handle nullable bytearrays and add null values in benchmark

benchmark/README.md

@@ -11,7 +11,8 @@ Each benchmark is executed with the following configuration:
 - Writing to a null output stream
 - All with the exact same schema generated by avro4k
 - Generating a maximum of use cases:
-  - unions (including nullable fields)
+  - nullable fields
+  - unions
   - arrays
   - records
   - enums
@@ -24,15 +25,15 @@ Each benchmark is executed with the following configuration:
 Computer: Macbook air M2
 
 ```
-Benchmark                                    Mode  Cnt      Score      Error  Units  Relative Difference (%)
-Avro4kBenchmark.read                        thrpt    5  20537.185 ±  135.318  ops/s    0.00%
-ApacheAvroReflectBenchmark.read             thrpt    5  20059.982 ±  241.854  ops/s   -2.32%
-Avro4kGenericWithApacheAvroBenchmark.read   thrpt    5   7591.527 ±  172.173  ops/s  -63.03%
-
-Avro4kBenchmark.write                       thrpt    5  41215.703 ± 1274.692  ops/s    0.00%
-ApacheAvroReflectBenchmark.write            thrpt    5  37188.260 ±  115.447  ops/s   -9.74%
-JacksonAvroBenchmark.write                  thrpt    5  30757.363 ± 1557.034  ops/s  -25.39%
-Avro4kGenericWithApacheAvroBenchmark.write  thrpt    5  21305.149 ±  830.640  ops/s  -48.33%
+Benchmark                                    Mode  Cnt      Score       Error  Units  Relative Difference (%)
+Avro4kBenchmark.read                        thrpt    5  21443.935 ±  2215.328  ops/s    0.00%
+ApacheAvroReflectBenchmark.read             thrpt    5  19803.543 ±   485.869  ops/s   -7.64%
+Avro4kGenericWithApacheAvroBenchmark.read   thrpt    5   8836.787 ±   404.874  ops/s  -58.79%
+
+Avro4kBenchmark.write                       thrpt    5  50565.556 ±   849.344  ops/s    0.00%
+ApacheAvroReflectBenchmark.write            thrpt    5  46872.768 ±  2406.622  ops/s   -7.30%
+JacksonAvroBenchmark.write                  thrpt    5  32349.182 ± 10105.111  ops/s  -36.01%
+Avro4kGenericWithApacheAvroBenchmark.write  thrpt    5  27471.887 ±   315.498  ops/s  -45.67%
 ```
 
 > [!WARNING]

benchmark/src/main/kotlin/com/github/avrokotlin/benchmark/ManualProfiling.kt

@@ -5,7 +5,7 @@ internal object ManualProfilingWrite {
     fun main(vararg args: String) {
         Avro4kBenchmark().apply {
             initTestData()
-            for (i in 0 until 100_000) {
+            for (i in 0 until 1_000_000) {
                 if (i % 1_000 == 0) println("Iteration $i")
                 write()
             }
@@ -17,7 +17,7 @@ internal object ManualProfilingRead {
     fun main(vararg args: String) {
         Avro4kBenchmark().apply {
             initTestData()
-            for (i in 0 until 100_000) {
+            for (i in 0 until 1_000_000) {
                 if (i % 1_000 == 0) println("Iteration $i")
                 read()
             }

benchmark/src/main/kotlin/com/github/avrokotlin/benchmark/internal/Clients.kt

@@ -41,31 +41,31 @@ class CharJacksonDeserializer : StdDeserializer<Char>(Char::class.java) {
 
 @Serializable
 internal data class Client(
-    val id: Long = 0,
-    val index: Int = 0,
-    val isActive: Boolean = false,
+    val id: Long,
+    val index: Int,
+    val isActive: Boolean,
     @Serializable(with = BigDecimalAsStringSerializer::class)
     @JsonFormat(shape = JsonFormat.Shape.STRING)
-    val balance: BigDecimal? = null,
-    val picture: ByteArray? = null,
-    val age: Int = 0,
-    val eyeColor: EyeColor? = null,
-    val name: String? = null,
+    val balance: BigDecimal?,
+    val picture: ByteArray?,
+    val age: Int,
+    val eyeColor: EyeColor?,
+    val name: String?,
     @JsonSerialize(using = CharJacksonSerializer::class)
     @JsonDeserialize(using = CharJacksonDeserializer::class)
-    val gender: Char? = null,
-    val company: String? = null,
-    val emails: Array<String> = emptyArray(),
-    val phones: LongArray = LongArray(0),
-    val address: String? = null,
-    val about: String? = null,
+    val gender: Char?,
+    val company: String?,
+    val emails: Array<String>,
+    val phones: LongArray,
+    val address: String?,
+    val about: String?,
     @Serializable(with = LocalDateSerializer::class)
-    val registered: LocalDate? = null,
-    val latitude: Double = 0.0,
-    val longitude: Float = 0.0f,
-    val tags: List<String?> = emptyList(),
-    val partner: Partner,
-    val map: Map<String, String> = emptyMap(),
+    val registered: LocalDate?,
+    val latitude: Double,
+    val longitude: Float,
+    val tags: List<String?>,
+    val partner: Partner?,
+    val map: Map<String, String>,
 )
 
 @Serializable
@@ -80,18 +80,18 @@ sealed interface Partner
 
 @Serializable
 internal class GoodPartner(
-    val id: Long = 0,
-    val name: String? = null,
+    val id: Long,
+    val name: String,
     @Serializable(with = InstantSerializer::class)
-    val since: Instant? = null
+    val since: Instant
 ) : Partner
 
 @Serializable
 internal class BadPartner(
-    val id: Long = 0,
-    val name: String? = null,
+    val id: Long,
+    val name: String,
     @Serializable(with = InstantSerializer::class)
-    val since: Instant? = null
+    val since: Instant
 ) : Partner
 
 @Serializable

benchmark/src/main/kotlin/com/github/avrokotlin/benchmark/internal/ClientsGenerator.kt

@@ -6,55 +6,50 @@ import kotlin.math.absoluteValue
 
 internal object ClientsGenerator {
     fun generate(size: Int) = Clients(
-        clients = buildList {
-            repeat(size) {
-                add(newClient())
+        buildList {
+            repeat(size) { index ->
+                add(Client(
+                    partner = when (RandomUtils.nextInt(4)) {
+                        0 -> GoodPartner(RandomUtils.nextLong(), RandomUtils.randomAlphabetic(30), Instant.ofEpochMilli(RandomUtils.nextLong()))
+                        1 -> BadPartner(RandomUtils.nextLong(), RandomUtils.randomAlphabetic(30), Instant.ofEpochMilli(RandomUtils.nextLong().absoluteValue))
+                        2 -> if (RandomUtils.nextBoolean()) Stranger.KNOWN_STRANGER else Stranger.UNKNOWN_STRANGER
+                        3 -> null
+                        else -> throw IllegalStateException("Unexpected value")
+                    },
+                    id = RandomUtils.nextLong().absoluteValue,
+                    index = RandomUtils.nextInt(0, Int.MAX_VALUE),
+                    isActive = RandomUtils.nextBoolean(),
+                    balance = if (index % 2 == 0) null else RandomUtils.randomBigDecimal(),
+                    picture = if (index % 3 == 0) null else RandomUtils.randomBytes(4048),
+                    age = RandomUtils.nextInt(0, 100),
+                    eyeColor = if (index % 2 == 0) null else EyeColor.entries[RandomUtils.nextInt(3)],
+                    name = if (index % 2 == 0) null else RandomUtils.randomAlphanumeric(20),
+                    gender = if (index % 2 == 0) null else if (RandomUtils.nextBoolean()) 'M' else 'F',
+                    company = if (index % 2 == 0) null else RandomUtils.randomAlphanumeric(20),
+                    emails = RandomUtils.stringArray(RandomUtils.nextInt(5, 10), 10),
+                    phones = RandomUtils.longArray(RandomUtils.nextInt(5, 10)),
+                    address = if (index % 2 == 0) null else RandomUtils.randomAlphanumeric(20),
+                    about = RandomUtils.randomAlphanumeric(20),
+                    registered = if (index % 3 == 0) null else
+                        LocalDate.of(
+                            1900 + RandomUtils.nextInt(110),
+                            1 + RandomUtils.nextInt(12),
+                            1 + RandomUtils.nextInt(28)
+                        ),
+                    latitude = RandomUtils.nextDouble(0.0, 90.0),
+                    longitude = RandomUtils.nextFloat(0.0f, 180.0f),
+                    tags = buildList {
+                        repeat(RandomUtils.nextInt(5, 25)) {
+                            add(if (it % 2 == 0) null else RandomUtils.randomAlphanumeric(10))
+                        }
+                    },
+                    map = buildMap {
+                        repeat(10) {
+                            put(RandomUtils.randomAlphanumeric(10), RandomUtils.randomAlphanumeric(10))
+                        }
+                    }
+                ))
             }
         }
     )
-
-    private fun newClient(): Client {
-        val u = Client(
-            partner = when (RandomUtils.nextInt(3)) {
-                0 -> GoodPartner(RandomUtils.nextLong(), RandomUtils.randomAlphabetic(30), Instant.ofEpochMilli(RandomUtils.nextLong()))
-                1 -> BadPartner(RandomUtils.nextLong(), RandomUtils.randomAlphabetic(30), Instant.ofEpochMilli(RandomUtils.nextLong().absoluteValue))
-                2 -> if (RandomUtils.nextBoolean()) Stranger.KNOWN_STRANGER else Stranger.UNKNOWN_STRANGER
-                else -> throw IllegalStateException("Unexpected value")
-            },
-            id = RandomUtils.nextLong().absoluteValue,
-            index = RandomUtils.nextInt(0, Int.MAX_VALUE),
-            isActive = RandomUtils.nextBoolean(),
-            balance = RandomUtils.randomBigDecimal(),
-            picture = RandomUtils.randomBytes(4048),
-            age = RandomUtils.nextInt(0, 100),
-            eyeColor = EyeColor.entries[RandomUtils.nextInt(3)],
-            name = RandomUtils.randomAlphanumeric(20),
-            gender = if (RandomUtils.nextBoolean()) 'M' else 'F',
-            company = RandomUtils.randomAlphanumeric(20),
-            emails = RandomUtils.stringArray(RandomUtils.nextInt(5, 10), 10),
-            phones = RandomUtils.longArray(RandomUtils.nextInt(5, 10)),
-            address = RandomUtils.randomAlphanumeric(20),
-            about = RandomUtils.randomAlphanumeric(20),
-            registered =
-            LocalDate.of(
-                1900 + RandomUtils.nextInt(110),
-                1 + RandomUtils.nextInt(12),
-                1 + RandomUtils.nextInt(28)
-            ),
-            latitude = RandomUtils.nextDouble(0.0, 90.0),
-            longitude = RandomUtils.nextFloat(0.0f, 180.0f),
-            tags = buildList {
-                repeat(RandomUtils.nextInt(5, 25)) {
-                    add(RandomUtils.randomAlphanumeric(10))
-                }
-            },
-            map = buildMap {
-                repeat(10) {
-                    put(RandomUtils.randomAlphanumeric(10), RandomUtils.randomAlphanumeric(10))
-                }
-            }
-        )
-        return u
-    }
 }
-

src/main/kotlin/com/github/avrokotlin/avro4k/internal/decoder/direct/AbstractAvroDirectDecoder.kt

@@ -18,7 +18,6 @@ import com.github.avrokotlin.avro4k.decodeResolvingInt
 import com.github.avrokotlin.avro4k.decodeResolvingLong
 import com.github.avrokotlin.avro4k.internal.UnexpectedDecodeSchemaError
 import com.github.avrokotlin.avro4k.internal.decoder.AbstractPolymorphicDecoder
-import com.github.avrokotlin.avro4k.internal.decoder.direct.AbstractAvroDirectDecoder.SizeGetter
 import com.github.avrokotlin.avro4k.internal.getElementIndexNullable
 import com.github.avrokotlin.avro4k.internal.isFullNameOrAliasMatch
 import com.github.avrokotlin.avro4k.internal.nonNullSerialName
@@ -46,7 +45,7 @@ internal abstract class AbstractAvroDirectDecoder(
     protected val binaryDecoder: org.apache.avro.io.Decoder,
 ) : AbstractInterceptingDecoder(), UnionDecoder {
     abstract override var currentWriterSchema: Schema
-    private var currentlyReadingCollection: Boolean = false
+    private var previousCollectionSize = -1
 
     override val serializersModule: SerializersModule
         get() = avro.serializersModule
@@ -64,60 +63,21 @@ internal abstract class AbstractAvroDirectDecoder(
             return decodeBytes() as T
         }
 
-        if (deserializer is AbstractCollectionSerializer<*, T, *> && currentWriterSchema.isCollectionOrMap()) {
-            var result: T = deserializer.merge(this, null)
-            with(result.collectionSizeGetter()) {
-                if (result.collectionSize() > 0) {
-                    currentlyReadingCollection = true
-                    var prevSize = result.collectionSize()
-                    while (true) {
-                        result = deserializer.merge(this@AbstractAvroDirectDecoder, result)
-                        val newSize = result.collectionSize()
-                        if (prevSize == newSize) {
-                            break
-                        }
-                        prevSize = newSize
-                    }
-                    currentlyReadingCollection = false
-                }
-            }
-            return result
+        if (deserializer is AbstractCollectionSerializer<*, T, *>) {
+            return decodeCollectionLike(deserializer)
         }
 
         return super<AbstractInterceptingDecoder>.decodeSerializableValue(deserializer)
     }
 
-    private fun Schema.isCollectionOrMap(): Boolean {
-        if (isUnion) {
-            this.types.forEach {
-                if (it.isCollectionOrMap()) {
-                    return true
-                }
-            }
-            return false
-        }
-        return this.type == Schema.Type.ARRAY || this.type == Schema.Type.MAP
-    }
-
-    fun interface SizeGetter<T> {
-        fun T.collectionSize(): Int
-    }
-
-    private fun <T> T.collectionSizeGetter(): SizeGetter<T> {
-        return when (this) {
-            is Collection<*> -> SizeGetter { size }
-            is Map<*, *> -> SizeGetter { size }
-            is Array<*> -> SizeGetter { size }
-            is BooleanArray -> SizeGetter { size }
-            is ByteArray -> SizeGetter { size }
-            is ShortArray -> SizeGetter { size }
-            is IntArray -> SizeGetter { size }
-            is LongArray -> SizeGetter { size }
-            is FloatArray -> SizeGetter { size }
-            is DoubleArray -> SizeGetter { size }
-            is CharArray -> SizeGetter { size }
-            else -> throw SerializationException("Unsupported collection type: ${this?.let { it::class }}")
-        }
+    @OptIn(InternalSerializationApi::class)
+    private fun <T> decodeCollectionLike(deserializer: AbstractCollectionSerializer<*, T, *>): T {
+        var result: T? = null
+        do {
+            result = deserializer.merge(this, result)
+        } while (previousCollectionSize > 0)
+        previousCollectionSize = -1
+        return result!!
     }
 
     override fun beginStructure(descriptor: SerialDescriptor): CompositeDecoder {
@@ -132,7 +92,7 @@ internal abstract class AbstractAvroDirectDecoder(
                 }) {
                     when (it.type) {
                         Schema.Type.ARRAY -> {
-                            AnyValueDecoder { ArrayBlockDirectDecoder(it, decodeFirstBlock = !currentlyReadingCollection, avro, binaryDecoder) }
+                            AnyValueDecoder { ArrayBlockDirectDecoder(it, decodeFirstBlock = previousCollectionSize == -1, { previousCollectionSize = it }, avro, binaryDecoder) }
                         }
 
                         Schema.Type.BYTES -> {
@@ -152,7 +112,7 @@ internal abstract class AbstractAvroDirectDecoder(
                 }) {
                     when (it.type) {
                         Schema.Type.MAP -> {
-                            AnyValueDecoder { MapBlockDirectDecoder(it, decodeFirstBlock = !currentlyReadingCollection, avro, binaryDecoder) }
+                            AnyValueDecoder { MapBlockDirectDecoder(it, decodeFirstBlock = previousCollectionSize == -1, { previousCollectionSize = it }, avro, binaryDecoder) }
                         }
 
                         else -> null

src/main/kotlin/com/github/avrokotlin/avro4k/internal/decoder/direct/CollectionsDirectDecoder.kt

@@ -34,6 +34,7 @@ internal class BytesDirectDecoder(
 internal class ArrayBlockDirectDecoder(
     private val arraySchema: Schema,
     private val decodeFirstBlock: Boolean,
+    private val onCollectionSizeDecoded: (Int) -> Unit,
     avro: Avro,
     binaryDecoder: org.apache.avro.io.Decoder,
 ) : AbstractAvroDirectDecoder(avro, binaryDecoder) {
@@ -44,7 +45,7 @@ internal class ArrayBlockDirectDecoder(
             binaryDecoder.readArrayStart().toInt()
         } else {
             binaryDecoder.arrayNext().toInt()
-        }
+        }.also { onCollectionSizeDecoded(it) }
     }
 
     override fun beginElement(
@@ -65,6 +66,7 @@ internal class ArrayBlockDirectDecoder(
 internal class MapBlockDirectDecoder(
     private val mapSchema: Schema,
     private val decodeFirstBlock: Boolean,
+    private val onCollectionSizeDecoded: (Int) -> Unit,
     avro: Avro,
     binaryDecoder: org.apache.avro.io.Decoder,
 ) : AbstractAvroDirectDecoder(avro, binaryDecoder) {
@@ -75,7 +77,7 @@ internal class MapBlockDirectDecoder(
             binaryDecoder.readMapStart().toInt()
         } else {
             binaryDecoder.mapNext().toInt()
-        }
+        }.also { onCollectionSizeDecoded(it) }
     }
 
     override fun beginElement(

src/main/kotlin/com/github/avrokotlin/avro4k/internal/decoder/generic/AbstractAvroGenericDecoder.kt

@@ -39,7 +39,7 @@ internal abstract class AbstractAvroGenericDecoder : AbstractDecoder(), AvroDeco
         get() = avro.serializersModule
 
     override fun <T> decodeSerializableValue(deserializer: DeserializationStrategy<T>): T {
-        if (deserializer.descriptor == ByteArraySerializer().descriptor) {
+        if (deserializer == ByteArraySerializer()) {
             // fast-path for ByteArray fields, to avoid slow-path with ArrayGenericDecoder
             @Suppress("UNCHECKED_CAST")
             return decodeBytes() as T