org.das2.qds.buffer.BufferDataSet

rank 1, 2, 3, and 4 datasets backed by NIO buffers. These have the advantage that data can be stored outside of the JVM, and in fact they can be backed by a huge file on disk. This code was copied from BinaryDataSource.

FLOAT

the data is in 4 byte floats.

TRUNCATEDFLOAT

the data is in 16 bit real that has exponent like a FLOAT but mantissa precision is reduced.

VAX_FLOAT

VAX floats.

INT24

three-byte ints.

UINT24

three-byte unsigned ints.

NYBBLE

four-bit unsigned ints.

INTEGER

4 byte signed integers, INT is canonical but INTEGER should be accepted.

UINT

4 byte unsigned integers. Note 4-byte signed ints are used to store the data which is unpacked in the value() method.

SHORT

2 byte short integer.

USHORT

2 byte unsigned short.

BYTE

1 byte signed byte.

UBYTE

1 byte unsigned byte.

BYTES

constructor units are in bytes.

BITS

constructor units are in bits.

bitCount

bitCount( Object type ) → int

Returns:

int

search for examples view on GitHub view source

byteCount

byteCount( Object type ) → int

return the number of bytes of each type (double=8, etc).

Parameters

type - DOUBLE, FLOAT, UBYTE, TIME28, etc.

Returns:

8, 4, 1, etc.

search for examples view on GitHub view source

compact

compact( ) → BufferDataSet

get ride of extra spaces between records.

Returns:

new BufferDataSet without gaps.

search for examples view on GitHub view source

getCompatibleComponentType

getCompatibleComponentType( ) → Class

return the Java type that is capable of containing elements of this dataset. For unsigned types, the next Java class is used, for example int.class is used to store unsigned shorts.

Returns:

double.class, float.class, long.class, etc.

search for examples view on GitHub view source

getType

getType( ) → Object

return the type of this dataset, for example BufferDataSet.INT, BufferDataSet.DOUBLE, etc...

Returns:

the type of this dataset.

search for examples view on GitHub view source

isCompact

isCompact( ) → boolean

returns true if the dataset is compact, meaning that there are no gaps between records, and no byte offset.

Returns:

true if the dataset is compact

search for examples view on GitHub view source

makeDataSet

makeDataSet( int rank, int reclen, int recoffs, int len0, int len1, int len2, int len3, java.nio.ByteBuffer buf, Object type ) → BufferDataSet

Make a BufferDataSet of the given type.

Parameters

rank - the rank (number of indeces) of the data.
reclen - length in bytes of each record. This may be longer than len1*len2*len3*byteCount(type)
recoffs - byte offset of each record
len0 - number of elements in the first index
len1 - number of elements in the second index
len2 - number of elements in the third index
len3 - number of elements in the fourth index
buf - ByteBuffer containing the data, which should be at least recoffs + reclen * len0 bytes long.
type - BufferDataSet.INT, BufferDataSet.DOUBLE, etc...

Returns:

BufferDataSet of the given type.

search for examples view on GitHub view source

makeDataSetBits

makeDataSetBits( int rank, int reclenbits, int recoffsbits, int len0, int len1, int len2, int len3, java.nio.ByteBuffer buf, Object type ) → BufferDataSet

support binary types that are not a multiple of 8 bits. This was added to support Nybbles, and 12-bit ints.

Parameters

rank -
reclenbits - number of bits per record
recoffsbits - number of bits offset. Note this must be a multiple of 8, for now.
len0 - number of elements in the first index
len1 - number of elements in the second index
len2 - number of elements in the third index
len3 - number of elements in the fourth index
buf - ByteBuffer containing the data, which should be at least recoffsbits/8 + reclenbits/8 * len0 bytes long.
type - BufferDataSet.NYBBLE, etc

Returns:

BufferDataSet of the given type.

search for examples view on GitHub view source

shouldAllocateDirect

shouldAllocateDirect( ) → int

return 1 if direct allocate should be used, 0 if not. Direct allocations are memory allocations outside of the JVM heap memory. (The internal variable has a -1 initial state, which is why this is not boolean.) This looks for 32bit Javas, and if more than 1/2 Gig is being used then it will allocate direct. This is because 32bit Javas cannot access any memory outside of 1Gig.

Returns:

1 or 0 if direct allocations should not be made.

See Also:

https://sourceforge.net/p/autoplot/bugs/1395/
http://stackoverflow.com/questions/807263/how-do-i-detect-which-kind-of-jre-is-installed-32bit-vs-64bit
http://stackoverflow.com/questions/3651737/why-the-odd-performance-curve-differential-between-bytebuffer-allocate-and-byt "How ByteBuffer works and why Direct (Byte) "How ByteBuffer works and why Direct (Byte)Buffers are the only truly useful now"


search for examples view on GitHub view source

typeFor

typeFor( java.lang.Class c ) → Object

Return the type for the given class. Note that there is a type for each native type (Byte,Short,Float,etc), but not a class for each type. (E.g. UBYTE is unsigned byte.)

Parameters

c - java class

Returns:

DOUBLE,FLOAT,etc.

search for examples view on GitHub view source