When using kotlinx.serialization outside the context of MongoDB, a format class is available to serialize and deserialize data. E.g. for JSON:

val jsonFormat = Json { ignoreUnknownKeys = false }
jsonFormat.decodeFromString<Person>("""{"name":"John Doe"}""")
jsonFormat.encodeToString(Person(name="Jane Doe"))

When using kotlinx.serialization with the Kotlin-MongoDB-Driver, such a format (e.g. BSON) is not available since it is hidden in the driver implementation.

This has two drawbacks:

1 Serialization tests

Serialization tests answer the question: "Am I inserting the data in the intended format into the database?". E.g. it does matter whether I am storing a date as BSON date or as string. Or I might want to ensure a specific type discriminator field "type" vs. the default "_t" is used.

This is currently only possible to test with starting up a MongoDB test-container, writing the data via the MongoDB client via a typed collection Collection<Person>, and reading it again via a raw collection such as Collection<Document> or Collection<BsonDocument>.

2 Serialize entire queries or aggregations (maybe less important)

Operations on collections such as collection.find or collection.aggregate take BSON and List<BSON> as parameters. BSON documents can either be created via a very cumbersome api such as:

val bson = BsonDocument()
    .append("key1", "value1")
    .append(
        "key2",
        BsonDocument()
            .append("key1", "value1")
            .append("key2", "value2")
    )

collection.find(bson)

or via the handy DSL

val findJohn = Filters.eq(Person::name, "John Doe")
// or
val findJane = Person::name eq "Jane Doe"

collection.find(findJohn)
collection.find(findJane)

Since BSON is very flexible and thus complex, the DSL will probably never cover all use-cases for queries and aggregations.

Exposing a BSON format would allow developers to write queries in a very flexible way. E.g.

@Serializable
class FindByIdQuery(
    @SerialName("_id")
    val id: String,
)

collection.find(bsonFormat.encode(FindByIdQuery("id")))