Lectures

We’ll break down how Kotlin collections are serialized to JSON via kotlinx.serialization: why List and Set turn into arrays [...], how Map<String, V> becomes a JSON object {...}, what happens with non-string keys ( Map<Int, ...> → string keys), and when it’s better to choose the List<Entry> format instead of a map. At the end — selection criteria and common integration mistakes.

We break down two key annotations from kotlinx.serialization that help keep the JSON format stable: @SerialName — to bind a Kotlin property name to the required field name in JSON (compatibility with old files and external APIs), and @Transient — to exclude fields from serialization (cache, debugging, temporary values) with a mandatory safe default. Plus — typical mistakes during refactoring and when reading “old” JSON.

We break down how enum and sealed class turn into JSON via kotlinx.serialization: why it’s important to lock down external values for an enum using @SerialName, why polymorphic serialization needs a discriminator ( classDiscriminator), how to pin the Ok/ Error variants as "ok"/ "error", and why the JSON can differ when serializing via the base type vs a concrete type.

We break down why you need a custom serializer in kotlinx.serialization: how to separate the internal representation (for example, Money as Long cents) from the external JSON contract (the string "12.34"), how to plug in a serializer via @Serializable( with = ...), what KSerializer<T> consists of ( descriptor, serialize, deserialize ), why it is better not to parse money via Double, and how to write fail-fast parsing with SerializationException.

We break down why cyclic references in a Kotlin object model turn JSON serialization into a “graph vs. tree” problem, which often leads to errors like StackOverflowError. We learn how to design a cycle-free “storage model” (the id-instead-of-reference strategy), how to dump state via BudgetDump, and then restore convenient in-memory links using buildState(). Finally, we cover nuances around stable identifiers, one-way models, and the role of @Transient.