feat: revamp the schema generation

.github/workflows/main.yml

@@ -16,7 +16,7 @@ jobs:
       - name: Checkout the repo
         uses: actions/checkout@v4
       - name: Setup Gradle
-        uses: gradle/gradle-build-action@v3
+        uses: gradle/actions/setup-gradle@v3
 
       - name: Run tests
         run: ./gradlew check
@@ -38,7 +38,7 @@ jobs:
       - name: Checkout the repo
         uses: actions/checkout@v4
       - name: Setup Gradle
-        uses: gradle/gradle-build-action@v3
+        uses: gradle/actions/setup-gradle@v3
 
       - name: deploy to sonatype snapshots
         run: ./gradlew publish

.github/workflows/pr.yml

@@ -13,7 +13,7 @@ jobs:
       - name: Checkout the repo
         uses: actions/checkout@v4
       - name: Setup Gradle
-        uses: gradle/gradle-build-action@v3
+        uses: gradle/actions/setup-gradle@v3
 
       - name: Run tests
         run: ./gradlew check

.github/workflows/release.yml

@@ -14,7 +14,7 @@ jobs:
       - name: Checkout the repo
         uses: actions/checkout@v4
       - name: Setup Gradle
-        uses: gradle/gradle-build-action@v3
+        uses: gradle/actions/setup-gradle@v3
 
       - name: publish release
         run: ./gradlew publishToSonatype closeAndReleaseSonatypeStagingRepository

build.gradle.kts

@@ -36,6 +36,7 @@ dependencies {
     api(libs.kotlinx.serialization.core)
     implementation(libs.kotlinx.serialization.json)
     implementation(libs.xerial.snappy)
+    implementation(kotlin("reflect"))
     testImplementation(libs.kotest.junit5)
     testImplementation(libs.kotest.core)
     testImplementation(libs.kotest.json)
@@ -44,9 +45,9 @@ dependencies {
 
 tasks.withType<KotlinCompile>().configureEach {
     kotlinOptions.jvmTarget = "1.8"
-    kotlinOptions.apiVersion = "1.6"
-    kotlinOptions.languageVersion = "1.6"
-    kotlinOptions.freeCompilerArgs += "-opt-in=kotlin.RequiresOptIn"
+    kotlinOptions.apiVersion = "1.8"
+    kotlinOptions.languageVersion = "1.8"
+    kotlinOptions.freeCompilerArgs += listOf("-opt-in=kotlinx.serialization.ExperimentalSerializationApi", "-Xcontext-receivers")
 }
 java {
     sourceCompatibility = JavaVersion.VERSION_1_8

src/main/kotlin/com/github/avrokotlin/avro4k/Avro.kt

@@ -9,11 +9,13 @@ import com.github.avrokotlin.avro4k.io.AvroEncodeFormat
 import com.github.avrokotlin.avro4k.io.AvroFormat
 import com.github.avrokotlin.avro4k.io.AvroInputStream
 import com.github.avrokotlin.avro4k.io.AvroOutputStream
-import com.github.avrokotlin.avro4k.schema.schemaFor
+import com.github.avrokotlin.avro4k.schema.AvroSerialDescriptorValueVisitor
+import com.github.avrokotlin.avro4k.serializer.BigDecimalAsStringSerializer
+import com.github.avrokotlin.avro4k.serializer.BigIntegerSerializer
+import com.github.avrokotlin.avro4k.serializer.URLSerializer
 import com.github.avrokotlin.avro4k.serializer.UUIDSerializer
 import kotlinx.serialization.BinaryFormat
 import kotlinx.serialization.DeserializationStrategy
-import kotlinx.serialization.ExperimentalSerializationApi
 import kotlinx.serialization.SerialFormat
 import kotlinx.serialization.SerializationStrategy
 import kotlinx.serialization.descriptors.SerialDescriptor
@@ -31,6 +33,7 @@ import java.nio.ByteBuffer
 import java.nio.file.Files
 import java.nio.file.Path
 import java.nio.file.Paths
+import java.util.concurrent.ConcurrentHashMap
 
 open class AvroInputStreamBuilder<T>(
     private val converter: (Any) -> T,
@@ -60,10 +63,12 @@ open class AvroInputStreamBuilder<T>(
                     val wschema = writerSchema ?: error("Writer schema needs to be supplied for Json format")
                     AvroDecodeFormat.Json(wschema, readerSchema ?: wschema)
                 }
+
                 is AvroFormat.BinaryFormat -> {
                     val wschema = writerSchema ?: error("Writer schema needs to be supplied for Binary format")
                     AvroDecodeFormat.Binary(wschema, readerSchema ?: wschema)
                 }
+
                 is AvroFormat.DataFormat -> AvroDecodeFormat.Data(writerSchema, readerSchema)
             }
         }
@@ -152,11 +157,12 @@ class AvroOutputStreamBuilder<T>(
     }
 }
 
-@OptIn(ExperimentalSerializationApi::class)
 class Avro internal constructor(
     internal val configuration: AvroInternalConfiguration,
     override val serializersModule: SerializersModule,
 ) : SerialFormat, BinaryFormat {
+    internal val schemaCache: MutableMap<SerialDescriptor, Schema> = ConcurrentHashMap()
+
     constructor(
         serializersModule: SerializersModule = defaultModule,
         configuration: AvroConfiguration = AvroConfiguration(),
@@ -167,7 +173,10 @@ class Avro internal constructor(
     companion object {
         val defaultModule =
             SerializersModule {
-                contextual(UUIDSerializer())
+                contextual(UUIDSerializer)
+                contextual(BigDecimalAsStringSerializer)
+                contextual(BigIntegerSerializer)
+                contextual(URLSerializer)
             }
         val default = Avro(defaultModule)
 
@@ -303,14 +312,13 @@ class Avro internal constructor(
         )
     }
 
-    fun schema(descriptor: SerialDescriptor): Schema =
-        schemaFor(
-            serializersModule,
-            descriptor,
-            descriptor.annotations,
-            configuration,
-            mutableMapOf()
-        ).schema()
+    fun schema(descriptor: SerialDescriptor): Schema {
+        return schemaCache.getOrPut(descriptor) {
+            lateinit var output: Schema
+            AvroSerialDescriptorValueVisitor(this) { output = it }.visitValue(descriptor)
+            return output
+        }
+    }
 
     fun <T> schema(serializer: SerializationStrategy<T>): Schema {
         return schema(serializer.descriptor)

src/main/kotlin/com/github/avrokotlin/avro4k/annotations.kt

@@ -2,27 +2,43 @@
 
 package com.github.avrokotlin.avro4k
 
+import com.github.avrokotlin.avro4k.schema.ClassElement
 import com.github.avrokotlin.avro4k.serializer.BigDecimalSerializer
 import kotlinx.serialization.ExperimentalSerializationApi
 import kotlinx.serialization.SerialInfo
-import kotlinx.serialization.descriptors.PrimitiveKind
-import org.apache.avro.LogicalTypes
+import org.apache.avro.LogicalType
 import org.apache.avro.Schema
-import org.apache.avro.SchemaBuilder
 import org.intellij.lang.annotations.Language
+import kotlin.reflect.KClass
 
 /**
- * When annotated on a property, overrides the namespace for the nested record.
+ * When annotated on a property, deeply overrides the namespace for all the nested named types (records, enums and fixed).
+ *
+ * Works with standard classes and inline classes.
  */
 @SerialInfo
 @Target(AnnotationTarget.PROPERTY)
-annotation class AvroNamespaceOverride(val value: String)
+annotation class AvroNamespaceOverride(
+    val value: String,
+)
 
+/**
+ * Adds a property to the Avro schema or field.
+ *
+ * Ignored in inline classes.
+ */
 @SerialInfo
+@Repeatable
 @Target(AnnotationTarget.PROPERTY, AnnotationTarget.CLASS)
 annotation class AvroProp(val key: String, val value: String)
 
+/**
+ * Adds a json property to the Avro schema or field.
+ *
+ * Ignored in inline classes.
+ */
 @SerialInfo
+@Repeatable
 @Target(AnnotationTarget.PROPERTY, AnnotationTarget.CLASS)
 annotation class AvroJsonProp(
     val key: String,
@@ -31,67 +47,33 @@ annotation class AvroJsonProp(
 
 /**
  * To be used with [BigDecimalSerializer] to specify the scale, precision, type and rounding mode of the decimal value.
+ *
+ * Can be used with [AvroFixed] to serialize value as a fixed type.
  */
 @SerialInfo
 @Target(AnnotationTarget.PROPERTY)
-annotation class AvroDecimalLogicalType(
+annotation class AvroDecimal(
     val scale: Int = 2,
     val precision: Int = 8,
-    val schema: LogicalDecimalTypeEnum = LogicalDecimalTypeEnum.BYTES,
 )
 
-enum class LogicalDecimalTypeEnum {
-    BYTES,
-    STRING,
-
-    /**
-     * Fixed requires the field annotated with [AvroFixed]
-     */
-    FIXED,
-}
-
-@SerialInfo
-@Target(AnnotationTarget.PROPERTY)
-annotation class AvroUuidLogicalType
-
-@SerialInfo
-@Target(AnnotationTarget.PROPERTY)
-annotation class AvroTimeLogicalType(val type: LogicalTimeTypeEnum)
-
-enum class LogicalTimeTypeEnum(val kind: PrimitiveKind, val schemaFor: () -> Schema) {
-    DATE(PrimitiveKind.INT, { LogicalTypes.date().addToSchema(SchemaBuilder.builder().intType()) }),
-    TIME_MILLIS(
-        PrimitiveKind.INT,
-        { LogicalTypes.timeMillis().addToSchema(SchemaBuilder.builder().intType()) }
-    ),
-    TIME_MICROS(
-        PrimitiveKind.LONG,
-        { LogicalTypes.timeMicros().addToSchema(SchemaBuilder.builder().longType()) }
-    ),
-    TIMESTAMP_MILLIS(
-        PrimitiveKind.LONG,
-        { LogicalTypes.timestampMillis().addToSchema(SchemaBuilder.builder().longType()) }
-    ),
-    TIMESTAMP_MICROS(
-        PrimitiveKind.LONG,
-        { LogicalTypes.timestampMicros().addToSchema(SchemaBuilder.builder().longType()) }
-    ),
-    LOCAL_TIMESTAMP_MILLIS(
-        PrimitiveKind.LONG,
-        { LogicalTypes.localTimestampMillis().addToSchema(SchemaBuilder.builder().longType()) }
-    ),
-    LOCAL_TIMESTAMP_MICROS(
-        PrimitiveKind.LONG,
-        { LogicalTypes.localTimestampMicros().addToSchema(SchemaBuilder.builder().longType()) }
-    ),
-}
-
+/**
+ * Adds documentation to:
+ * - a record's field
+ * - a record
+ * - an enum
+ *
+ * Ignored in inline classes.
+ */
 @SerialInfo
 @Target(AnnotationTarget.PROPERTY, AnnotationTarget.CLASS)
 annotation class AvroDoc(val value: String)
 
 /**
  * Adds aliases to a field of a record. It helps to allow having different names for the same field for better compatibility when changing a schema.
+ *
+ * Ignored in inline classes.
+ *
  * @param value The aliases for the annotated property. Note that the given aliases won't be changed by the configured [AvroConfiguration.fieldNamingStrategy].
  */
 @SerialInfo
@@ -106,12 +88,44 @@ annotation class AvroAlias(vararg val value: String)
 @Target(AnnotationTarget.PROPERTY)
 annotation class AvroFixed(val size: Int)
 
+/**
+ * Sets the default avro value for a record's field.
+ *
+ * Ignored in inline classes.
+ */
 @SerialInfo
 @Target(AnnotationTarget.PROPERTY)
 annotation class AvroDefault(
     @Language("JSON") val value: String,
 )
 
+/**
+ * This annotation indicates that the annotated enum class should be serialized as an Avro enum with the given default value.
+ *
+ * It must be annotated on an enum class. Otherwise, it will be ignored.
+ */
 @SerialInfo
 @Target(AnnotationTarget.CLASS)
-annotation class AvroEnumDefault(val value: String)
+annotation class AvroEnumDefault(val value: String)
+
+/**
+ * Allows to specify the schema of a property.
+ */
+@SerialInfo
+@Target(AnnotationTarget.PROPERTY)
+annotation class AvroSchema(val value: KClass<out AvroSchemaSupplier>)
+
+interface AvroSchemaSupplier {
+    fun getSchema(classElement: ClassElement): Schema
+}
+
+/**
+ * Allows to specify the logical type applied on the generated schema of a property.
+ */
+@SerialInfo
+@Target(AnnotationTarget.PROPERTY)
+annotation class AvroLogicalType(val value: KClass<out AvroLogicalTypeSupplier>)
+
+interface AvroLogicalTypeSupplier {
+    fun getLogicalType(classElement: ClassElement): LogicalType
+}

src/main/kotlin/com/github/avrokotlin/avro4k/decoder/FromAvroValue.kt

@@ -2,35 +2,17 @@ package com.github.avrokotlin.avro4k.decoder
 
 import kotlinx.serialization.SerializationException
 import org.apache.avro.generic.GenericData
-import org.apache.avro.generic.GenericEnumSymbol
-import org.apache.avro.util.Utf8
 import java.nio.ByteBuffer
 
-interface FromAvroValue<T, R> {
-    fun fromValue(value: T): R
-}
-
-object StringFromAvroValue : FromAvroValue<Any?, String> {
-    override fun fromValue(value: Any?): String {
+object StringFromAvroValue {
+    fun fromValue(value: Any?): String {
         return when (value) {
-            is String -> value
-            is Utf8 -> value.toString()
+            is CharSequence -> value.toString()
             is GenericData.Fixed -> String(value.bytes())
             is ByteArray -> String(value)
-            is CharSequence -> value.toString()
             is ByteBuffer -> String(value.array())
             null -> throw SerializationException("Cannot decode <null> as a string")
             else -> throw SerializationException("Unsupported type for String [is ${value::class.qualifiedName}]")
         }
     }
-}
-
-object EnumFromAvroValue : FromAvroValue<Any, String> {
-    override fun fromValue(value: Any): String {
-        return when (value) {
-            is GenericEnumSymbol<*> -> value.toString()
-            is String -> value
-            else -> value.toString()
-        }
-    }
 }

src/main/kotlin/com/github/avrokotlin/avro4k/decoder/ListDecoder.kt

@@ -72,7 +72,7 @@ class ListDecoder(
     override fun fieldSchema(): Schema = schema.elementType
 
     override fun decodeEnum(enumDescriptor: SerialDescriptor): Int {
-        val symbol = EnumFromAvroValue.fromValue(array[index]!!)
+        val symbol = array[index]!!.toString()
         return (0 until enumDescriptor.elementsCount).find { enumDescriptor.getElementName(it) == symbol } ?: -1
     }
 

src/main/kotlin/com/github/avrokotlin/avro4k/decoder/RecordDecoder.kt

@@ -135,7 +135,7 @@ class RecordDecoder(
     }
 
     override fun decodeEnum(enumDescriptor: SerialDescriptor): Int {
-        val symbol = EnumFromAvroValue.fromValue(fieldValue()!!)
+        val symbol = fieldValue()!!.toString()
         val enumValueByEnumName =
             (0 until enumDescriptor.elementsCount).associateBy { enumDescriptor.getElementName(it) }
 

src/main/kotlin/com/github/avrokotlin/avro4k/encoder/RecordEncoder.kt

@@ -4,7 +4,6 @@ import com.github.avrokotlin.avro4k.AvroInternalConfiguration
 import com.github.avrokotlin.avro4k.ListRecord
 import com.github.avrokotlin.avro4k.Record
 import com.github.avrokotlin.avro4k.schema.extractNonNull
-import com.github.avrokotlin.avro4k.schema.unwrapValueClass
 import kotlinx.serialization.ExperimentalSerializationApi
 import kotlinx.serialization.SerializationException
 import kotlinx.serialization.descriptors.PolymorphicKind
@@ -38,6 +37,10 @@ interface StructureEncoder : FieldEncoder {
     }
 }
 
+@ExperimentalSerializationApi
+internal val SerialDescriptor.unwrapValueClass: SerialDescriptor
+    get() = if (isInline) getElementDescriptor(0) else this
+
 @ExperimentalSerializationApi
 class RecordEncoder(
     private val schema: Schema,