Allow disabling of conversion of cache data

[phxql]

Task List

• Steps to reproduce provided

Steps to Reproduce

1. Write a method which returns a huge map and annotate it with @Cacheable
2. Call that method
3. Call it again

Expected Behaviour

2nd method call is at least as fast as 1st.

Actual Behaviour

2nd method call is a lot slower.

Environment Information

• Operating System: Fedora 31
• Micronaut Version: 1.3.0
• JDK Version: AdoptJDK 11.0.6

Hi,

the DefaultSyncCache uses the conversion service to convert the stuff coming from the actual cache implementation. The DefaultConversionService contains some code which iterates over all the keys from map (because it implements Iterable) and tries to convert them. I'm not sure why it's doing this, but this caused a huge performance hit in our application. We're using the default in-memory caffeine cache.

We worked around by wrapping the huge map in an object which doesn't implement Iterable and now it's fast again.

[graemerocher]

Should be checking if conversion is needed before attempting conversion

[jameskleeh]

@phxql So you understand the root of this behavior. When the map is stored in the cache, any information about its generics are lost. We don't know what the key type or value type of the map is. When we retrieve the map, it is converted to being a Map<String, String> (or whatever the return type is). What we can do is add an option to the @CacheConfig annotation to allow disabling of the conversion.

[alvarosanchez]

What if we would allow to define key and value types ?

[phxql]

@phxql So you understand the root of this behavior. When the map is stored in the cache, any information about its generics are lost. We don't know what the key type or value type of the map is. When we retrieve the map, it is converted to being a Map<String, String> (or whatever the return type is). What we can do is add an option to the @CacheConfig annotation to allow disabling of the conversion.

Fine for me :)

[jameskleeh]

This was implemented in the cache module

[graemerocher]

@jameskleeh so can the issue be closed?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[alvarosanchez]

@jameskleeh can this be closed?

[jameskleeh]

@alvarosanchez It seems the code wasn't ever merged. See https://github.com/micronaut-projects/micronaut-cache/tree/cache_conversion

[alvarosanchez]

@jameskleeh were you done with it? If so, can you please create a PR? Otherwise, what is left?

[erkieh]

We got hit by this issue as well. Would be nice if this got merged.