Effective Kotlin Item 52: Consider associating elements to a map
It is not uncommon to have a bigger set of elements, in which we need to find elements by some key many times. It might be:
- A class storing configuration, loaded from some file or files.
- A network repository storing downloaded data.
- An in-memory repository (such repositories are often used for tests).
Those data might represent some list of users, ids, configurations, etc. They are generally fetched as a list, and it is tempting to represent them in our memory the same way:
However, this is rarely the best way to store those elements. Notice how the data we load are used. We most often access an element by some identifier or name (it is typically connected to how we design data to have unique keys in databases). Finding an element in a list has linear complexity (
n is the size of the list, or more concretely, it takes on average
n/2 comparisons to find an element in a list). It is especially problematic for bigger lists because finding each element requires comparing it with many other elements. A better solution to this problem is to use a
Map instead of a
List. Kotlin by default uses a hash map (concretely
LinkedHashMap), and as we described in Item 43: Respect the contract of hashCode, the performance of finding an element when we use the hash map is much better. On the JVM finding an element will take one comparison only. This is thanks to the fact that the size of the used hash map is adjusted to the size of the map itself (given that the
hashCode function is implemented properly).
For example, this is the
InMemoryRepo from before, but implemented to use a map instead of a list:
Most other operations, like modifying or iterate over those data (likely using collection processing methods like
sum, etc.) have more or less the same performance for the standard map and list. But finding an element by key is much faster.
Associating elements to keys
The crucial point is how do we transform from a list to a map and vice versa. For that, we use the
associate* functions. The most common one is
associateBy that builds a map by associating each element to the key specified by the key selector function. This key selector function (typically lambda expression next to
associateBy) might point to the property we use to identify the element.
Notice, that keys in a map must be unique or duplicates are removed. This is why we should only associate by a unique identifier (to group by something that is not unique, use
To transform from the map to list instead, you can use the
values property of
Here are some sample usages of the technique:
This technique is important, but it is not for all cases. It is more useful when we need to access those elements frequently. This is why it is especially important on the backend, where those collections might be accessed many times per second. It is not so important on the frontend (by that I mean also Android or iOS) where a user will access this repository at most a few times. We also need to remember that mapping from a list to a map takes some time as well, so if we do it often, it might hurt our performance as well.