CRS("+axis=wsu") support

[dmi3kno]

Following this, axis orientation was first introduced into proj4 in 2010 and became part of standard in 2012 (listed now among common parameters

However, when I try to "turn my world upside down" I get the error. Does it follow that the proj4 version used by CRS() does not (yet?) support +axis param?

> data("wrld_simpl", package = "maptools")
> spTransform(wrld_simpl, "+proj=latlong +axis=wsd")
Error in CRS(CRSobj) : 
  northings must follow eastings: +proj=latlong +axis=wsd

[edzer]

No, this is a case that is not allowed by sp, and caught before passing stuff on to proj.4. Pkg sf allows it:

library(sf)
s  = st_as_sf(wrld_simpl)
s2 = st_transform(s, "+proj=latlong +axis=wsd")
plot(s2)

but doesn't seem to do much.

[rsbivand]

Re: email from Nick Tate 5/1-18

Using that example, I see:

library(rgdal)
data("wrld_simpl", package = "maptools")
t0 <- spTransform(wrld_simpl, CRS("+proj=cea +lat_ts=37.5 +ellps=WGS84 +datum=WGS84 +units=m +axis=wsu +no_defs +towgs84=0,0,0"))
plot(t0)

which is upside-down - is that what you wanted? This doesn't work for sf:

library(sf)
s  = st_as_sf(wrld_simpl)
s2 = st_transform(s, "+proj=cea +lat_ts=37.5 +ellps=WGS84 +datum=WGS84 +units=m +axis=wsu +no_defs +towgs84=0,0,0")
plot(st_geometry(s2))

which is not upside-down, and reverses the signs on the bounding box y-axis

> st_bbox(s2)
     xmin      ymin      xmax      ymax 
-15916580  -8011546  15916580   7960704 
> bbox(t0)
        min      max
x -15916580 15916580
y  -7960704  8011546

I tried using the with_geos argument FALSE in st_transform, but I think the outcome was the same - unsure because I had an older lwgeom attached.

I'm copying this into the GitHub issue.

Email:

Quick question: I’m playing around with various projection transformations via sf (st_transform) and ggplot.
I notice that it is possible in PROJ.4 to influence axis orientation see http://proj4.org/parameters.html#axis-orientation via the +axis parameter.
But I am unsure whether I can effect this in R.

The proj4 definition below is for the standard Hobo Dyer (from tmap definitions)
crs <- "+proj=cea +lat_ts=37.5 +ellps=WGS84 +datum=WGS84 +units=m +no_defs +towgs84=0,0,0"

P.S. it seems some comment on this has appeared.
#42
Though as Edzer notes, it doesn’t seem to do much...

I suspect what I want is to replace the default +axis=enu with +axis=wsu

[edzer]

s3 = lwgeom::st_transform_proj(s, "+proj=cea +lat_ts=37.5 +ellps=WGS84 +datum=WGS84 +units=m +axis=wsu +no_defs +towgs84=0,0,0")
plot(st_geometry(s3))

does plot the world upside down; sf::st_transform takes the GDAL path to transformations, and it seems that that silently ignores the +axis=wsu.

[rsbivand]

Good, that is what I suspected.

[edzer]

With upside down, do you imply only y is affected? In any case, they are both; +axis=esu only affects y. In the bounding box example above, xmin equals -xmax, so you cannot see the swap.

[edzer]

Interestingly, following this suggestion,

s2 = st_transform(s, "+proj=cea +lat_ts=37.5 +ellps=WGS84 +datum=WGS84 +units=m +axis=wsu +no_defs +towgs84=0,0,0 +wktext")
plot(st_geometry(s2))

does work as intended.