Package uk.ac.starlink.table.storage
Class AdaptiveByteStore
java.lang.Object
uk.ac.starlink.table.storage.AdaptiveByteStore
- All Implemented Interfaces:
ByteStore
ByteStore which adopts a hybrid approach between use of memory
and use of disk. Bytes are written into an array in memory up to
a given size limit; if the amount written exceeds this limit,
it's all put in a temporary file instead.
This class is intended to be a general purpose StoragePolicy implementation that does something sensible most of the time. The details of the implementation may be changed following experience.
The current implementation uses ByteBuffer.allocateDirect(int)
for byte arrays in memory apart from rather small ones.
On most OSes this corresponds to using malloc()
,
thus avoiding heavy use of JVM heap memory.
Note very large arrays are still stored on disk, not directly
allocated.
- Since:
- 5 Nov 2009
- Author:
- Mark Taylor
-
Constructor Summary
ConstructorsConstructorDescriptionConstructs a new store with a default memory limit.AdaptiveByteStore
(int memLimit) Constructs a new store with a given maximum memory limit. -
Method Summary
Modifier and TypeMethodDescriptionvoid
close()
Tidies up.void
copy
(OutputStream out) Takes all the data written so far into this store's sink stream and copies it to a destination stream.static int
Calculates the default memory limit used by this class.long
Returns the number of bytes currently stored in this object.Returns an output stream which can be used to write to the store.Returns an array of byte buffers containing the bytes written to this store.
-
Constructor Details
-
AdaptiveByteStore
Constructs a new store with a given maximum memory limit.- Parameters:
memLimit
- maximum size of in-memory buffer- Throws:
IOException
-
AdaptiveByteStore
Constructs a new store with a default memory limit.- Throws:
IOException
-
-
Method Details
-
getOutputStream
Description copied from interface:ByteStore
Returns an output stream which can be used to write to the store. May be called multiple times; always returns the same object. Note that this is not in general buffered - it is the responsibility of the user to take steps like wrapping it in aBufferedOutputStream
for efficiency if required.- Specified by:
getOutputStream
in interfaceByteStore
- Returns:
- data sink stream
-
getLength
public long getLength()Description copied from interface:ByteStore
Returns the number of bytes currently stored in this object. -
copy
Description copied from interface:ByteStore
Takes all the data written so far into this store's sink stream and copies it to a destination stream. The output stream is not closed.- Specified by:
copy
in interfaceByteStore
- Parameters:
out
- data destination stream- Throws:
IOException
-
toByteBuffers
Description copied from interface:ByteStore
Returns an array of byte buffers containing the bytes written to this store. The stored bytes are all of the bytes from the first buffer in the returned array, followed by all in the second, etc. In many cases the returned array will, and probably should, contain a single buffer, but if the written byte count exceedsInteger.MAX_VALUE
, more than one will be required. The limit of each buffer indicates the number of bytes it contains.Usual usage will be to write all data, then call this method once; this model may affect implementation decisions about efficiency.
- Specified by:
toByteBuffers
in interfaceByteStore
- Returns:
- byte buffer containing bytes written
- Throws:
IOException
- if there is an I/O error
-
close
public void close()Description copied from interface:ByteStore
Tidies up. Should be called when the data in this object is no longer required. This object may no longer be usable following a call to this method. -
getDefaultLimit
public static int getDefaultLimit()Calculates the default memory limit used by this class.- Returns:
- default memory limit
-