diff options
author | Patrick Wendell <patrick@databricks.com> | 2015-04-21 21:04:04 -0700 |
---|---|---|
committer | Patrick Wendell <patrick@databricks.com> | 2015-04-21 21:04:04 -0700 |
commit | 70f9f8ff38560967f2c84de77263a5455c45c495 (patch) | |
tree | 26510a73c4c18826ed8d0064a4d6e4bec4a5f0f9 /core | |
parent | 3a3f7100f4ead9b7ac50e9711ac50b603ebf6bea (diff) | |
download | spark-70f9f8ff38560967f2c84de77263a5455c45c495.tar.gz spark-70f9f8ff38560967f2c84de77263a5455c45c495.tar.bz2 spark-70f9f8ff38560967f2c84de77263a5455c45c495.zip |
[MINOR] Comment improvements in ExternalSorter.
1. Clearly specifies the contract/interactions for users of this class.
2. Minor fix in one doc to avoid ambiguity.
Author: Patrick Wendell <patrick@databricks.com>
Closes #5620 from pwendell/cleanup and squashes the following commits:
8d8f44f [Patrick Wendell] [Minor] Comment improvements in ExternalSorter.
Diffstat (limited to 'core')
-rw-r--r-- | core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala | 27 |
1 files changed, 19 insertions, 8 deletions
diff --git a/core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala b/core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala index 79a1a8a0da..79a695fb62 100644 --- a/core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala +++ b/core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala @@ -53,7 +53,18 @@ import org.apache.spark.storage.{BlockObjectWriter, BlockId} * probably want to pass None as the ordering to avoid extra sorting. On the other hand, if you do * want to do combining, having an Ordering is more efficient than not having it. * - * At a high level, this class works as follows: + * Users interact with this class in the following way: + * + * 1. Instantiate an ExternalSorter. + * + * 2. Call insertAll() with a set of records. + * + * 3. Request an iterator() back to traverse sorted/aggregated records. + * - or - + * Invoke writePartitionedFile() to create a file containing sorted/aggregated outputs + * that can be used in Spark's sort shuffle. + * + * At a high level, this class works internally as follows: * * - We repeatedly fill up buffers of in-memory data, using either a SizeTrackingAppendOnlyMap if * we want to combine by key, or an simple SizeTrackingBuffer if we don't. Inside these buffers, @@ -65,11 +76,11 @@ import org.apache.spark.storage.{BlockObjectWriter, BlockId} * aggregation. For each file, we track how many objects were in each partition in memory, so we * don't have to write out the partition ID for every element. * - * - When the user requests an iterator, the spilled files are merged, along with any remaining - * in-memory data, using the same sort order defined above (unless both sorting and aggregation - * are disabled). If we need to aggregate by key, we either use a total ordering from the - * ordering parameter, or read the keys with the same hash code and compare them with each other - * for equality to merge values. + * - When the user requests an iterator or file output, the spilled files are merged, along with + * any remaining in-memory data, using the same sort order defined above (unless both sorting + * and aggregation are disabled). If we need to aggregate by key, we either use a total ordering + * from the ordering parameter, or read the keys with the same hash code and compare them with + * each other for equality to merge values. * * - Users are expected to call stop() at the end to delete all the intermediate files. * @@ -259,8 +270,8 @@ private[spark] class ExternalSorter[K, V, C]( * Spill our in-memory collection to a sorted file that we can merge later (normal code path). * We add this file into spilledFiles to find it later. * - * Alternatively, if bypassMergeSort is true, we spill to separate files for each partition. - * See spillToPartitionedFiles() for that code path. + * This should not be invoked if bypassMergeSort is true. In that case, spillToPartitionedFiles() + * is used to write files for each partition. * * @param collection whichever collection we're using (map or buffer) */ |