mapPlot() problem with space-containing plots

[dankelley]

This is a placeholder for work I'm doing in the "dk" branch, to try to fix a problem with missing grid lines (OAR figure, left). I am not going to document very much here, because nobody else really ought to be playing with "dk" (which might not even build, at any time). Sample code:

library(oce)
data(coastlineWorld)
mapPlot(coastlineCut(coastlineWorld, -45),projection="+proj=ortho +lon_0=-45", col="gray")

It all stems from an attempt at efficiency in maps that span a small portion of the earth (for which there's no need to draw a lot of non-intersecting grid lines). If there is a lot of space around a world view, the code cannot detect any intersections, because what it's doing is to trace along the par("usr") perimeter, attempting inverse map projections. My plan is to trace a grid through the whole plot are, not just the perimeter. That should help in this case, but also I think it can help with the UHL problems, if I store information about the mapping. (The whole challenge with UHL is making it general, so the problem goes away with any legal projection, including both lon and lat shifts.)

I don't have any particular timetable for fixing this issue, which is why it's going in the dk branch, where I can experiment a lot without users getting tangled up with changes.

[richardsc]

Also notice issue #616 on the orthographic plot ... 😢

[dankelley]

Yes, when I go to my envisioned grid-based scheme, we'll know whether Antarctica should be completed with line going to the pole, or something else. I think coastlineCut could be altered, to fix up such things. I also think it could be altered -- and this is the holy grail -- to use the edge of the earth, in space-revealing plots.

[dankelley]

I think this works ok now in the "develop" branch. I've updated a blog item on this (http://dankelley.github.io/r/2015/04/03/oce-proj.html) as an additional test. This grid-based method is what I should have been doing from the start, instead of the perimeter-trace method. I think the blog is a good test suite for such things, and I'll be using it from now on.