
* #ORM\OneToMany(targetEntity="Translation", mappedBy="origin", cascade={"persist", "remove"}, orphanRemoval=true, fetch="EAGER")
* #ORM\OrderBy({"locale"="ASC"})
*/
private $translations;
Could it be some kind of configuration error?
Solved it – the root cause was apcu. When I switched from:
doctrine:
orm:
metadata_cache_driver: apcu
query_cache_driver: apcu
result_cache_driver: apcu
to
doctrine:
orm:
metadata_cache_driver: apc
query_cache_driver: apc
result_cache_driver: apc
I got down to 9 requests instead of 16 (dev) and 101 (prod via apcu). Maybe it was some server side issue with the module.
I'm writing writing a Spring Restful microservice that relies on Spring Data Rest and Spring Data Neo4J.
We don't want to expose the internal Neo4J node identifier in the HAL links of the JSON response. The reason being (as far as I understand) that these identifiers are reused by Neo4J in case of node deletion. If this is the case this will present us with data integrity problems. And so we'd rather use a natural key, for example, UUID. Please correct me if my assumption regarding the reuse of neo4J node Ids is wrong.
What we want to achieve is using a Natural Key for all CRUD operations instead of the node Id (ie: PUT http://localhost:8080/apiname/5448ae86-fe87-4daf-bfb5-985522a1cf14 with some body).
Our first approach was to extend NodeGraphRepositoryImp . E.g.
protected Node getById(UUID id) {
Node node = (Node) this.findByPropertyValue("uuid_id", id);
return template.getNode(node.getId());
}
And instantiate it by injecting a customized Neo4jTemplate in the Neo4J configuration.
However this approach doesn't work for the following reasons:
1) I cannot define any other idProperty different from the one annotated with #GraphId in the NodeEntity class. As a result Spring Data Rest ends up trying to assign the natural key (e.g. UUID) to the neo4j node id field (that is a Long) and fails on type conversion.
2) It seems that the Spring Data Neo4J is not using our custom class that extends NodeGraphRepositoryImpl but the original not extended NodeGraphRepositoryImpl class instead.
Maybe this approach is wrong. Could you please recommend a way to achieve it?
Thanks a lot for your help.
Spring Data Rest introduced a BackendIdConverter for overriding/customizing the field that gets exposed in the URIs.
Please have a look at DATAREST-155 (https://jira.spring.io/plugins/servlet/mobile#issue/DATAREST-155)
Can you give it a try?
You can also have a look at the following thread:
How can you customise self, parent, children links in spring data rest with neo4j
I found strange queries in log and detected source:
echo app()->getCountSql(); // 2
$table->findAll();
echo app()->getCountSql(); // 21
$table is an autogenerated repository for page\route class
after loading routes doctrine loading page for each route in separate query.
i checked proxy genereation, i manualy created Proxy__CG__\page\page and Proxy__CG__\page\route objects and it works fine. I tested metadatas and in route assosiation "page" i found fetch = 2 (LEZY).
What can be reason for this?
i've found answer in one of open doctrine bugs http://www.doctrine-project.org/jira/browse/DDC-1390, i hope this can save some one few hours.
Is it possible to generate fixtures from an existing DB in Symfony2/Doctrine? How could I do that?
Example:
I have defined 15 entities and my symfony2 application is working. Now some people are able to browse to the application and by using it it had inserted about 5000 rows until now. Now I want the stuff inserted as fixtures, but I don’t want to do this by hand. How can I generate them from the DB?
There's no direct manner within Doctrine or Symfony2, but writing a code generator for it (either within or outside of sf2) would be trivial. Just pull each property and generate a line of code to set each property, then put it in your fixture loading method. Example:
<?php
$i = 0;
$entities = $em->getRepository('MyApp:Entity')->findAll();
foreach($entities as $entity)
{
$code .= "$entity_{$i} = new MyApp\Entity();\n";
$code .= "$entity_{$i}->setMyProperty('" . addslashes($entity->getMyProperty()); . "'); \n");
$code .= "$manager->persist($entity_{$i}); \n $manager->flush();";
++$i;
}
// store code somewhere with file_put_contents
As I understand your question, you have two databases: the first is already in production and filled with 5000 rows, the second one is a new database you want to use for new test and development. Is that right ?
If it is, I suggest you to create in you test environment two entity manager: the first will be the 'default' one, which will be used in your project (your controllers, etc.). The second one will be used to connect to your production database. You will find here how to deal with multiple entity manager : http://symfony.com/doc/current/cookbook/doctrine/multiple_entity_managers.html
Then, you should create a Fixture class which will have access to your container. There is an "how to" here : http://symfony.com/doc/current/bundles/DoctrineFixturesBundle/index.html#using-the-container-in-the-fixtures.
Using the container, you will have access to both entity manager. And this is the 'magic': you will have to retrieve the object from your production database, and persist them in the second entity manager, which will insert them in your test database.
I point your attention to two points:
If there are relationship between object, you will have to take care to those dependencies: owner side, inversed side, ...
If you have 5000 rows, take care on the memory your script will use. Another solution may be use native sql to retrieve all the rows from your production database and insert them in your test database. Or a SQL script...
I do not have any code to suggest to you, but I hope this idea will help you.
I assume that you want to use fixtures (and not just dump the production or staging database in the development database) because a) your schema changes and the dumps would not work if you update your code or b) you don't want to dump the hole database but only want to extend some custom fixtures. An example I can think of is: you have 206 countries in your staging database and users add cities to those countries; to keep the fixtures small you only have 5 countries in your development database, however you want to add the cities that the user added to those 5 countries in the staging database to the development database
The only solution I can think of is to use the mentioned DoctrineFixturesBundle and multiple entity managers.
First of all you should configure two database connections and two entity managers in your config.yml
doctrine:
dbal:
default_connection: default
connections:
default:
driver: %database_driver%
host: %database_host%
port: %database_port%
dbname: %database_name%
user: %database_user%
password: %database_password%
charset: UTF8
staging:
...
orm:
auto_generate_proxy_classes: %kernel.debug%
default_entity_manager: default
entity_managers:
default:
connection: default
mappings:
AcmeDemoBundle: ~
staging:
connection: staging
mappings:
AcmeDemoBundle: ~
As you can see both entity managers map the AcmeDemoBundle (in this bundle I will put the code to load the fixtures). If the second database is not on your development machine, you could just dump the SQL from the other machine to the development machine. That should be possible since we are talking about 500 rows and not about millions of rows.
What you can do next is to implement a fixture loader that uses the service container to retrieve the second entity manager and use Doctrine to query the data from the second database and save it to your development database (the default entity manager):
<?php
namespace Acme\DemoBundle\DataFixtures\ORM;
use Doctrine\Common\DataFixtures\FixtureInterface;
use Doctrine\Common\Persistence\ObjectManager;
use Symfony\Component\DependencyInjection\ContainerAwareInterface;
use Symfony\Component\DependencyInjection\ContainerInterface;
use Acme\DemoBundle\Entity\City;
use Acme\DemoBundle\Entity\Country;
class LoadData implements FixtureInterface, ContainerAwareInterface
{
private $container;
private $stagingManager;
public function setContainer(ContainerInterface $container = null)
{
$this->container = $container;
$this->stagingManager = $this->container->get('doctrine')->getManager('staging');
}
public function load(ObjectManager $manager)
{
$this->loadCountry($manager, 'Austria');
$this->loadCountry($manager, 'Germany');
$this->loadCountry($manager, 'France');
$this->loadCountry($manager, 'Spain');
$this->loadCountry($manager, 'Great Britain');
$manager->flush();
}
protected function loadCountry(ObjectManager $manager, $countryName)
{
$country = new Country($countryName);
$cities = $this->stagingManager->createQueryBuilder()
->select('c')
->from('AcmeDemoBundle:City', 'c')
->leftJoin('c.country', 'co')
->where('co.name = :country')
->setParameter('country', $countryName)
->getQuery()
->getResult();
foreach ($cities as $city) {
$city->setCountry($country);
$manager->persist($city);
}
$manager->persist($country);
}
}
What I did in the loadCountry method was that I load the objects from the staging entity manager, add a reference to the fixture country (the one that already exists in your current fixtures) and persist it using the default entity manager (your development database).
Sources:
DoctrineFixturesBundle
How to work with Multiple Entity Managers
you could use https://github.com/Webonaute/DoctrineFixturesGeneratorBundle
It add ability to generate fixtures for single entity using commands like
$ php bin/console doctrine:generate:fixture --entity=Blog:BlogPost --ids="12 534 124" --name="bug43" --order="1"
Or you can create full snapshot
php app/console doctrine:generate:fixture --snapshot --overwrite
The Doctrine Fixtures are useful because they allow you to create objects and insert them into the database. This is especially useful when you need to create associations or say, encode a password using one of the password encoders. If you already have the data in a database, you shouldn't really need to bring them out of that format and turn it into PHP code, only to have that PHP code insert the same data back into the database. You could probably just do an SQL dump and then re-insert them into your database again that way.
Using a fixture would make more sense if you were initiating your project but wanted to use user input to create it. If you had in your config file the default user, you could read that and insert the object.
The AliceBundle can help you doing this. Indeed it allows to load fixtures with YAML (or PHP array) files.
For instance you can define your fixtures with:
Nelmio\Entity\Group:
group1:
name: Admins
owner: '#user1->id'
Or with the same structure in a PHP array. It's WAY easier than generating working PHP code.
It also supports references:
Nelmio\Entity\User:
# ...
Nelmio\Entity\Group:
group1:
name: Admins
owner: '#user1'
In the doctrine_fixture cookbook, you can see in the last example how to get the service container in your entity.
With this service container, you can retrieve the doctrine service, then the entity manager. With the entity manager, you will be able to get all the data from your database you need.
Hope this will help you!