Fix bug of scan aggregate index returning empty non-end continuation #3397
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -33,6 +33,7 @@ | |
| import com.apple.foundationdb.record.KeyRange; | ||
| import com.apple.foundationdb.record.RecordCoreException; | ||
| import com.apple.foundationdb.record.RecordCursorContinuation; | ||
| import com.apple.foundationdb.record.RecordCursorProto; | ||
| import com.apple.foundationdb.record.RecordCursorResult; | ||
| import com.apple.foundationdb.record.ScanProperties; | ||
| import com.apple.foundationdb.record.TupleRange; | ||
|
|
@@ -42,11 +43,13 @@ | |
| import com.apple.foundationdb.subspace.Subspace; | ||
| import com.apple.foundationdb.tuple.Tuple; | ||
| import com.google.protobuf.ByteString; | ||
| import com.google.protobuf.InvalidProtocolBufferException; | ||
| import com.google.protobuf.ZeroCopyByteString; | ||
|
|
||
| import javax.annotation.Nonnull; | ||
| import javax.annotation.Nullable; | ||
| import java.util.Arrays; | ||
| import java.util.Objects; | ||
| import java.util.Optional; | ||
| import java.util.concurrent.CompletableFuture; | ||
|
|
||
|
|
@@ -61,22 +64,26 @@ public abstract class KeyValueCursorBase<K extends KeyValue> extends AsyncIterat | |
| private final int prefixLength; | ||
| @Nonnull | ||
| private final CursorLimitManager limitManager; | ||
| private final int valuesLimit; | ||
| // the pointer may be mutated, but the actual array must never be mutated or continuations will break | ||
| @Nullable | ||
| private byte[] lastKey; | ||
| @Nonnull | ||
| private final SerializationMode serializationMode; | ||
|
|
||
| protected KeyValueCursorBase(@Nonnull final FDBRecordContext context, | ||
| @Nonnull final AsyncIterator<K> iterator, | ||
| int prefixLength, | ||
| @Nonnull final CursorLimitManager limitManager, | ||
| int valuesLimit, | ||
| @Nonnull final SerializationMode serializationMode) { | ||
| super(context.getExecutor(), iterator); | ||
|
|
||
| this.context = context; | ||
| this.prefixLength = prefixLength; | ||
| this.limitManager = limitManager; | ||
| this.valuesLimit = valuesLimit; | ||
| this.serializationMode = serializationMode; | ||
|
|
||
| context.instrument(FDBStoreTimer.DetailEvents.GET_SCAN_RANGE_RAW_FIRST_CHUNK, iterator.onHasNext()); | ||
| } | ||
|
|
@@ -131,21 +138,23 @@ public RecordCursorResult<K> getNext() { | |
|
|
||
| @Nonnull | ||
| private RecordCursorContinuation continuationHelper() { | ||
| return new Continuation(lastKey, prefixLength, serializationMode); | ||
| } | ||
|
|
||
| public static class Continuation implements RecordCursorContinuation { | ||
| @Nullable | ||
| private final byte[] lastKey; | ||
| private final int prefixLength; | ||
| private final SerializationMode serializationMode; | ||
|
|
||
| public Continuation(@Nullable final byte[] lastKey, final int prefixLength, final SerializationMode serializationMode) { | ||
| // Note that doing this without a full copy is dangerous if the array is ever mutated. | ||
| // Currently, this never happens and the only thing that changes is which array lastKey points to. | ||
| // However, if logic in KeyValueCursor or KeyValue changes, this could break continuations. | ||
| // To resolve it, we could resort to doing a full copy here, although that's somewhat expensive. | ||
| this.lastKey = lastKey; | ||
| this.prefixLength = prefixLength; | ||
| this.serializationMode = serializationMode; | ||
| } | ||
|
|
||
| @Override | ||
|
|
@@ -156,21 +165,83 @@ public boolean isEnd() { | |
| @Nonnull | ||
| @Override | ||
| public ByteString toByteString() { | ||
| if (serializationMode == SerializationMode.TO_OLD) { | ||
| // lastKey = null when source iterator hit limit that we passed down. | ||
| if (lastKey == null) { | ||
| return ByteString.EMPTY; | ||
| } | ||
| ByteString base = ZeroCopyByteString.wrap(lastKey); | ||
| // when prefixLength == lastKey.length, toByteString() also returns ByteString.EMPTY | ||
| return base.substring(prefixLength, lastKey.length); | ||
| } else { | ||
| return toProto().toByteString(); | ||
| } | ||
| } | ||
|
|
||
| @Nullable | ||
| @Override | ||
| public byte[] toBytes() { | ||
| if (lastKey == null) { | ||
| return null; | ||
| } | ||
| ByteString byteString = toByteString(); | ||
| return byteString.isEmpty() ? new byte[0] : byteString.toByteArray(); | ||
| } | ||
|
|
||
| @Nullable | ||
| public byte[] getInnerContinuationInBytes() { | ||
| if (lastKey == null) { | ||
| return null; | ||
| } | ||
| return Arrays.copyOfRange(lastKey, prefixLength, lastKey.length); | ||
| } | ||
|
|
||
| @Nonnull | ||
| public ByteString getInnerContinuationInByteString() { | ||
| if (lastKey == null) { | ||
| return ByteString.EMPTY; | ||
| } | ||
| ByteString base = ZeroCopyByteString.wrap(lastKey); | ||
| return base.substring(prefixLength, lastKey.length); | ||
| } | ||
|
|
||
| public static byte[] fromRawBytes(@Nullable byte[] rawBytes, SerializationMode serializationMode) { | ||
| if (rawBytes == null) { | ||
| return null; | ||
| } | ||
| if (serializationMode == SerializationMode.TO_OLD) { | ||
| return rawBytes; | ||
| } | ||
| try { | ||
| RecordCursorProto.KeyValueCursorContinuation continuationProto = RecordCursorProto.KeyValueCursorContinuation.parseFrom(rawBytes); | ||
| if (continuationProto.hasPrefixLength()) { | ||
| return continuationProto.getContinuation().toByteArray(); | ||
| } else { | ||
| // parseFrom can parse an old serialization result as the new proto, wrong deserialization | ||
| return rawBytes; | ||
| } | ||
| } catch (InvalidProtocolBufferException ipbe) { | ||
| return rawBytes; | ||
| } | ||
| } | ||
|
|
||
| @Nonnull | ||
| private RecordCursorProto.KeyValueCursorContinuation toProto() { | ||
| RecordCursorProto.KeyValueCursorContinuation.Builder builder = RecordCursorProto.KeyValueCursorContinuation.newBuilder(); | ||
| if (lastKey == null) { | ||
|
There was a problem hiding this comment. The
So, I don't think we'll ever hit this code path. I think it also points to the possibility that this isn't quite the right format for the continuation, though that's something that I said in a comment on There was a problem hiding this comment. Thanks for catching this! I removed the |
||
| // proto.hasContinuation() = false when lastKey = null | ||
| return builder.setPrefixLength(prefixLength).build(); | ||
| } else { | ||
| ByteString base = ZeroCopyByteString.wrap(Objects.requireNonNull(lastKey)); | ||
| // proto.hasContinuation() = ByteString.EMPTY when prefixLength = lastKey.length | ||
| return builder.setContinuation(base.substring(prefixLength, lastKey.length)).setPrefixLength(prefixLength).build(); | ||
| } | ||
| } | ||
| } | ||
|
|
||
| public enum SerializationMode { | ||
| TO_OLD, | ||
| TO_NEW | ||
|
Comment on lines
+243
to
+244
There was a problem hiding this comment. Using "old" and "new" here may be fine (it's certainly what we did with #3254). The danger is that we leave it in for too long and then we're confused what's "new" about it. I think the goal is for this to be pretty temporary, and for us to move pretty quickly to a mode where we only have the new way, and if that's the case, then we're probably okay with this nomenclature |
||
| } | ||
|
|
||
| /** | ||
|
|
@@ -208,9 +279,11 @@ public abstract static class Builder<T extends Builder<T>> { | |
| private StreamingMode streamingMode; | ||
| private KeySelector begin; | ||
| private KeySelector end; | ||
| protected SerializationMode serializationMode; | ||
|
There was a problem hiding this comment. It seems like there's some confusion about the lifecycle of this value. It's a member variable of the builder, and there's a setter, but it also gets cribbed from the execute properties during |
||
|
|
||
| protected Builder(@Nonnull Subspace subspace) { | ||
| this.subspace = subspace; | ||
| this.serializationMode = SerializationMode.TO_OLD; | ||
| } | ||
|
|
||
| /** | ||
|
|
@@ -247,10 +320,12 @@ protected void prepare() { | |
| prefixLength = calculatePrefixLength(); | ||
|
|
||
| reverse = scanProperties.isReverse(); | ||
|
|
||
| if (continuation != null) { | ||
| byte[] realContinuation = KeyValueCursorBase.Continuation.fromRawBytes(continuation, serializationMode); | ||
| final byte[] continuationBytes = new byte[prefixLength + realContinuation.length]; | ||
| System.arraycopy(lowBytes, 0, continuationBytes, 0, prefixLength); | ||
| System.arraycopy(realContinuation, 0, continuationBytes, prefixLength, realContinuation.length); | ||
| if (reverse) { | ||
| highBytes = continuationBytes; | ||
| highEndpoint = EndpointType.CONTINUATION; | ||
|
|
@@ -334,6 +409,11 @@ public T setHigh(@Nonnull byte[] highBytes, @Nonnull EndpointType highEndpoint) | |
| return self(); | ||
| } | ||
|
|
||
| public T setSerializationMode(@Nonnull final SerializationMode serializationMode) { | ||
| this.serializationMode = serializationMode; | ||
| return self(); | ||
| } | ||
|
|
||
| /** | ||
| * Calculate the key prefix length for the returned values. This will be used to derive the primary key used in | ||
| * the calculated continuation. | ||
|
|
||
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does this need to be
publicjust so that theRecordQueryIndexPlancan update its continuation appropriately?There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes
RecordIndexPlancalls this method:KeyValueCursorBase.Continuation.fromRawBytes(continuation, serializationMode);