@InterfaceAudience.Public public interface DataType<T>
DataType
is the base class for all HBase data types. Data type implementations are
designed to be serialized to and deserialized from byte[]. Serialized representations can retain
the natural sort ordering of the source object, when a suitable encoding is supported by the
underlying implementation. This is a desirable feature for use in rowkeys and column qualifiers.
DataType
s are different from Hadoop
ImmutableBytesWritable
s in two significant ways. First,
DataType
describes how to serialize a value, it does not encapsulate a serialized value.
Second, DataType
implementations provide hints to consumers about relationships between
the POJOs they represent and richness of the encoded representation.
Data type instances are designed to be stateless, thread-safe, and reused. Implementations should
provide static final
instances corresponding to each variation on configurable
parameters. This is to encourage and simplify instance reuse. For instance, order-preserving
types should provide static ASCENDING and DESCENDING instances. It is also encouraged for
implementations operating on Java primitive types to provide primitive implementations of the
encode
and decode
methods. This advice is a performance consideration to clients
reading and writing values in tight loops.
Modifier and Type | Method and Description |
---|---|
T |
decode(PositionedByteRange src)
Read an instance of
T from the buffer src . |
int |
encode(PositionedByteRange dst,
T val)
Write instance
val into buffer dst . |
Class<T> |
encodedClass()
Inform consumers over what type this
DataType operates. |
int |
encodedLength(T val)
Inform consumers how long the encoded
byte[] will be. |
Order |
getOrder()
Retrieve the sort
Order imposed by this data type, or null when natural ordering is not
preserved. |
boolean |
isNullable()
Indicates whether this instance supports encoding null values.
|
boolean |
isOrderPreserving()
Indicates whether this instance writes encoded
byte[] 's which preserve the natural sort
order of the unencoded value. |
boolean |
isSkippable()
Indicates whether this instance is able to skip over it's encoded value.
|
int |
skip(PositionedByteRange src)
Skip
src 's position forward over one encoded value. |
boolean isOrderPreserving()
byte[]
's which preserve the natural sort
order of the unencoded value.true
when natural order is preserved, false
otherwise.Order getOrder()
Order
imposed by this data type, or null when natural ordering is not
preserved. Value is either ascending or descending. Default is assumed to be
Order.ASCENDING
.boolean isNullable()
DataType
s that support null should
treat null as comparing less than any non-null value for default sort ordering purposes.true
when null is supported, false
otherwise.boolean isSkippable()
DataType
s that
are not skippable can only be used as the right-most field of a Struct
.int encodedLength(T val)
byte[]
will be.val
- The value to check.val
.aClass<T> encodedClass()
DataType
operates. Useful when working with bare
DataType
instances.int skip(PositionedByteRange src)
src
's position forward over one encoded value.src
- the buffer containing the encoded value.T decode(PositionedByteRange src)
T
from the buffer src
.src
- the buffer containing the encoded value.int encode(PositionedByteRange dst, T val)
val
into buffer dst
.dst
- the buffer containing the encoded value.val
- the value to encode onto dst
.Copyright © 2007–2020 The Apache Software Foundation. All rights reserved.