Right now the gcammaptools API takes a monolithic approach to specifying plots. For example, plot_GCAM sets the aesthetics, draws the graticule, draws the base map, colors the map if necessary, sets the color sale, applies the map projection, and sets theme elements. Bundling all that functionality into a single function makes it hard to reuse any of this functionality. For example, plot_GCAM_grid uses plot_GCAM to draw its base map, but this causes some problems with the legend title that have to be worked around, and fine-tuning the color scale is likely to be a major hassle.

What we need is a set of customized ggplot elements that we can chain together according to ggplot conventions, so we could write something like

plot <- gcam_basemap(...) + gcam_grid(...) + gcam_colorscale(...) + gcam_theme(...)

Each one of those could provide a set of sensible defaults to narrow down the range of options and provide a consistent house style for geographic plots, while still allowing us to remix as required.

It would be a useful utility to map differences between scenarios within the function rather than switching between wide and long dataframe formats.

This will allow us to remove the dependency on mapproj and eliminate the orientation argument to our plot functions.

The geojson file for GCAM-USA is missing entirely, so that needs to be updated.

The one for GCAM-China has a practical problem; it is painfully slow to render. I think this is due to excessively high resolution in the China subregion boundaries:

filter(map.chn, id<=32) %>% nrow / nrow(map.chn)
[1] 0.3350259

So, over 65% of the vertices in that data set are in the subregions (all of which have id values >= 33). I think we could do with a bit less resolution in those subregion boundaries, especially if it speeds up map rendering.

Along those lines, all of the map data could have its resolution reduced a bit. It's nice that we have all of the detail around the deltas of major rivers, all the islands of the Canadian arctic, and the fjords of Norway, but they don't really show up at the scales of our plots, and they slow the rendering dramatically.

Example:

> gcammaptools::plot_GCAM_grid(co2grid, 'value')
Error in import_mapdata(mapdata) : object 'map.rgn32' not found

When you attach the namespace it works as expected.

> library(gcammaptools)
> gcammaptools::plot_GCAM_grid(co2grid, 'value')
although coordinates are longitude/latitude, st_intersection assumes that they are planar

However, you shouldn't have to attach the package namespace in order to get the package to work. Notably, as it stands the function can't be used from another package.

The basin235 map is blocky. Update to a more detailed version of the map.

After changing the water basin variable name to 'region' to allow for use of add_region_ID the following code:

groundwater <- add_region_ID(groundwater, lookupfile = basin235)

results in a data frame that is all 0's except for 'region' and 'id'. These two variables are also reversed, id values in the region variable and vis versa, which is likely causing add_region_ID to work incorrectly.

Starting over and changing the water basin variable name to 'id' allows add_region_ID to work, but provides 'id' variables as water basin names rather than water basin numbers and 'region' as the water basin number as class character rather than numeric

Original:

                            region       total Units               scenario year
1                   Africa_Eastern  0.30006688    EJ GCAM-China-Ref-d633d8b 2020
2                  Africa_Northern  1.43266595    EJ GCAM-China-Ref-d633d8b 2020
3                  Africa_Southern  0.36476475    EJ GCAM-China-Ref-d633d8b 2020
4                   Africa_Western  0.43875343    EJ GCAM-China-Ref-d633d8b 2020
5                               AH  0.87788441    EJ GCAM-China-Ref-d633d8b 2020
6                        Argentina  0.71153013    EJ GCAM-China-Ref-d633d8b 2020
7                     Australia_NZ  1.19388696    EJ GCAM-China-Ref-d633d8b 2020
8                               BJ  0.22668671    EJ GCAM-China-Ref-d633d8b 2020
9                           Brazil  2.67441480    EJ GCAM-China-Ref-d633d8b 2020
10                          Canada  2.43183550    EJ GCAM-China-Ref-d633d8b 2020
...
18                   Europe_Non_EU  2.07783819    EJ GCAM-China-Ref-d633d8b 2020
19 European Free Trade Association  0.74806197    EJ GCAM-China-Ref-d633d8b 2020

After add_region_ID:

                            region      total Units               scenario year id
1                                          NA    EJ                   <NA>   NA  0
2                   Africa_Eastern  0.3000669    EJ GCAM-China-Ref-d633d8b 2020  2
3                  Africa_Northern  1.4326660    EJ GCAM-China-Ref-d633d8b 2020  3
4                  Africa_Southern  0.3647647    EJ GCAM-China-Ref-d633d8b 2020  4
5                   Africa_Western  0.4387534    EJ GCAM-China-Ref-d633d8b 2020  5
6                            Anhui  0.8778844    EJ GCAM-China-Ref-d633d8b 2020 44
7                        Argentina  0.7115301    EJ GCAM-China-Ref-d633d8b 2020 31
8                     Australia_NZ  1.1938870    EJ GCAM-China-Ref-d633d8b 2020  6
9                     Beijing  Shi  0.0000000    EJ                      0    0 33
10                          Brazil  2.6744148    EJ GCAM-China-Ref-d633d8b 2020  7
...
18                European_Eastern  0.0000000    EJ                      0    0 14
19                 European_Non_EU  0.0000000    EJ                      0    0 15

This is NOT good, as warnings from R CMD CHECK are generally things you need to pay attention to and fix. But:

• rgdal wouldn't install on Travis via apt-get approach
• So I installed it via r_binary_packages
• But the binary is perhaps an old version? Anyway, the vignette wouldn't build b/c of some undefined symbol error in rgeos or rgdal
• So I added '--no-build-vignettes' to the R CMD CHECK and BUILD arguments
• But this resulted in a warning of no matching vignette html/pdf file
• So, I set warnings_are_errors: false

At some point, we'd like to revisit this--either figure out how to do a source rgdal installation; or include the vignette-build products in the repo I guess.

See commits in PR #3 for various .travis.yml changes I tried.

plot_GCAM(map.rgn32, proj_type = 'EPSG', proj = 4985) gives:

How to plot pies for different fuel types based on GCAM outputs on the base map (like on mapUS49)?
like the picture below:

Another one of these darn things.

ERROR: hard-coded installation path: please report to the package maintainer and use ‘--no-staged-install’

Honestly, I'm not sure where the problem is here. All of the places where we use a path from the package data appear to be within the guidelines established here:
https://developer.r-project.org/Blog/public/2019/02/14/staged-install/index.html

Probably we'll just convert all of the stuff currently read from the package library to package data objects and be done with it.

The command plot_GCAM(map.rgn32, proj_type = 'EPSG', proj = 3395) gives the error

Error in if (needClosed && any(vapply(x, NotClosed, TRUE))) stop("polygons not (all) closed") : 
  missing value where TRUE/FALSE needed

Needs more looking into.

If you're making a difference plot, then you want 0 to be at the center of the diverging color scale, and you want the positive and negative sides to have the same scale. Here's an example where that doesn't happen:

The zero point of the color scale is all the way over in the deep blue, which is misleading (you can also see that Antarctica has been assigned a value of 0, a described in #10 ).

When calling plot_GCAM() where the gcam_df parameter contains more columns than the specified col column for plotting, the rest of the columns get filtered out. This makes sense internally, but the user may want to keep that data, for example with ggplot2::facet_wrap().

For example, calls like plot_GCAM(map.basin235, col='value', gcam_df=df, gcam_key='id') + ggplot2::facet_wrap(~scenario) fail, when that is a reasonable thing to do.

This is a basic projection that we use for examples. It should not be cut off at the North Pole.

plot_GCAM(map.rgn32, proj = "eck3")

Clean up the existing implementation to accommodate version 2 simplified structure and use cases.

Clean up existing test suite. Investigate the shape processing aspect of the choropleth function in maps.R.

Specifically, remove use of depreciated functions.

The function add_region_id() is now complicated enough to warrant several tests. We currently have none.

A number of documentation links don't work, i.e. devtools::check() is reporting errors on them. I'm about to open a PR that changes them (for now) to just \code{}, i.e. removes the links, because I want to make sure the package passes R CMD CHECK for #1 . They include

• \code{\link{mapproj::coord_map}} in diag_header:43
• \code{\link{map.usa}} in gcammaptools:36
• \code{\link{ggalt::coord_proj}} and in Map_Functions:49
• \code{\link{mapproj::coord_map}} in Map_Functions:440

The latest version adds new capabilities for plotting gridded data, but neither the quick start guide on the front page, nor the examples vignette mention it. I can put something in the quick start shortly, but to get it working in the vignette I'll need to scare up an example data set.

Hey @rplzzz want me to add continuous integration to the repo?

The examples vignette needs to be updated to work with the new changes on the alter-proj4-method branch.

Need to fix the display order of the plot_GCAM_grid function so the boundary does not impede the grid.

I see gcam14_colors is exported, but not 32 colors. Seems like it should be, but I'm just starting to look at the package, so might be my error...

I noticed that the Antarctica polygon for the 32-region GCAM shapefile is distorted at the bottom. This occurs in the Robinson, Eckert-III and Winkel-Tripel projections. I think it can be fixed by replacing the Antarctica polygon in the shapefile. I can work on this.

addRegionID assigns the null region a value of 0 in all of its columns (including the Units column). This causes Antarctica to be plotted as if it has data, instead of being drawn in neutral grey.

Error:
[1] "There was an error at - `General` - building your map:" <simpleError in process_data(map_data, data_key_field, data_col, data_key_field, shape_obj, shape_data_field, shape_key_field): unused argument (shape_key_field)>

Executed Code:

shp_file <- '/Users/d3y010/projects/gcammaptools/world_shape_file/TM_WORLD_BORDERS_SIMPL-0.3.shp'
csv_file <- '/Users/d3y010/projects/gcammaptools/data.csv'
test <- gcammaptools::choropleth(shape_data = shp_file,
                                 shape_label_field = NULL,
                                 map_data = csv_file,
                                 data_key_field = "Country",
                                 shape_key_field = "NAME",
                                 output_file = "/Users/d3y010/projects/gcammaptools/test_output.png",
                                 data_col = "X2015",
                                 map_legend_title = "Population",
                                 map_title = "2005 Population",
                                 simplify = TRUE,
                                 bin_method = "pretty",
                                 bins = 8,
                                 map_palette_reverse = FALSE,
                                 map_palette_type = "seq")

Let me know if you need version specifics.

process_data/shape/raster etc functionality to be all inclusive in one call if possible

Somewhere Antarctica as region 0 is not being added.

When installing ggalt through github, you need to include the location of CRAN repositories to get the dependent packages.

Need to fix (nudge) the map at global extent to display the full graticule

Right now the package check stage throws several warnings on the alter-proj4-method branch. I've disabled warnings_are_errors on the CI for that branch, but ultimately the package needs to pass checks with no warnings.

Maybe we should let the map FILL_COLOR be an optional parameter of plot_GCAM. It shouldn't matter too much when plotting just to show region boundaries, but may affect the look of gridded data.

The plot_GCAM_grid function crops raster data to the bounds of the map, but it is not exact. Cells that are mostly within bounds are still included which causes the edge of the plot to be pushed past the edge of the map.

With maps for GCAM USA being added to gcammaptools, we will need to provide examples on how to use it. Also, we should have an example for China as well.

Create choropleth map functionality that ingests a polygon vector dataset (e.g., shapefile, sf object, etc.) that contains a field to bin. Optionally, ingest a complementary tabular (e.g., CSV file, data.frame, etc.) dataset to supply the values as linked to the polygons by some common key.

I'd like to plot land cover by GCAM region and AEZ, but the mapping files aren't available at that level.

The current development version of ggalt did not support eck3, robinson and witntri projections resulting in the following error when using plot_GCAM:

Error in zero_range(from) : x must be length 1 or 2

The issue was fixed by installing and older commit (12/30/16):
devtools::install_github('hrbrmstr/ggalt',ref="017b4824f3")

plot_GCAM(map.rgn32, extent = EXTENT_CHINA) gives the following map:

Pakistan and Afghanistan are being filtered out, when they should be visible.