Euclid::Fits::ColumnInfo< T > Struct Template Reference

#include <Column.h>

Class Description

template<typename T>
struct Euclid::Fits::ColumnInfo< T >

Column metadata, i.e. { name, unit, repeatCount } and the value type as the template parameter.

Binary table columns are either scalar (repeatCount = 1) or vector (repeatCount > 1). In the case of vector columns, each cell of the column contains repeatCount values. Here is an example of a 4-row table with a scalar column and a vector column with a repeat count of 3:

Row	repeatCount = 1	repeatCount = 3
0	00	00, 01, 02
1	10	10, 11, 12
2	20	20, 21, 22
3	30	30, 31, 32

For performance, the values are stored sequentially in a 1D array as follows:

int repeat1[] = { 00, 10, 20, 30 };
int repeat3[] = { 00, 01, 02, 10, 11, 12, 20, 21, 22, 30, 31, 32 };

The only exception to this is string columns, which are vector columns – they should have a repeat count greater than the maximum number of characters in a cell – but each cell contains only one string:

Row	repeatCount = 6
0	"ZERO"
1	"ONE"
2	"TWO"
3	"THREE"

The data array is a simple array of std::strings:

std::string data[] = { "ZERO", "ONE", "TWO", "THREE", };

but the repeat count should be at least 6 (beware of the null terminating character).

Note

Since the values are stored sequentially even for vector columns, a scalar column can be "fold" into a vector column by just setting a repeat count greater than 1, and vice-versa. This trick allows writing scalar columns as vector columns, which is what CFitsIO recommends for performance. Indeed, with CFitsIO, it is much faster to write 1 row with a repeat count of 10.000 than 10.000 rows with a repeat count of 1. This is because binary tables are written row-wise in the Fits file. CFitsIO uses an internal buffer, which can be exploited to optimize reading and writing. This is generally handled through the "iterator function" provided by CFitsIO.

Fortunately, this complexity is already embedded in EleFits internals: the buffer is used optimally when reading and writing several columns. In general, it is nearly as fast to read and write scalar columns as vector columns with EleFits. Therefore, users are encouraged to consider the repeat count as a meaningful value, rather than as an optimization trick.

See also

Optimization and good practices

Data classes

Public Types
using	Value = T
	The value type.

Public Attributes
std::string	name
	Column name.
std::string	unit = ""
	Column unit.
long	repeatCount = 1
	Repeat count of the column, i.e., number of values per cell. More...

---

The documentation for this struct was generated from the following file:

• EleFitsData/EleFitsData/Column.h