Map vs Filter: Performance

[bdkjones]

I'd like to know how the performance of Realm's Map type compares to the performance of just using filter on a List.

For example, suppose I have a model where I have 400,000 FileItem instances (a subclass of Object), each of which represents a file on disk. I need to quickly get the FileItem object for path /some/file/path.txt. Which route is more performant:

1. Create a Map<String: FileItem> property, where the key is the path of each file?

2. Create a List<FileItem> property and then just use .where({ $0.path == somePath })?

General Observation

There's often very little documentation about the best performance path in Realm. There's a few tricks, like not using () on the declaration of collections, and structuring filter closures so that the most performant checks come first (e.g. numbers before strings), but there's lot of situations like the above where I wonder how things work under the hood.

Where the Swift SDK hits Realm Core, is the Map object just a convenient "wrapper" around the same database queries that filter will generate? Or is Map, like Dictionary, actually an optimized data structure for random access by key?

I wish the docs would offer more information about performance and how to maximize it. It's hard to implement a complex app multiple ways to discover performance by testing.

[dmorgereth]

What do you mean by "not using () on the declaration of collections"?