This was one of the outstanding features that we didn't get done in time for 4.2. And it turned out to be on the one hand not very hard, on the other hand fairly instrumental to finding quite a few bugs that were still lurking here and there in the planner code. I think more of the commits in this series end up fixing old issues than adding new capabilities. And certainly 3/4 of the time working on this was spent on observing the issues, understanding their causes and coming up with reasonable fixes for them.
That said, as usual I have hidden some very cute bugs in the new code, just to make sure that you guys read what I commit and test it. So please get started, as always the first one to find a bug in my code will win a valuable price! So where are the limitations of the current implementation? It seems to work quite well in my testing (but who knows once Henrik, Joakim, Robert and the other usual suspects start doing their testing). But there is one rather interesting issue: right now we don't save whether a sample was manually entered or created by our deco (or ascent) algorithm. Or more precisely, we've never saved this in the past and we still don't save this information to disk. Which means that when you re-plan such a dive the dive points table will contain ALL waypoints of the dive, including the ascent and deco stops (and all such points will have handlers in the profile). This is of course undesirable (even though it isn't really hard to get rid of them - especially in the dive points table it's really easy to just delete them all and they get automagically recreated by the planning algorithm). For now I have only implemented an in memory solution. The idea is that while you are planning a dive you may go back and forth and try multiple things. And as long as you don't save the planned dive, restart Subsurface and read it in again, you can continue to make changes as you'd expect. Only the dive data points that you entered before are in the table and have handles in the profile. All the other poinst (gas changes, ascent, deco stops, transitions) will get recalculated on the fly as if you never left planning mode. But once the planned dive is written to disk and read back we are back to what I describe above - ALL waypoints are treated as if they were user entered. I have a hunch that this is sufficient for most reasonable use cases, but the simply fact how hard it is to explain this makes me wonder if we shouldn't save this data to disk, anyway. It obviously would be easy to add to both XML and git save and parse routines... but I didn't want to add this until the likely users of this feature have commented if that would be important to them. And until all of you had some time to poke holes into my architecture and implementation for this. /D _______________________________________________ subsurface mailing list subsurface@hohndel.org http://lists.hohndel.org/cgi-bin/mailman/listinfo/subsurface
