`plotTS` margin behaviour conflicts with `drawPalette`

[richardsc]

It appears that plotTS() imposes it's own margins regardless of what have previously been set. An example is when using drawPalette to add a colorbar.

library(oce)
data(ctd)
col <- oceColorsJet(100)[rescale(ctd[['pressure']], rlow=1, rhigh=100)]
drawPalette(range(ctd[['pressure']]), col=oceColorsJet(100))
plotTS(ctd, col=col)

[dankelley]

Yes, and even below doesn't do the right thing. I'll look into this. Thanks

library(oce)
data(ctd)
col <- oceColorsJet(100)[rescale(ctd[['pressure']], rlow=1, rhigh=100)]
drawPalette(range(ctd[['pressure']]), col=oceColorsJet(100))
plotTS(ctd, col=col, mar=par('mar'))

[dankelley]

It's not at all clear to me why plotTS() should have mar as an arg, since a person can do

par(mar=MAR)
plot(...)

just about as easily as

plot(..., mar=MAR)

but, looking at the code, I'm a bit scared to just remove the arg, since its use depends also on args inset and add (see ctd.R near line 1802).

[dankelley]

In the develop branch, I just put in a temporary debug=10 ability. It will be removed later when I can think this through.

It gets part of the way to success, but now the isopycnal labels are running into the palette. I'll look more at perhaps 3pm my time.

## TEMPORARY action for debug=10 while working on issue 317.
library(oce)
data(ctd)
col <- oceColorsJet(100)[rescale(ctd[['pressure']], rlow=1, rhigh=100)]
drawPalette(range(ctd[['pressure']]), col=oceColorsJet(100))
plotTS(ctd, col=col, debug=10) # 10 means to avoid setting mar

[dankelley]

Below is a demo of the over-pltting into palette, with above test code

[dankelley]

The code below gives better results.

## TEMPORARY action for debug=10 while working on issue 317.
library(oce)
data(ctd)
col <- oceColorsJet(100)[rescale(ctd[['pressure']], rlow=1, rhigh=100)]
drawPalette(range(ctd[['pressure']]), col=oceColorsJet(100), mai=c(0, 0.5, 0, 0.5))
plotTS(ctd, col=col, debug=10) # 10 means to avoid setting mar

[dankelley]

@richardsc -- if the prev seems ok, it will be a temporary workaround. I think a fair bit of testing is needed on inset and add before I alter mar, however.

I don't want to break a lot of code by removing the mar arg (and its default). Maybe you can test some of your code with debug=10 to see?

PS. I could also do this as

plotTS(..., mar=NULL)

which, come to think of it, is more sensible than using debug=10, but there's not much sense in doing that if I'm going to remove the mar in a day or two, anyway.

[richardsc]

The fix with the debug=10 option looks pretty good to me. I don't fully understand the reasons behind why the mar argument is specified like that -- probably just legacy coding from the early days of "oce"?

As for the rho label and palette overplotting, perhaps the default for rotateRhoLabels should be TRUE, which saves a bit of space between the axis and the palette. I guess your fix above is just to add a bit more to the palette mar/mai before calling plotTS() which is an ok solution too.

[dankelley]

In develop, I have switched it so that mar=NULL will avoid setting margins. @richardsc -- I'd like to release to CRAN in a few days, with the recent CNV bug fix, so I hope this quick change in args (from the temporary debug=10 method) is okay for you.

[richardsc]

Looks good to me!

From personal preference I'd set rotateRhoLabels=TRUE by default, but I'll leave that as your call.