@anidotnet - ok no worries Anindya I'll try to create something. But have a rookie question. The in-memory instance doesn't seem to store anything in the heap. I'm always getting null or actually nothing is returned. what am I doing wrong here?
// In Memory instance
Nitrite db = Nitrite.builder()
.openOrCreate();
// Creating repository
ObjectRepository<Employee> repository = db.getRepository(Employee.class);
// Inserting Employee
Employee emp = new Employee();
emp.setEmpId(12564);
emp.setAddress("12 Some Street");
repository.insert(employees);
// Listing all the employees
List<Employee> employees = repository.find().toList();
for (Employee employee: employees) {
System.out.println(employee);
}Output - No records are being returned.
Any clue what am I doing wrong here?
It works fine if I store them in a file. Its just the in-memory. Doesn't want to store anything.
I'm running it on Jetty as a web application. Not sure if that makes any difference though.
send me at anindya @ dizitart.com
a. If you want to maintain 2 types of db, you must create 2 separate Nitrite variables.
b. In your closeConnection method, you are closing the db. In-memory db is not persistent. As soon as you close the db, it looses data unlike file store db.
Thanks Anindya. Appreciate your response. Actually I just want to maintain the in-memory version of the db. The github code that I provided was more of an example to demonstrate 2 different behaviors. Unfortunately, I see the same behavior even if I don't close the connection and/or use a different variables.
public final class DatabaseConnection {
private static final DatabaseConnection DB_CONN = new DatabaseConnection();
private Nitrite db;
private ObjectRepository<Employee> repository;
private DatabaseConnection(){
db = Nitrite.builder().openOrCreate();
repository = db.getRepository(Employee.class);
}
public static DatabaseConnection getInstance() {
return DB_CONN;
}
/**
* Inserting into inmemory
* @param employee
* @return
*/
public List<Employee> insertIntoMemory(Employee employee) {
return insert(employee);
}
/**
* Find from memory
* @param offset
* @param size
* @return
*/
public List<Employee> findFromMemory(int offset, int size){
return find(offset, size);
}
private List<Employee> insert(Employee employee){
List<Employee> employees = new ArrayList<>();
WriteResult wr = null;
try {
wr = repository.insert(employee);
System.out.println("\n Affected Rows : " + wr.getAffectedCount());
return repository.find().toList();
} catch (Exception e) {
System.out.println("Unable to insert transaction. ");
e.printStackTrace();
}
return employees;
}
private List<Employee> find(int offset, int size) {
List<Employee> employees = new ArrayList<>();
try {
employees = repository.find(limit(offset, size)).toList();
} catch (Exception e) {
System.out.println("Unable to find transactions. ");
e.printStackTrace();
}
return employees;
}
}
And here is the result
HTTP/1.1 200 OK
Date: Thu, 05 Dec 2019 18:30:39 GMT
Content-Type: application/json
Content-Length: 365
Server: Jetty(9.4.22.v20191022)
[
{
"empId": "f843e184-46f0-4967-b3e1-fecbb4dfcf23",
"address": "From Memory. 123, North Pole."
},
{
"empId": "6a75a311-5cb7-44fd-9ca8-1b7a86ea3b22",
"address": "From Memory. 123, North Pole."
},
{
"empId": "c2e3c339-26b3-4bb9-8823-41bc94e90065",
"address": "From Memory. 123, North Pole."
},
{
"empId": "1d40b65a-4ef7-4acf-973e-2b327f591629",
"address": "From Memory. 123, North Pole."
}
]
Response code: 200 (OK); Time: 35ms; Content length: 365 bytes
Not only closing the store after every call, you are also calling
Nitrite.builder().openOrCreate()before every insert call, which is actually resetting the memory store before every insert.
Hi, I am not sure I understand the find API for a specific use case.. say I have some JSON as follows...
"tags": [
{
"type": "example"
}
],How do I write a find query that returns documents that have a tag where the key is type or documents that have a type field with the value example?
You have to use
elemMatch filter here.public void testCollectionField() {
Document document = createDocument("name", "John")
.put("tags", new Document[] {
createDocument("type", "example").put("other", "value"),
createDocument("type", "another-example").put("other", "some-other-value")
});
NitriteCollection example = db.getCollection("example");
example.insert(document);
document = createDocument("name", "Jane")
.put("tags", new Document[] {
createDocument("type", "example2").put("other", "value2"),
createDocument("type", "another-example2").put("other", "some-other-value2")
});
example.insert(document);
DocumentCursor cursor = example.find(elemMatch("tags", eq("type", "example")));
for (Document doc : cursor) {
System.out.println(doc);
}
}
nitrite currently does not support indexing on collection field, so for now you have to search un-indexed.
About the datagate
Interested ?
The idea is to re-write it as a separate project (as a library), and offer a standalone version as well
This way, we can allow people who want to build apps with the datagate to integrated it better instead of integrating via the database
Especially when it comest to stuff like access rights and access control and sync policies
I'm also thinking we should look at the sync client and use something more real time like a websocket or SSE instead of polling
Making things a bit faster and have higher data consistency
What do you think ?
Your point about refactoring the sync client, I definitely have it in my mind for the next major release 4.x. But I did not understand your proposal about datagate server. What use case do you have in mind? Care to elaborate? As for me, I was planning to rewrite it using golang as a separate application to squeeze maximum throughput during sync. Implementing it as a library I did not have it in my mind unless there is a valid use case.
The use case is simple : I, as someone who wants to build a mobile application that uses datagate, need integrated authentication and access rights (Ie, I can only see and edit my authorized data), need multiple login tokens (from my different clients), may need to modify data from web/desktop clients directly to the database, etc. In most applications, I'll also need a higher level of consistency for some entities, like invoice serials and stock numbers , etc. So a lot of the time, I'll be running the datagate alongside my application server, and getting data from both. Right now, this means that I'll be integrating via the database and running 2 servers. If the datagate were a library with its own endpoints integrated into my application, this wouldn't be the case, and I have more levels of consistency possible.
@TareqK in its current form you can do the same with datagate application. It's a spring boot application, so you should be good to integrate datagate into your application, provided your application is also a spring boot application. You just have to delete the application file
NitriteDataGate.java from datagate.
How are you planning to build a server component which is framework agnostic? You have to use some kind of framework to expose the endpoints. Moreover, what you are proposing is directly against microservice architecture. You can modify datagate source to fit your access control, but I'll suggest separate your application logic from datagate.
It can be kept agnostic by writing it as a plain servlet or websocket and using it in any servlet compatible environment. I mean we are going to rewrite it using websockets anyway, which means everything we use will be in the javax.ws package which is compatible with a lot of frameworks of varying complexity (spring , jersey, rest easy, spark, etc).
And as for the Microservices paradigm, it shouldn't be forced on users, they should be able to choose anyway.
Here are the setups for a clustered and for a single instance setup mode
Here is the Sequence diagram for how messages will be transferred across multiple clusters to reach all clients subscribed to a collection
This kind of setup, whether single mode or listening to a kafka cluster, is implemented using websockets from the javax.ws package and json-rpc, meaning that it supports all server environments and frameworks that support it. We can provide the websocket class, with callbacks for authentication, session management, authorization, etc, for people to override as they see fit(with of course, a default implementation for us). These things can be passed as parameters in a configuration object or the web.xml(or servlet holder), allowing easy configuration via text or code. I personally believe that doing this, we can get rid of a lot of performance issues(polling is expensive on the client and server side), and increase overall service quality. Additionally, we should publish the json-rpc API, allowing anyone to create a datagate client the way they see fit, including via javascript for a web application (a la pouch db) or for other kinds of devices(IOT comes to mind), allowing a more extensible ecosystem based on nitrite
I know that nitrite is designed to be a simple database, but nitrite datagate, i think , should be something that can be used to build scalable applications