Strategy for low level direct access to sparse FITS data

[ChrisArnault]

=> We consider the BinTable use case ie. N columns x M rows

We want to only access [a subset of columns] AND/OR [a subset of rows]

ie. Is it possible to specify:

"select (col1, col2, ..., colp) from file"
"select (row1, row2, ..., rowq) from file"
"select (col1, col2, ..., colp) and (row1, row2, ..., rowq) from file"

so that only relevant data bytes are really read ?

• This must be a very low level mechanism
• Ensure data locality => assume some shuffle
• How to evaluate the threshold of data proportion where the mechanism would become unefficient.
• Is it feasible in cfitsio ?

[JulienPeloton]

See also #18