Support For Creating Index in MongoTemplate based On Repository [DATAMONGO-2190]

[spring-projects-issues]

Ankur Pathak opened DATAMONGO-2190 and commented

I have my mongo repository like this:
|@QueryEntity
@Document
public class User {
    ``@Indexed
    ``private String name;
     
     {{... }
}
}|

I create Index programatically using MongoTemplate like this:

 
mongoOps.indexOps(User.``class``).
  ``ensureIndex(``new Index().on(``"name"``, Direction.ASC));

Why I have to give Index infromation while creating 
Index programtically using Mongo Template?
The below information in above method called is redundant:

new Index().on(``"name"``, Direction.ASC)

as same information is already on my Repositry by:

@Indexed
    ``private String name;

So actual call should be like:

mongoOps.indexOps(User.``class``).``ensureIndex()

We can have above version of ensureIndex to add more Indexes
programatically not specified on Repository.

But version of ensureIndex which takes Index information
from repository and create Indexes is a must.

This way of creating Indexes is also very helpful during tests else ours
test keep on failing with DuplicateKeyExcepton. Please check this
article for reference:
https://dzone.com/articles/spring-integration-tests
I tried this way of this article but if also failing because of API Changes?
Can you provide quick fix or some solution for this use case

---

Affects: 2.1.4 (Lovelace SR4)

Reference URL: https://dzone.com/articles/spring-integration-tests

Issue Links:

• DATAMONGO-2188 Add configuration option to enable/disable automatic index creation

[spring-projects-issues]

Mark Paluch commented

Care to elaborate what you're trying to achieve? Also care to explain what you mean with repository? The description above does not contain a repository interface.

Automatic index creation can be used through either one of the following approaches:

•  Implicit creation as soon as you interact (query, persist, create a repository instance) for the first time with a class that is annotated with @Document and one of our @Indexed annotations.
• Creation on startup: By registering annotated entity classes in MongoMappingContext.initialEntitySet, indexes are created on startup

[spring-projects-issues]

Ankur Pathak commented

Repostorry  =  Annotated Document Class + Corresponding Repository which extends MongoRepository

I have Document User class above and it is indexed on Field name.

Now I drop my collection user with:

template.dropCollection(User.class) 

and it will drop my Index on name Field.along with Collection.

Now I want to create collection:

template.createCollection(User.class)

and it will not create my index on name Field.

I want to create my Index as well:and create it by:

template.indexOps(User.``class``).
  ``ensureIndex(``new Index().on(``"name"``, Direction.ASC));

But to create Index I have to give redundant Information new Index().on(``"name"``, Direction.ASC) because

this information is already present in Document User class by annotation @Indexed.

It will be great if I don't have to give this extra information and have method call like:
template.indexOps(User.``class``).
  ``ensureIndex(``);

and it takes Index information from Document User class(Specified by Index Annotation

like @Indexed, @CompoundIndex etc)

 

Dropping collecation and recreating them along with Index I have to do at the begining of all my Tests. Otherwise my tests fails with DuplicateKeyException. And giving that extra information to create index is mess and will require lot of extra work in CI kind of environment.

 

Whatever way of creating Indexex you have mention that happens automatically or takes redundant information. I am looking for programatic  way of creating Index which:

1. Dont take any extra iformation about indexes
2. Collects index information from Document Class User

 

[spring-projects-issues]

Mark Paluch commented

Thanks for the additional details. Spring Data MongoDB's auto-index creation feature creates indexes on the first access to the entity. That being said, if you remove a collection then Spring Data is not able to determine that this happened and so it does not re-create indexes.

With the next version (see DATAMONGO-2188) we're going to deprecate this feature as it causes a couple of downstream issues that we're not able to resolve. IndexOperations for instance is based on collection names therefore we do not have sufficient context to derive entirely indexes.

You can use IndexResolver to derive index definitions and make sure to create these yourself:

MongoMappingContext mappingContext = new MongoMappingContext();
IndexResolver indexResolver = IndexResolver.create(mappingContext);
Iterable<? extends IndexDefinitionHolder> definitions = indexResolver.resolveIndexFor(MyEntity.class);
IndexOperations indexOperations = template.indexOps(MyEntity.class);
definitions.forEach(indexOperations::ensureIndex);

Closing this ticket as the desired functionality will be available with version 2.2