J.1.3 File size - float vs. double

For a really large number of bodies (to take the example of particle pushers) filesize becomes important for many analysis programs which become I/O bound:

$\bullet$ It's not always necessary to keep information in double precision. Images are also stored in doubles, in good faith with old C, where all math is intrinsically done in double anyhow. This would save a factor of 2 in space. There exist data i/o routines with force float/double conversions (get_data_coerced()).

$\bullet$ It's not always needed to keep all 6 phase space coordinates, besides the data structure of snapshots has phase coordinates rather well tied in, it's not easy yet to separate positions and velocities, and only store positions in a datafile. This would save another factor of 2.

$\bullet$ Images: totally unneeded to have them in double precision. Should become 'float'. Since we have used the real type