[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
d.masterdim and slices
- To: zzdev@xxxxxxxxxx
- Subject: d.masterdim and slices
- From: Tuukka Hastrup <Tuukka.Hastrup@xxxxxx>
- Date: Mon, 24 Jul 2000 15:32:53 +0300 (EEST)
Now that I have d.masterdim and d.cellcreation implemented, I'd need
some specs about how to make this work correctly with slices. (That's
why the current code is not actually so stable with sliced spaces.)
At the moment, it works like this:
On every slice, there is a single rank along d.cellcreation starting
from the home cell of that slice. This rank lists the cells in creation
order, and can be used to enumerate through the cells of this slice.
The slice home cells form a rank along d.slices, starting from the home
cell of the space.
The space has one rank listing all the dimensions ever used in some
slice. It starts on the space home cell. No dimension is ever removed
by the code. It's user-modifialbe, though, so one can go and mess it up
:-) d.slicedims could be handy. Also, we should clearly spec witch
dimensions are listed: those loaded to gZZ (bad), those ever used by
the code (current), those having at least one connection (clear), those
never totally disconnected (interesting) etc.
(Remember the decisions should leave space for d.clone, d.version etc)
- Is there a home cell on each slice, or only one per space?
- Are slices differentiable in the space? (Is there one rank per slice
along d.cellcreation? Is there one rank per slice along d.masterdim?
etc. Are there two operations: getSpaceHome and getSliceHome)
- Is there a slice that is not a slice ;-) Can we make s.0 be otherwise
empty but just contain the information for the compound space? This
way, no need to separate getSpaceHome and getSliceHome, as getSpaceHome