shunping commented on code in PR #30317:
URL: https://github.com/apache/beam/pull/30317#discussion_r1500882918
##########
sdks/java/harness/src/test/java/org/apache/beam/fn/harness/state/FakeBeamFnStateClient.java:
##########
@@ -174,6 +215,124 @@ public CompletableFuture<StateResponse>
handle(StateRequest.Builder requestBuild
response =
StateResponse.newBuilder().setAppend(StateAppendResponse.getDefaultInstance());
break;
+ case ORDERED_LIST_GET:
+ {
+ long start = request.getOrderedListGet().getRange().getStart();
+ long end = request.getOrderedListGet().getRange().getEnd();
+
+ KvCoder<Long, Integer> coder = KvCoder.of(VarLongCoder.of(),
VarIntCoder.of());
+ long sortKey = start;
+ int index = 0;
+ if (request.getOrderedListGet().getContinuationToken().size() > 0) {
+ try {
+ // The continuation format here is the sort key (long) followed
by an index (int)
+ KV<Long, Integer> cursor =
+
coder.decode(request.getOrderedListGet().getContinuationToken().newInput());
+ sortKey = cursor.getKey();
+ index = cursor.getValue();
+ } catch (IOException e) {
+ throw new RuntimeException(e);
+ }
+ }
+
+ ByteString continuationToken;
+ ByteString returnBlock = ByteString.EMPTY;
+ ;
+ try {
+ if (sortKey < start || sortKey >= end) {
+ throw new IndexOutOfBoundsException("sort key out of range");
+ }
+
+ NavigableSet<Long> subset =
+ orderedListKeys
+ .getOrDefault(request.getStateKey(), new TreeSet<>())
+ .subSet(sortKey, true, end, false);
+
+ // get the effective sort key currently, can throw
NoSuchElementException
+ Long nextSortKey = subset.first();
+
+ StateKey.Builder keyBuilder = request.getStateKey().toBuilder();
+
keyBuilder.getOrderedListUserStateBuilder().setSortKey(nextSortKey);
+ List<ByteString> byteStrings =
+ data.getOrDefault(keyBuilder.build(),
Collections.singletonList(ByteString.EMPTY));
+
+ // get the block specified in continuation token, can throw
IndexOutOfBoundsException
+ returnBlock = byteStrings.get(index);
+
+ if (byteStrings.size() > index + 1) {
+ // more blocks from this sort key
+ index += 1;
+ } else {
+ // finish navigating the current sort key and need to find the
next one,
+ // can throw NoSuchElementException
+ nextSortKey = subset.tailSet(nextSortKey, false).first();
+ index = 0;
+ }
+
+ ByteStringOutputStream outputStream = new ByteStringOutputStream();
+ try {
+ KV<Long, Integer> cursor = KV.of(nextSortKey, index);
+ coder.encode(cursor, outputStream);
+ } catch (IOException e) {
+ throw new RuntimeException(e);
+ }
+ continuationToken = outputStream.toByteString();
+ } catch (NoSuchElementException | IndexOutOfBoundsException e) {
+ continuationToken = ByteString.EMPTY;
+ }
+ response =
+ StateResponse.newBuilder()
+ .setOrderedListGet(
+ OrderedListStateGetResponse.newBuilder()
+ .setData(returnBlock)
+ .setContinuationToken(continuationToken));
+ }
+ break;
+
+ case ORDERED_LIST_UPDATE:
+ for (OrderedListRange r :
request.getOrderedListUpdate().getDeletesList()) {
+ List<Long> keysToRemove =
+ new ArrayList<>(
+ orderedListKeys
+ .getOrDefault(request.getStateKey(), new TreeSet<>())
+ .subSet(r.getStart(), true, r.getEnd(), false));
+
+ for (Long l : keysToRemove) {
+ StateKey.Builder keyBuilder = request.getStateKey().toBuilder();
+ keyBuilder.getOrderedListUserStateBuilder().setSortKey(l);
+ data.remove(keyBuilder.build());
+ orderedListKeys.get(request.getStateKey()).remove(l);
+ }
+ }
+
+ for (OrderedListEntry e :
request.getOrderedListUpdate().getInsertsList()) {
+ StateKey.Builder keyBuilder = request.getStateKey().toBuilder();
+
keyBuilder.getOrderedListUserStateBuilder().setSortKey(e.getSortKey());
+
+ ByteStringOutputStream outStream = new ByteStringOutputStream();
+
+ try {
+ InstantCoder.of().encode(Instant.ofEpochMilli(e.getSortKey()),
outStream);
+ } catch (IOException ex) {
+ throw new RuntimeException(ex);
+ }
+ // In the response, the value encoded bytes are placed before the
timestamp encoded bytes.
Review Comment:
ACK.
I find a better way to do this. Specifically, I can reuse
TimestampedValueCoder but also implement a special coder for ByteString that
doesn't append string length during the encoding phase. Then I can use
`TimestampedValueCoder.of(NewCoder)` to encode TimestampedValue, like
https://github.com/apache/beam/blob/33e748d175079c68f4fc2935a396cb10a4582637/sdks/java/harness/src/test/java/org/apache/beam/fn/harness/state/FakeBeamFnStateClient.java#L341
This can both avoid the decoding overhead in the fake client, but it also
leverage TimestampedValueCoder without explicitly documenting the expected
buffer layout in the comment like I did before
https://github.com/apache/beam/blob/5d7cd5e2e2e6b966ddb47cd78135b58013622d23/sdks/java/harness/src/test/java/org/apache/beam/fn/harness/state/FakeBeamFnStateClient.java#L319
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscr...@beam.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
