com.apple.foundationdb.record.provider.foundationdb

Class SplitHelper

java.lang.Object

com.apple.foundationdb.record.provider.foundationdb.SplitHelper

@API(value=INTERNAL)
public class SplitHelper
extends Object

Helper classes for splitting records across multiple key-value pairs.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	SplitHelper.FoundSplitWithoutStartException Exception thrown when only part of a split record is found.
static class	SplitHelper.KeyValueUnsplitter This cursor may exceed out-of-band limits in order to ensure that it only ever stops in between (split) records.
static class	SplitHelper.SingleKeyUnsplitter Unsplit a single record from a given range scan.
static class	SplitHelper.SizeInfo Accumulator for key-value sizes while loading / saving split records.

Field Summary

Fields

Modifier and Type	Field and Description
static long	RECORD_VERSION Special split point added to end of the key when a record is associated with a version.
static int	SPLIT_RECORD_SIZE If a record is greater than this size (in bytes), it will be split into multiple kv pairs.
static long	START_SPLIT_RECORD Minimum index used when a record is split.
static long	UNSPLIT_RECORD Value added to the end of the key when a record is not split.

Method Summary

Modifier and Type	Method and Description
static void	deleteSplit(FDBRecordContext context, Subspace subspace, Tuple key, boolean splitLongRecords, boolean missingUnsplitRecordSuffix, boolean clearBasedOnPreviousSizeInfo, FDBStoredSizes previousSizeInfo) Delete the serialized representation of a record.
static CompletableFuture<Boolean>	keyExists(ReadTransaction tr, FDBTransactionContext context, Subspace subspace, Tuple key, boolean splitLongRecords, boolean missingUnsplitRecordSuffix) Checks to see if a given key exists.
static CompletableFuture<com.apple.foundationdb.record.provider.foundationdb.FDBRawRecord>	loadWithSplit(ReadTransaction tr, FDBRecordContext context, Subspace subspace, Tuple key, boolean splitLongRecords, boolean missingUnsplitRecordSuffix, SplitHelper.SizeInfo sizeInfo) Load serialized byte array that may be split among several keys.
static void	saveWithSplit(FDBRecordContext context, Subspace subspace, Tuple key, byte[] serialized, FDBRecordVersion version) Save serialized representation using multiple keys if necessary.
static void	saveWithSplit(FDBRecordContext context, Subspace subspace, Tuple key, byte[] serialized, FDBRecordVersion version, boolean splitLongRecords, boolean omitUnsplitSuffix, boolean clearBasedOnPreviousSizeInfo, FDBStoredSizes previousSizeInfo, SplitHelper.SizeInfo sizeInfo) Save serialized representation using multiple keys if necessary, clearing only as much as needed.
static Tuple	unpackKey(Subspace subspace, KeyValue kv)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

SPLIT_RECORD_SIZE

public static final int SPLIT_RECORD_SIZE

If a record is greater than this size (in bytes), it will be split into multiple kv pairs.

See Also:

Constant Field Values

RECORD_VERSION

public static final long RECORD_VERSION

Special split point added to end of the key when a record is associated with a version.

See Also:

Constant Field Values

UNSPLIT_RECORD

public static final long UNSPLIT_RECORD

Value added to the end of the key when a record is not split.

See Also:

Constant Field Values

START_SPLIT_RECORD

public static final long START_SPLIT_RECORD

Minimum index used when a record is split. The index is appended to the key, one for each split portion.

See Also:

Constant Field Values

Method Detail

saveWithSplit

public static void saveWithSplit(@Nonnull
                                 FDBRecordContext context,
                                 @Nonnull
                                 Subspace subspace,
                                 @Nonnull
                                 Tuple key,
                                 @Nonnull
                                 byte[] serialized,
                                 @Nullable
                                 FDBRecordVersion version)

Save serialized representation using multiple keys if necessary.

Parameters:

context - write transaction

subspace - subspace to save in

key - key within subspace

serialized - serialized representation

version - the version to store inline with this record

saveWithSplit

public static void saveWithSplit(@Nonnull
                                 FDBRecordContext context,
                                 @Nonnull
                                 Subspace subspace,
                                 @Nonnull
                                 Tuple key,
                                 @Nonnull
                                 byte[] serialized,
                                 @Nullable
                                 FDBRecordVersion version,
                                 boolean splitLongRecords,
                                 boolean omitUnsplitSuffix,
                                 boolean clearBasedOnPreviousSizeInfo,
                                 @Nullable
                                 FDBStoredSizes previousSizeInfo,
                                 @Nullable
                                 SplitHelper.SizeInfo sizeInfo)

Save serialized representation using multiple keys if necessary, clearing only as much as needed.

Parameters:

context - write transaction

subspace - subspace to save in

key - key within subspace

serialized - serialized representation

version - the version to store inline with this record

splitLongRecords - true if multiple keys should be used; if false, serialized must fit in a single key

omitUnsplitSuffix - if splitLongRecords is false, then this will omit a suffix added to the end of the key if true for backwards-compatibility reasons

clearBasedOnPreviousSizeInfo - if splitLongRecords, whether to use previousSizeInfo to determine how much to clear

previousSizeInfo - if clearBasedOnPreviousSizeInfo, the FDBStoredSizes for any old record, or null if there was no old record

sizeInfo - optional size information to populate

loadWithSplit

public static CompletableFuture<com.apple.foundationdb.record.provider.foundationdb.FDBRawRecord> loadWithSplit(@Nonnull
                                                                                                                ReadTransaction tr,
                                                                                                                @Nonnull
                                                                                                                FDBRecordContext context,
                                                                                                                @Nonnull
                                                                                                                Subspace subspace,
                                                                                                                @Nonnull
                                                                                                                Tuple key,
                                                                                                                boolean splitLongRecords,
                                                                                                                boolean missingUnsplitRecordSuffix,
                                                                                                                @Nullable
                                                                                                                SplitHelper.SizeInfo sizeInfo)

Load serialized byte array that may be split among several keys.

Parameters:

tr - read transaction

context - transaction context

subspace - subspace containing serialized value

key - key within subspace

splitLongRecords - true if multiple keys should be used; if false, serialized must fit in a single key

missingUnsplitRecordSuffix - if splitLongRecords is false and this is true, this will assume keys are missing a suffix for backwards compatibility reasons

sizeInfo - optional size information to populate

Returns:

the merged byte array

keyExists

public static CompletableFuture<Boolean> keyExists(@Nonnull
                                                   ReadTransaction tr,
                                                   @Nonnull
                                                   FDBTransactionContext context,
                                                   @Nonnull
                                                   Subspace subspace,
                                                   @Nonnull
                                                   Tuple key,
                                                   boolean splitLongRecords,
                                                   boolean missingUnsplitRecordSuffix)

Checks to see if a given key exists.

Parameters:

tr - read transaction

context - transaction context

subspace - subspace containing serialized value

key - key within subspace

splitLongRecords - true if multiple keys should be used; if false, serialized must fit in a single key

missingUnsplitRecordSuffix - if splitLongRecords is false and this is true, this will assume keys are missing a suffix for backwards compatibility reasons

Returns:

true if the provided key exists, false otherwise.

deleteSplit

public static void deleteSplit(@Nonnull
                               FDBRecordContext context,
                               @Nonnull
                               Subspace subspace,
                               @Nonnull
                               Tuple key,
                               boolean splitLongRecords,
                               boolean missingUnsplitRecordSuffix,
                               boolean clearBasedOnPreviousSizeInfo,
                               @Nullable
                               FDBStoredSizes previousSizeInfo)

Delete the serialized representation of a record. This possibly requires deleting multiple keys, but this method makes an effort to clear only as much as needed.

Parameters:

context - write transaction

subspace - subspace to delete from

key - primary key of the record to delete within subspace

splitLongRecords - true if multiple keys should be used; if false, serialized must fit in a single key

missingUnsplitRecordSuffix - if splitLongRecords is false and this is true, this will assume keys are missing a suffix for backwards compatibility reasons

clearBasedOnPreviousSizeInfo - if splitLongRecords, whether to use previousSizeInfo to determine how much to clear

previousSizeInfo - if clearBasedOnPreviousSizeInfo, the FDBStoredSizes for any old record, or null if there was no old record

unpackKey

public static Tuple unpackKey(@Nonnull
                              Subspace subspace,
                              @Nonnull
                              KeyValue kv)