CompressFloat.h
Classes
- CompressFloat -- Virtual column engine to scale a table float array (full description)
class CompressFloat : public BaseMappedArrayEngine<Float, Short>
Interface
- Public Members
- CompressFloat (const String& virtualColumnName, const String& storedColumnName, Float scale, Float offset = 0)
- CompressFloat (const String& virtualColumnName, const String& storedColumnName, const String& scaleColumnName, const String& offsetColumnName, Bool autoScale = True)
- CompressFloat (const Record& spec)
- ~CompressFloat()
- virtual String dataManagerType() const
- virtual String dataManagerName() const
- virtual Record dataManagerSpec() const
- static String className()
- static void registerClass()
- Private Members
- CompressFloat (const CompressFloat&)
- CompressFloat& operator= (const CompressFloat&)
- virtual DataManager* clone() const
- virtual void create (uInt initialNrrow)
- virtual void prepare()
- virtual void reopenRW()
- virtual void addRowInit (uInt startRow, uInt nrrow)
- virtual void getArray (uInt rownr, Array<Float>& array)
- virtual void putArray (uInt rownr, const Array<Float>& array)
- virtual void getSlice (uInt rownr, const Slicer& slicer, Array<Float>& array)
- virtual void putSlice (uInt rownr, const Slicer& slicer, const Array<Float>& array)
- virtual void getArrayColumn (Array<Float>& array)
- virtual void putArrayColumn (const Array<Float>& array)
- virtual void getColumnSlice (const Slicer& slicer, Array<Float>& array)
- virtual void putColumnSlice (const Slicer& slicer, const Array<Float>& array)
- void scaleOnGet (Float scale, Float offset, Array<Float>& array, const Array<Short>& target)
- void scaleOnPut (Float scale, Float offset, const Array<Float>& array, Array<Short>& target)
- void scaleColumnOnGet (Array<Float>& array, const Array<Short>& target)
- void scaleColumnOnPut (const Array<Float>& array, Array<Short>& target)
- Float getScale (uInt rownr)
- Float getOffset (uInt rownr)
- void findMinMax (Float& minVal, Float& maxVal, const Array<Float>& array) const
- void makeScaleOffset (Float& scale, Float& offset, Float minVal, Float maxVal) const
- void putPart (uInt rownr, const Slicer& slicer, const Array<Float>& array, Float scale, Float offset)
- void putFullPart (uInt rownr, const Slicer& slicer, Array<Float>& fullArray, const Array<Float>& partArray, Float minVal, Float maxVal)
- Public Members
- static DataManager* makeObject (const String& dataManagerType, const Record& spec)
Review Status
- Reviewed By:
- UNKNOWN
- Date Reviewed:
- before2004/08/25
- Programs:
- Tests:
Prerequisite
- VirtualColumnEngine
- VirtualArrayColumn
Synopsis
CompressFloat is a virtual column engine which scales an array
of one type to another type to save disk storage.
This resembles the classic AIPS compress method which scales the
data from float to short.
The scale factor and offset values can be given in two ways:
- As a fixed values which is used for all arrays in the column.
These values have to be given when constructing of the engine.
- As the name of a column. In this way each array in the
column has its own scale and offset value.
By default it uses auto-scaling (see below).
Otherwise the scale and offset value in a row must be put
before the array is put and should not be changed anymore.
Auto-scaling means that the engine will determine the scale
and offset value itself when an array (or a slice) is put.
It does it by mapping the values in the array to the range [-32767,32767].
At each put the scale/offset values are changed as needed.
Note that with auto-scaling putSlice can be somewhat
slower, because the entire array might need to be rescaled.
As in FITS the scale and offset values are used as:
True_value = Stored_value * scale + offset;
An engine object should be used for one column only, because the stored
column name is part of the engine. If it would be used for more than
one column, they would all share the same stored column.
When the engine is bound to a column, it is checked if the name
of that column matches the given virtual column name.
The engine can be used for a column containing any kind of array
(thus direct or indirect, fixed or variable shaped)) as long as the
virtual array can be stored in the stored array. Thus a fixed shaped
virtual can use a variable shaped stored, but not vice versa.
A fixed shape indirect virtual can use a stored with direct arrays.
This class can also serve as an example of how to implement
a virtual column engine.
Motivation
This class allows to store data in a smaller representation.
It is needed to resemble the classic AIPS compress option.
Because the engine can serve only one column, it was possible to
combine the engine and the column functionality in one class.
Example
// Create the table description and 2 columns with indirect arrays in it.
// The Int column will be stored, while the double will be
// used as virtual.
TableDesc tableDesc ("", TableDesc::Scratch);
tableDesc.addColumn (ArrayColumnDesc<Short> ("storedArray"));
tableDesc.addColumn (ArrayColumnDesc<Float> ("virtualArray"));
tableDesc.addColumn (ScalarColumnDesc<Float> ("scale"));
tableDesc.addColumn (ScalarColumnDesc<Float> ("offset"));
// Create a new table using the table description.
SetupNewTable newtab (tableDesc, "tab.data", Table::New);
// Create the array scaling engine (with auto-scale)
// and bind it to the float column.
CompressFloat scalingEngine("virtualArray", "storedArray",
"scale", "offset");
newtab.bindColumn ("virtualArray", scalingEngine);
// Create the table.
Table table (newtab);
// Store a 3-D array (with dim. 2,3,4) into each row of the column.
// The shape of each array in the column is implicitly set by the put
// function. This will also set the shape of the underlying Int array.
ArrayColumn data (table, "virtualArray");
Array<double> someArray(IPosition(4,2,3,4));
someArray = 0;
for (uInt i=0, i<10; i++) { // table will have 10 rows
table.addRow();
data.put (i, someArray)
}
Member Description
CompressFloat (const String& virtualColumnName, const String& storedColumnName, Float scale, Float offset = 0)
Construct an engine to scale all arrays in a column with
the given offset and scale factor.
StoredColumnName is the name of the column where the scaled
data will be put and must have data type Short.
The virtual column using this engine must have data type Float.
CompressFloat (const String& virtualColumnName, const String& storedColumnName, const String& scaleColumnName, const String& offsetColumnName, Bool autoScale = True)
Construct an engine to scale the arrays in a column.
The scale and offset values are taken from a column with
the given names. In that way each array has its own scale factor
and offset value.
An exception is thrown if these columns do not exist.
VirtualColumnName is the name of the virtual column and is used to
check if the engine gets bound to the correct column.
StoredColumnName is the name of the column where the scaled
data will be put and must have data type Short.
The virtual column using this engine must have data type Float.
Construct from a record specification as created by getmanagerSpec().
Destructor is mandatory.
Return the type name of the engine (i.e. its class name).
Get the name given to the engine (is the virtual column name).
Record a record containing data manager specifications.
Return the name of the class.
This includes the names of the template arguments.
Register the class name and the static makeObject "constructor".
This will make the engine known to the table system.
Copy constructor is only used by clone().
(so it is made private).
CompressFloat& operator= (const CompressFloat&)
Assignment is not needed and therefore forbidden
(so it is made private and not implemented).
Clone the engine object.
virtual void create (uInt initialNrrow)
Initialize the object for a new table.
It defines the keywords containing the engine parameters.
virtual void prepare()
Preparing consists of setting the writable switch and
adding the initial number of rows in case of create.
Furthermore it reads the keywords containing the engine parameters.
Reopen the engine for read/write access.
It makes the column writable if the underlying column is writable.
virtual void addRowInit (uInt startRow, uInt nrrow)
Add rows to the table.
If auto-scaling, it initializes the scale column with 0
to indicate that no data has been processed yet.
virtual void getArray (uInt rownr, Array<Float>& array)
Get an array in the given row.
This will scale and offset from the underlying array.
virtual void putArray (uInt rownr, const Array<Float>& array)
Put an array in the given row.
This will scale and offset to the underlying array.
virtual void getSlice (uInt rownr, const Slicer& slicer, Array<Float>& array)
Get a section of the array in the given row.
This will scale and offset from the underlying array.
virtual void putSlice (uInt rownr, const Slicer& slicer, const Array<Float>& array)
Put into a section of the array in the given row.
This will scale and offset to the underlying array.
Get an entire column.
This will scale and offset from the underlying array.
Put an entire column.
This will scale and offset to the underlying array.
Get a section of all arrays in the column.
This will scale and offset from the underlying array.
virtual void putColumnSlice (const Slicer& slicer, const Array<Float>& array)
Put a section of all arrays in the column.
This will scale and offset to the underlying array.
void scaleOnGet (Float scale, Float offset, Array<Float>& array, const Array<Short>& target)
Scale and/or offset target to array.
This is meant when reading an array from the stored column.
It optimizes for scale=1 and/or offset=0.
void scaleOnPut (Float scale, Float offset, const Array<Float>& array, Array<Short>& target)
Scale and/or offset array to target.
This is meant when writing an array into the stored column.
It optimizes for scale=1 and/or offset=0.
Scale and/or offset target to array for the entire column.
When the scale and offset are fixed, it will do the entire array.
Otherwise it iterates through the array and applies the scale
and offset per row.
Scale and/or offset array to target for the entire column.
When the scale and offset are fixed, it will do the entire array.
Otherwise it iterates through the array and applies the scale
and offset per row.
Float getScale (uInt rownr)
Get the scale value for this row.
Float getOffset (uInt rownr)
Get the offset value for this row.
void findMinMax (Float& minVal, Float& maxVal, const Array<Float>& array) const
Find minimum and maximum from the array data.
NaN and infinite values are ignored. If no values are finite,
minimum and maximum are set to NaN.
void makeScaleOffset (Float& scale, Float& offset, Float minVal, Float maxVal) const
Make scale and offset from the minimum and maximum of the array data.
If minVal is NaN, scale is set to 0.
void putPart (uInt rownr, const Slicer& slicer, const Array<Float>& array, Float scale, Float offset)
Put a part of an array in a row using given scale/offset values.
void putFullPart (uInt rownr, const Slicer& slicer, Array<Float>& fullArray, const Array<Float>& partArray, Float minVal, Float maxVal)
Fill the array part into the full array and put it using the
given min/max values.
Define the "constructor" to construct this engine when a
table is read back.
This "constructor" has to be registered by the user of the engine.
If the engine is commonly used, its registration can be added
to the registerAllCtor function in DataManager.cc.
That function gets automatically invoked by the table system.