kscience/dataforge-core
7
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases 12 Wiki Activity

DataSet flow to sequence

Browse Source
This commit is contained in:
Alexander Nozik 2022-04-24 09:57:33 +03:00
parent eaa9d40d60
commit 77857289f0
No known key found for this signature in database
GPG Key ID: F7FCF2DD25C71357
15 changed files with 68 additions and 70 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
CHANGELOG.md
View File

@ -11,6 +11,7 @@
- KTor 2.0
- DataTree `items` call is blocking.
- DataSet `getData` is no longer suspended and renamed to `get`
- DataSet operates with sequences of data instead of flows
### Deprecated

9
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/actions/MapAction.kt
View File

@ -1,7 +1,6 @@
package space.kscience.dataforge.actions
import kotlinx.coroutines.CoroutineScope
import kotlinx.coroutines.flow.map
import kotlinx.coroutines.launch
import space.kscience.dataforge.data.*
import space.kscience.dataforge.meta.Meta
@ -50,7 +49,7 @@ internal class MapAction<in T : Any, out R : Any>(
meta: Meta,
scope: CoroutineScope?,
): DataSet<R> {
suspend fun mapOne(data: NamedData<T>): NamedData<R> {
fun mapOne(data: NamedData<T>): NamedData<R> {
// Creating a new environment for action using **old** name, old meta and task meta
val env = ActionEnv(data.name, data.meta, meta)
@ -75,16 +74,16 @@ internal class MapAction<in T : Any, out R : Any>(
return newData.named(newName)
}
val flow = dataSet.flowData().map(::mapOne)
val sequence = dataSet.dataSequence().map(::mapOne)
return ActiveDataTree(outputType) {
populateWith(flow)
populateWith(sequence)
scope?.launch {
dataSet.updates.collect { name ->
//clear old nodes
remove(name)
//collect new items
populateWith(dataSet.flowChildren(name).map(::mapOne))
populateWith(dataSet.children(name).map(::mapOne))
}
}
}

3
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/actions/ReduceAction.kt
View File

@ -3,7 +3,6 @@ package space.kscience.dataforge.actions
import kotlinx.coroutines.CoroutineScope
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.flow
import kotlinx.coroutines.flow.fold
import space.kscience.dataforge.data.*
import space.kscience.dataforge.meta.Meta
import space.kscience.dataforge.meta.MutableMeta
@ -84,7 +83,7 @@ internal class ReduceAction<T : Any, R : Any>(
override fun CoroutineScope.transform(set: DataSet<T>, meta: Meta, key: Name): Flow<NamedData<R>> = flow {
ReduceGroupBuilder<T, R>(inputType, this@transform, meta).apply(action).buildGroups(set).forEach { group ->
val dataFlow: Map<Name, Data<T>> = group.set.flowData().fold(HashMap()) { acc, value ->
val dataFlow: Map<Name, Data<T>> = group.set.dataSequence().fold(HashMap()) { acc, value ->
acc.apply {
acc[value.name] = value.data
}

12
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/actions/SplitAction.kt
View File

@ -2,10 +2,6 @@ package space.kscience.dataforge.actions
import kotlinx.coroutines.CoroutineScope
import kotlinx.coroutines.FlowPreview
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.asFlow
import kotlinx.coroutines.flow.flatMapConcat
import kotlinx.coroutines.flow.map
import kotlinx.coroutines.launch
import space.kscience.dataforge.data.*
import space.kscience.dataforge.meta.Laminate
@ -58,14 +54,14 @@ internal class SplitAction<T : Any, R : Any>(
scope: CoroutineScope?,
): DataSet<R> {
suspend fun splitOne(data: NamedData<T>): Flow<NamedData<R>> {
fun splitOne(data: NamedData<T>): Sequence<NamedData<R>> {
val laminate = Laminate(data.meta, meta)
val split = SplitBuilder<T, R>(data.name, data.meta).apply(action)
// apply individual fragment rules to result
return split.fragments.entries.asFlow().map { (fragmentName, rule) ->
return split.fragments.entries.asSequence().map { (fragmentName, rule) ->
val env = SplitBuilder.FragmentRule<T, R>(fragmentName, laminate.toMutableMeta()).apply(rule)
//data.map<R>(outputType, meta = env.meta) { env.result(it) }.named(fragmentName)
@OptIn(DFInternal::class) Data(outputType, meta = env.meta, dependencies = listOf(data)) {
@ -75,13 +71,13 @@ internal class SplitAction<T : Any, R : Any>(
}
return ActiveDataTree<R>(outputType) {
populateWith(dataSet.flowData().flatMapConcat(transform = ::splitOne))
populateWith(dataSet.dataSequence().flatMap (transform = ::splitOne))
scope?.launch {
dataSet.updates.collect { name ->
//clear old nodes
remove(name)
//collect new items
populateWith(dataSet.flowChildren(name).flatMapConcat(transform = ::splitOne))
populateWith(dataSet.children(name).flatMap(transform = ::splitOne))
}
}
}

22
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/DataSet.kt
View File

@ -1,7 +1,9 @@
package space.kscience.dataforge.data
import kotlinx.coroutines.*
import kotlinx.coroutines.flow.*
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.emptyFlow
import kotlinx.coroutines.flow.mapNotNull
import space.kscience.dataforge.data.Data.Companion.TYPE_OF_NOTHING
import space.kscience.dataforge.meta.Meta
import space.kscience.dataforge.meta.set
@ -23,7 +25,7 @@ public interface DataSet<out T : Any> {
/**
* Traverse this provider or its child. The order is not guaranteed.
*/
public fun flowData(): Flow<NamedData<T>>
public fun dataSequence(): Sequence<NamedData<T>>
/**
* Get data with given name.
@ -34,8 +36,8 @@ public interface DataSet<out T : Any> {
/**
* Get a snapshot of names of top level children of given node. Empty if node does not exist or is a leaf.
*/
public suspend fun listTop(prefix: Name = Name.EMPTY): List<Name> =
flowData().map { it.name }.filter { it.startsWith(prefix) && (it.length == prefix.length + 1) }.toList()
public fun listTop(prefix: Name = Name.EMPTY): List<Name> =
dataSequence().map { it.name }.filter { it.startsWith(prefix) && (it.length == prefix.length + 1) }.toList()
// By default, traverses the whole tree. Could be optimized in descendants
public companion object {
@ -50,13 +52,15 @@ public interface DataSet<out T : Any> {
//private val nothing: Nothing get() = error("this is nothing")
override fun flowData(): Flow<NamedData<Nothing>> = emptyFlow()
override fun dataSequence(): Sequence<NamedData<Nothing>> = emptySequence()
override fun get(name: Name): Data<Nothing>? = null
}
}
}
public operator fun <T: Any> DataSet<T>.get(name:String): Data<T>? = get(name.parseAsName())
public interface ActiveDataSet<T : Any> : DataSet<T> {
/**
* A flow of updated item names. Updates are propagated in a form of [Flow] of names of updated nodes.
@ -72,8 +76,8 @@ public val <T : Any> DataSet<T>.updates: Flow<Name> get() = if (this is ActiveDa
/**
* Flow all data nodes with names starting with [branchName]
*/
public fun <T : Any> DataSet<T>.flowChildren(branchName: Name): Flow<NamedData<T>> =
this@flowChildren.flowData().filter {
public fun <T : Any> DataSet<T>.children(branchName: Name): Sequence<NamedData<T>> =
this@children.dataSequence().filter {
it.name.startsWith(branchName)
}
@ -81,7 +85,7 @@ public fun <T : Any> DataSet<T>.flowChildren(branchName: Name): Flow<NamedData<T
* Start computation for all goals in data node and return a job for the whole node
*/
public fun <T : Any> DataSet<T>.startAll(coroutineScope: CoroutineScope): Job = coroutineScope.launch {
flowData().map {
dataSequence().map {
it.launch(this@launch)
}.toList().joinAll()
}
@ -89,7 +93,7 @@ public fun <T : Any> DataSet<T>.startAll(coroutineScope: CoroutineScope): Job =
public suspend fun <T : Any> DataSet<T>.join(): Unit = coroutineScope { startAll(this).join() }
public suspend fun DataSet<*>.toMeta(): Meta = Meta {
flowData().collect {
dataSequence().forEach {
if (it.name.endsWith(DataSet.META_KEY)) {
set(it.name, it.meta)
} else {

10
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/DataSetBuilder.kt
View File

@ -30,7 +30,7 @@ public interface DataSetBuilder<in T : Any> {
}
//Set new items
dataSet.flowData().collect {
dataSet.dataSequence().forEach {
data(name + it.name, it.data)
}
}
@ -148,7 +148,7 @@ public suspend inline fun <reified T : Any> DataSetBuilder<T>.static(
*/
@DFExperimental
public suspend fun <T : Any> DataSetBuilder<T>.populateFrom(tree: DataSet<T>): Unit = coroutineScope {
tree.flowData().collect {
tree.dataSequence().forEach {
//TODO check if the place is occupied
data(it.name, it.data)
}
@ -159,3 +159,9 @@ public suspend fun <T : Any> DataSetBuilder<T>.populateWith(flow: Flow<NamedData
data(it.name, it.data)
}
}
public suspend fun <T : Any> DataSetBuilder<T>.populateWith(sequence: Sequence<NamedData<T>>) {
sequence.forEach {
data(it.name, it.data)
}
}

8
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/DataTree.kt
View File

@ -43,18 +43,18 @@ public interface DataTree<out T : Any> : DataSet<T> {
override val meta: Meta get() = items[META_ITEM_NAME_TOKEN]?.meta ?: Meta.EMPTY
override fun flowData(): Flow<NamedData<T>> = flow {
override fun dataSequence(): Sequence<NamedData<T>> = sequence {
items.forEach { (token, childItem: DataTreeItem<T>) ->
if (!token.body.startsWith("@")) {
when (childItem) {
is DataTreeItem.Leaf -> emit(childItem.data.named(token.asName()))
is DataTreeItem.Node -> emitAll(childItem.tree.flowData().map { it.named(token + it.name) })
is DataTreeItem.Leaf -> yield(childItem.data.named(token.asName()))
is DataTreeItem.Node -> yieldAll(childItem.tree.dataSequence().map { it.named(token + it.name) })
}
}
}
}
override suspend fun listTop(prefix: Name): List<Name> =
override fun listTop(prefix: Name): List<Name> =
getItem(prefix).tree?.items?.keys?.map { prefix + it } ?: emptyList()
override fun get(name: Name): Data<T>? = when (name.length) {

2
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/GroupRule.kt
View File

@ -43,7 +43,7 @@ public interface GroupRule {
): Map<String, DataSet<T>> {
val map = HashMap<String, ActiveDataTree<T>>()
set.flowData().collect { data ->
set.dataSequence().forEach { data ->
val tagValue = data.meta[key]?.string ?: defaultTagValue
map.getOrPut(tagValue) { ActiveDataTree(set.dataType) }.data(data.name, data.data)
}

2
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/StaticDataTree.kt
View File

@ -54,7 +54,7 @@ internal class StaticDataTree<T : Any>(
set(name, DataTreeItem.Node(dataSet))
} else {
coroutineScope {
dataSet.flowData().collect {
dataSet.dataSequence().forEach {
data(name + it.name, it.data)
}
}

9
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/dataFilter.kt
View File

@ -24,8 +24,8 @@ public fun <T : Any> DataSet<T>.filter(
override val meta: Meta get() = this@filter.meta
override fun flowData(): Flow<NamedData<T>> =
this@filter.flowData().filter { predicate(it.name, it.data) }
override fun dataSequence(): Sequence<NamedData<T>> =
this@filter.dataSequence().filter { predicate(it.name, it.data) }
override fun get(name: Name): Data<T>? = this@filter.get(name)?.takeIf {
predicate(name, it)
@ -48,7 +48,8 @@ else object : ActiveDataSet<T> {
override val meta: Meta get() = this@withNamePrefix.meta
override fun flowData(): Flow<NamedData<T>> = this@withNamePrefix.flowData().map { it.data.named(prefix + it.name) }
override fun dataSequence(): Sequence<NamedData<T>> =
this@withNamePrefix.dataSequence().map { it.data.named(prefix + it.name) }
override fun get(name: Name): Data<T>? =
name.removeHeadOrNull(name)?.let { this@withNamePrefix.get(it) }
@ -66,7 +67,7 @@ public fun <T : Any> DataSet<T>.branch(branchName: Name): DataSet<T> = if (branc
override val meta: Meta get() = this@branch.meta
override fun flowData(): Flow<NamedData<T>> = this@branch.flowData().mapNotNull {
override fun dataSequence(): Sequence<NamedData<T>> = this@branch.dataSequence().mapNotNull {
it.name.removeHeadOrNull(branchName)?.let { name ->
it.data.named(name)
}

32
dataforge-data/src/commonMain/kotlin/space/kscience/dataforge/data/dataTransform.kt
View File

@ -1,9 +1,7 @@
package space.kscience.dataforge.data
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.fold
import kotlinx.coroutines.flow.map
import kotlinx.coroutines.flow.toList
import space.kscience.dataforge.meta.Meta
import space.kscience.dataforge.meta.MutableMeta
import space.kscience.dataforge.meta.seal
@ -100,11 +98,11 @@ public inline fun <K, T : Any, reified R : Any> Map<K, Data<T>>.reduceToData(
* Transform a [Flow] of [NamedData] to a single [Data].
*/
@DFInternal
public suspend fun <T : Any, R : Any> Flow<NamedData<T>>.reduceToData(
public inline fun <T : Any, R : Any> Sequence<NamedData<T>>.reduceToData(
outputType: KType,
coroutineContext: CoroutineContext = EmptyCoroutineContext,
meta: Meta = Meta.EMPTY,
transformation: suspend (Flow<NamedData<T>>) -> R,
crossinline transformation: suspend (Sequence<NamedData<T>>) -> R,
): Data<R> = Data(
outputType,
meta,
@ -115,10 +113,10 @@ public suspend fun <T : Any, R : Any> Flow<NamedData<T>>.reduceToData(
}
@OptIn(DFInternal::class)
public suspend inline fun <T : Any, reified R : Any> Flow<NamedData<T>>.reduceToData(
public inline fun <T : Any, reified R : Any> Sequence<NamedData<T>>.reduceToData(
coroutineContext: CoroutineContext = EmptyCoroutineContext,
meta: Meta = Meta.EMPTY,
noinline transformation: suspend (Flow<NamedData<T>>) -> R,
crossinline transformation: suspend (Sequence<NamedData<T>>) -> R,
): Data<R> = reduceToData(typeOf<R>(), coroutineContext, meta) {
transformation(it)
}
@ -126,15 +124,15 @@ public suspend inline fun <T : Any, reified R : Any> Flow<NamedData<T>>.reduceTo
/**
* Fold a flow of named data into a single [Data]
*/
public suspend inline fun <T : Any, reified R : Any> Flow<NamedData<T>>.foldToData(
public inline fun <T : Any, reified R : Any> Sequence<NamedData<T>>.foldToData(
initial: R,
coroutineContext: CoroutineContext = EmptyCoroutineContext,
meta: Meta = Meta.EMPTY,
noinline block: suspend (result: R, data: NamedData<T>) -> R,
crossinline block: suspend (result: R, data: NamedData<T>) -> R,
): Data<R> = reduceToData(
coroutineContext, meta
) {
it.fold(initial, block)
it.fold(initial) { acc, t -> block(acc, t) }
}
//DataSet operations
@ -147,7 +145,7 @@ public suspend fun <T : Any, R : Any> DataSet<T>.map(
block: suspend (T) -> R,
): DataTree<R> = DataTree<R>(outputType) {
populateWith(
flowData().map {
dataSequence().map {
val newMeta = it.meta.toMutableMeta().apply(metaTransform).seal()
Data(outputType, newMeta, coroutineContext, listOf(it)) {
block(it.await())
@ -165,20 +163,20 @@ public suspend inline fun <T : Any, reified R : Any> DataSet<T>.map(
public suspend fun <T : Any> DataSet<T>.forEach(block: suspend (NamedData<T>) -> Unit) {
contract { callsInPlace(block, InvocationKind.EXACTLY_ONCE) }
flowData().collect {
dataSequence().forEach {
block(it)
}
}
public suspend inline fun <T : Any, reified R : Any> DataSet<T>.reduceToData(
public inline fun <T : Any, reified R : Any> DataSet<T>.reduceToData(
coroutineContext: CoroutineContext = EmptyCoroutineContext,
meta: Meta = Meta.EMPTY,
noinline transformation: suspend (Flow<NamedData<T>>) -> R,
): Data<R> = flowData().reduceToData(coroutineContext, meta, transformation)
crossinline transformation: suspend (Sequence<NamedData<T>>) -> R,
): Data<R> = dataSequence().reduceToData(coroutineContext, meta, transformation)
public suspend inline fun <T : Any, reified R : Any> DataSet<T>.foldToData(
public inline fun <T : Any, reified R : Any> DataSet<T>.foldToData(
initial: R,
coroutineContext: CoroutineContext = EmptyCoroutineContext,
meta: Meta = Meta.EMPTY,
noinline block: suspend (result: R, data: NamedData<T>) -> R,
): Data<R> = flowData().foldToData(initial, coroutineContext, meta, block)
crossinline block: suspend (result: R, data: NamedData<T>) -> R,
): Data<R> = dataSequence().foldToData(initial, coroutineContext, meta, block)

7
dataforge-data/src/jvmMain/kotlin/space/kscience/dataforge/data/select.kt
View File

@ -2,7 +2,6 @@ package space.kscience.dataforge.data
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.filter
import kotlinx.coroutines.flow.map
import space.kscience.dataforge.meta.Meta
import space.kscience.dataforge.misc.DFExperimental
import space.kscience.dataforge.names.Name
@ -45,19 +44,19 @@ public fun <R : Any> DataSet<*>.select(
&& (namePattern == null || name.matches(namePattern))
&& filter(name, datum.meta)
override fun flowData(): Flow<NamedData<R>> = this@select.flowData().filter {
override fun dataSequence(): Sequence<NamedData<R>> = this@select.dataSequence().filter {
checkDatum(it.name, it.data)
}.map {
@Suppress("UNCHECKED_CAST")
it as NamedData<R>
}
override fun get(name: Name): Data<R>? = this@select.get(name)?.let { datum ->
override fun get(name: Name): Data<R>? = this@select[name]?.let { datum ->
if (checkDatum(name, datum)) datum.castOrNull(type) else null
}
override val updates: Flow<Name> = this@select.updates.filter {
val datum = this@select.get(it) ?: return@filter false
val datum = this@select[it] ?: return@filter false
checkDatum(it, datum)
}
}

6
dataforge-workspace/src/commonMain/kotlin/space/kscience/dataforge/workspace/TaskResult.kt
View File

@ -1,7 +1,5 @@
package space.kscience.dataforge.workspace
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.map
import space.kscience.dataforge.data.DataSet
import space.kscience.dataforge.meta.Meta
import space.kscience.dataforge.names.Name
@ -25,7 +23,7 @@ public interface TaskResult<out T : Any> : DataSet<T> {
*/
public val taskMeta: Meta
override fun flowData(): Flow<TaskData<T>>
override fun dataSequence(): Sequence<TaskData<T>>
override fun get(name: Name): TaskData<T>?
}
@ -36,7 +34,7 @@ private class TaskResultImpl<out T : Any>(
override val taskMeta: Meta,
) : TaskResult<T>, DataSet<T> by dataSet {
override fun flowData(): Flow<TaskData<T>> = dataSet.flowData().map {
override fun dataSequence(): Sequence<TaskData<T>> = dataSet.dataSequence().map {
workspace.wrapData(it, it.name, taskName, taskMeta)
}

7
dataforge-workspace/src/jvmTest/kotlin/space/kscience/dataforge/workspace/DataPropagationTest.kt
View File

@ -1,6 +1,5 @@
package space.kscience.dataforge.workspace
import kotlinx.coroutines.flow.single
import kotlinx.coroutines.runBlocking
import space.kscience.dataforge.context.Context
import space.kscience.dataforge.context.PluginFactory
@ -16,7 +15,7 @@ class DataPropagationTestPlugin : WorkspacePlugin() {
val allData by task<Int> {
val selectedData = workspace.data.select<Int>()
val result: Data<Int> = selectedData.flowData().foldToData(0) { result, data ->
val result: Data<Int> = selectedData.dataSequence().foldToData(0) { result, data ->
result + data.await()
}
data("result", result)
@ -58,7 +57,7 @@ class DataPropagationTest {
fun testAllData() {
runBlocking {
val node = testWorkspace.produce("Test.allData")
assertEquals(4950, node.flowData().single().await())
assertEquals(4950, node.dataSequence().single().await())
}
}
@ -66,7 +65,7 @@ class DataPropagationTest {
fun testSingleData() {
runBlocking {
val node = testWorkspace.produce("Test.singleData")
assertEquals(12, node.flowData().single().await())
assertEquals(12, node.dataSequence().single().await())
}
}
}

8
dataforge-workspace/src/jvmTest/kotlin/space/kscience/dataforge/workspace/SimpleWorkspaceTest.kt
View File

@ -2,8 +2,6 @@
package space.kscience.dataforge.workspace
import kotlinx.coroutines.flow.first
import kotlinx.coroutines.flow.single
import kotlinx.coroutines.runBlocking
import org.junit.jupiter.api.Timeout
import space.kscience.dataforge.context.*
@ -161,7 +159,7 @@ class SimpleWorkspaceTest {
fun testWorkspace() {
runBlocking {
val node = workspace.runBlocking("sum")
val res = node.flowData().single()
val res = node.dataSequence().single()
assertEquals(328350, res.await())
}
}
@ -171,7 +169,7 @@ class SimpleWorkspaceTest {
fun testMetaPropagation() {
runBlocking {
val node = workspace.produce("sum") { "testFlag" put true }
val res = node.flowData().single().await()
val res = node.dataSequence().single().await()
}
}
@ -194,7 +192,7 @@ class SimpleWorkspaceTest {
fun testFilter() {
runBlocking {
val node = workspace.produce("filterOne")
assertEquals(12, node.flowData().first().await())
assertEquals(12, node.dataSequence().first().await())
}
}
}