Support for encoding/decoding a sequence of values (#20)

* Support for streaming via Reader and Appendable
* Handle Microsoft Excel's insistence on using a byte order marker
* Cleaning up new unit tests for FetchSourceTest
* Removed commented debugging println's for FetchSource
* Sequence encoding and decoding
* Streaming serialization bug fix
* Cleanup formatting
* Refactor asynchronous API to `CsvRecordReader` and `CsvRecordWriter`
* Fix blocking initialization of readers

---------

Co-authored-by: Sven Obser <[email protected]>

Author: UnknownJoe796

gradle/libs.versions.toml

@@ -2,13 +2,15 @@
 dokka = "1.9.20"
 junit-jupiter = "5.11.0"
 kotlin = "2.1.0"
+kotlinx-coroutines = "1.9.0"
 kotlinx-serialization-core = "1.7.3"
 nexus-publish = "0.4.0"
 nexus-staging = "0.30.0"
 researchgate-release = "3.0.2"
 
 [libraries]
 junit-jupiter = { module = "org.junit.jupiter:junit-jupiter", version.ref = "junit-jupiter" }
+kotlinx-coroutines-test = { module = "org.jetbrains.kotlinx:kotlinx-coroutines-test", version.ref = "kotlinx-coroutines" }
 kotlinx-serialization-core = { module = "org.jetbrains.kotlinx:kotlinx-serialization-core", version.ref = "kotlinx-serialization-core" }
 
 [plugins]

library/build.gradle.kts

@@ -15,6 +15,7 @@ dependencies {
 
     testImplementation(kotlin("test-junit5"))
     testImplementation(libs.junit.jupiter)
+    testImplementation(libs.kotlinx.coroutines.test)
 }
 
 kotlin {

library/src/main/kotlin/kotlinx/serialization/csv/CsvRecordReader.kt

@@ -0,0 +1,48 @@
+package kotlinx.serialization.csv
+
+import kotlinx.serialization.ExperimentalSerializationApi
+import kotlinx.serialization.KSerializer
+import kotlinx.serialization.builtins.ListSerializer
+import kotlinx.serialization.csv.decode.CsvReader
+import kotlinx.serialization.csv.decode.FetchSource
+import kotlinx.serialization.csv.decode.RecordListCsvDecoder
+import kotlinx.serialization.encoding.CompositeDecoder.Companion.DECODE_DONE
+import java.io.Reader
+
+/**
+ * Record reader that allows reading CSV line-by-line.
+ */
+interface CsvRecordReader<T : Any> : Iterator<T> {
+    /**
+     * Read next record
+     */
+    fun read(): T? = if (hasNext()) next() else null
+}
+
+/**
+ * Parse CSV line-by-line from the given [input].
+ *
+ * @param deserializer The deserializer used to parse the given CSV string.
+ * @param input The CSV reader to parse. This function *does not close the reader*.
+ */
+@ExperimentalSerializationApi
+fun <T : Any> Csv.recordReader(deserializer: KSerializer<T>, input: Reader): CsvRecordReader<T> {
+    val decoder = RecordListCsvDecoder(
+        csv = this,
+        reader = CsvReader(FetchSource(input), config)
+    )
+    val listDescriptor = ListSerializer(deserializer).descriptor
+    var previousValue: T? = null
+
+    return object : CsvRecordReader<T> {
+        override fun hasNext(): Boolean =
+            decoder.decodeElementIndex(listDescriptor) != DECODE_DONE
+
+        override fun next(): T {
+            val index = decoder.decodeElementIndex(listDescriptor)
+            return decoder.decodeSerializableElement(listDescriptor, index, deserializer, previousValue).also {
+                previousValue = it
+            }
+        }
+    }
+}

library/src/main/kotlin/kotlinx/serialization/csv/CsvRecordWriter.kt

@@ -0,0 +1,34 @@
+package kotlinx.serialization.csv
+
+import kotlinx.serialization.ExperimentalSerializationApi
+import kotlinx.serialization.KSerializer
+import kotlinx.serialization.csv.encode.CsvWriter
+import kotlinx.serialization.csv.encode.RecordListCsvEncoder
+
+/**
+ * Record writer that allows writing CSV line by line.
+ */
+fun interface CsvRecordWriter<T : Any> {
+    /**
+     * Write next record.
+     */
+    fun write(record: T)
+}
+
+/**
+ * Create [CsvRecordWriter] that allows writing CSV line-by-line.
+ *
+ * @param serializer The serializer used to serialize the given object.
+ * @param output The output where the CSV will be written.
+ */
+@ExperimentalSerializationApi
+fun <T : Any> Csv.recordWriter(serializer: KSerializer<T>, output: Appendable): CsvRecordWriter<T> {
+    val encoder = RecordListCsvEncoder(
+        csv = this,
+        writer = CsvWriter(output, config)
+    )
+
+    return CsvRecordWriter {
+        encoder.encodeSerializableValue(serializer, it)
+    }
+}

library/src/main/kotlin/kotlinx/serialization/csv/decode/CsvReader.kt

@@ -5,7 +5,17 @@ import kotlinx.serialization.csv.config.CsvConfig
 /**
  * Reader that parses CSV input.
  */
-internal class CsvReader(private val source: Source, private val config: CsvConfig) {
+internal class CsvReader(source: Source, private val config: CsvConfig) {
+
+    private val source: Source by lazy {
+        source.also {
+            // Skip Microsoft Excel's byte order marker, should it appear.
+            // This has to happen lazily to avoid blocking read calls during the initialization of the CsvReader.
+            if (source.peek() == '\uFEFF') {
+                source.read()
+            }
+        }
+    }
 
     val offset
         get() = source.offset
@@ -21,11 +31,6 @@ internal class CsvReader(private val source: Source, private val config: CsvConf
 
     private var marks = arrayListOf<Int>()
 
-    init {
-        // Skip Microsoft Excel's byte order marker, should it appear:
-        read("\uFEFF")
-    }
-
     /**
      * Read value in the next column.
      */

library/src/main/kotlin/kotlinx/serialization/csv/decode/FetchSource.kt

@@ -1,5 +1,6 @@
 package kotlinx.serialization.csv.decode
 
+import java.io.EOFException
 import java.io.Reader
 
 internal class FetchSource(
@@ -20,18 +21,27 @@ internal class FetchSource(
     override var offset: Int = 0
         private set
 
-    private var next: Char? = getChar()
+    private var queue = ArrayList<Char>(2048)
+    private var marks = ArrayList<Int>(2048)
+    private var queueOffset = 0
+
+    private var next: Char? = null
+        get() {
+            if (field == null && nextPosition == 0) {
+                // Reading first char has to happen lazily to avoid blocking read calls
+                // during the initialization of the FetchSource.
+                field = getChar()
+            }
+            return field
+        }
+
     private fun nextChar(): Char {
-        val n = next ?: throw IllegalStateException("Out of characters")
+        val nextChar = next ?: throw EOFException("No more characters to read.")
         next = getChar()
         nextPosition++
-        return n
+        return nextChar
     }
 
-    private var queue = ArrayList<Char>(2048)
-    private var marks = ArrayList<Int>(2048)
-    private var queueOffset = 0
-
     override fun canRead(): Boolean = offset <= nextPosition
 
     override fun read(): Char? {

library/src/main/kotlin/kotlinx/serialization/csv/encode/CsvWriter.kt

@@ -9,7 +9,7 @@ import kotlinx.serialization.csv.config.QuoteMode
  * To write one CSV record, call [beginRecord], followed by multiple calls to [printColumn] and
  * finally call [endRecord] to finish the record.
  */
-internal class CsvWriter(private val sb: Appendable, private val config: CsvConfig) {
+internal class CsvWriter(private val output: Appendable, private val config: CsvConfig) {
 
     var isFirstRecord = true
     private var isFirstColumn = true
@@ -20,7 +20,7 @@ internal class CsvWriter(private val sb: Appendable, private val config: CsvConf
      */
     fun beginRecord() {
         if (!isFirstRecord) {
-            sb.append(config.recordSeparator)
+            output.append(config.recordSeparator)
         }
     }
 
@@ -64,19 +64,19 @@ internal class CsvWriter(private val sb: Appendable, private val config: CsvConf
                 escapeCharacters = "$escapeChar$delimiter$quoteChar$recordSeparator",
                 escapeChar = escapeChar
             )
-            sb.append(escapedValue)
+            output.append(escapedValue)
         } else if (mode == WriteMode.QUOTED || mode == WriteMode.ESCAPED) {
             val escapedValue = value.replace("$quoteChar", "$quoteChar$quoteChar")
-            sb.append(quoteChar).append(escapedValue).append(quoteChar)
+            output.append(quoteChar).append(escapedValue).append(quoteChar)
         } else {
-            sb.append(value)
+            output.append(value)
         }
     }
 
     /** End the current column (which writes the column delimiter). */
     private fun nextColumn() {
         if (!isFirstColumn) {
-            sb.append(config.delimiter)
+            output.append(config.delimiter)
         }
         isFirstColumn = false
     }