Nitrite/Java: Using rocksdb as storage engine

[tmoliveira]

Hello

I am using nitriteDb with rocksdb all is working properly in Linux and MacOS. However, in Windows, if the path to the db contains any special character (just like 'ç' or 'õ') it fails to open the DB. I read that there is a compilation env $WITH_WINDOWS_UTF8_FILENAMES$ that seems to solve this issue but I am not sure how to use it. I also have dig into the java library and not found any option to solve this.

Any tip on this?

Thanks in advance

[anidotnet]

This environment variable is for building the rocksdb from source using cmake tool. Nitrite is using pre-built jar for rocksdb from maven. So facebook can only answer if they build the source using the said flag on or not.

Otherwise, you can build the rockdb jar from source setting this custom flag on and supply it as nitrite dependency in your application.

[anidotnet]

Judging from the cmake file of rockdb here - https://github.com/facebook/rocksdb/blob/1a343bc393b5328c9825896a20fc750fb4de3e1c/CMakeLists.txt#L69

I see the default value of the flag is off. So you need to build the whole thing from source setting the flag on and create your own rockdb jni jar and supply with your application.

[tmoliveira]

Thanks @anidotnet! I will try it out. One more question: Which one do you advise to use as a storage engine of Nitrite, rocksDB or MapDB? Do you have any thought on this? What are the main concerns when using one or another?

Thanks again @anidotnet

[anidotnet]

I would not suggest to use MapDb yet, because its performance is not at per. If you are using nitrite in desktop or android app, go for mvstore, but if you are using it for server side and want to deal with millions of records then go for rocksdb.

We also have a jmh project for nitrite - https://github.com/nitrite/nitrite-jmh You can use it to gauge the performance for your use case and decide accordingly.

[tmoliveira]

Thanks, @anidotnet. I will keep using rocksDB. After testing the three, this one is the one with a better performance/memory usage tradeoff. As you said, I need to deal with millions of records, and rocksDB offer the best suit for this use case.