v0.6

[aaronw22]

Describe your changes

Closes #66, closes #77, closes #63, closes #72, closes #69, closes #62, closes #70, closes #65, closes #74, closes #75, closes #67, closes #68.

Do the following before requesting a review

• [x] I have written tests covering functions I have added or changed.
• [x] I have run tests and they all pass.
• [x] I have ensured all new functions will show up in the _pkgdown.yml file.
• [x] I have updated DESCRIPTION with any new package dependencies.
• [x] I have updated NEWS.md with a brief description of my changes.
• [x] I have updated the package documentation with devtools::document().
• [x] I have updated the package website with pkgdown::build_site_github_pages().

[aaronw22]

Just realised we'll need to rewrite almost all the vignettes as well.

[JackBuckley]

Add a new diverging palette for scale_..._e61. This will be helpful when making maps.

[JackBuckley]

Could we create a default base map for different cities from OSM data? I did this for Sydney and it could be handy to pre-build this into the package if the files aren't too large.

[JackBuckley]

Potentially create templates for making state based charts with zoom boxes for the capital city. I've only done this for Sydney so far and it's fiddly.

[JackBuckley]

Do we need a new version of mpanel for the spatial maps? Or maybe we just need to not apply the auto scaling for most of the spatial maps when using the save_e61 function.

[aaronw22]

RESOLVED

@JackBuckley Something in ggplot cannot handle vectors with more than 1 class.

This one is caused by graph_data$date being class IDate and Date.

Triggered on the x-var code but would be sensible to ensure it can handle these classes for both axes. Should just be a matter of adjusting the if statements to handle inputs of length > 1.

> ggplot(graph_data, aes(x = date, y = rate)) +
+   geom_line(colour = e61_teallight)
Error in if (x_var_class == "numeric" | x_var_class == "integer") { : 
  the condition has length > 1

[aaronw22]

@JackBuckley

Fixed the format problem. But...

Some post-saving functionality present in single-panel save is still missing in the mpanel version.

The (invisible) return of the functions were also not the same. I have fixed that.

There's a bit of duplicate code across the two now. Likely to be some efficiencies (and reduced future copy-pasting and error-prevention from writing more helper functions for joint functionality).

Also don't like the fact that the mpanel save lives in mpanel_e61.R while single-panel lives in save_e61.R.

[JackBuckley]

Additional issues:

• remove spacing between mpanels
• fix issue that allows subtitle to extend beyond chart width
• add manual control of label wrapping back
• plot labels are too big in mpanels
• errors when title or y label are supplied as empty strings ""

[aaronw22]

@JackBuckley I think these changes have completely stopped format_flipped_bar() from working entirely.

Presumably because format_flipped_bar() works by adding some theme arguments, which would get ignored by save_e61. One solution would be to add (yet another) argument to save_e61 that formats the flipped graph more nicely.

Ignore the above. I have a different issue with flipped bars though. Which is that they aren't being scaled correctly. See graph. This would be much too small if it were shrunk to the size of a normal graph.

Secondary comment moved to own comment.

[aaronw22]

(Moved from another comment)

Is it time we start asking whether we are putting too much into save_e61? There may be potential to reduce the number of arguments by collapsing closely-related arguments into single arguments that take vectors or named lists.

[JackBuckley]

Can't seem to reply to these comments so quote replying to this one.

I agree that save_e61 has a lot of arguments, but almost all of them are there to provide detailed manual control for the expert user. Adding them to a vector would make the function a bit cleaner, but also a bit harder to use. I'm happy with how it currently is, but also not against you condensing a few things.

(Moved from another comment)

Is it time we start asking whether we are putting too much into save_e61? There may be potential to reduce the number of arguments by collapsing closely-related arguments into single arguments that take vectors or named lists.

[JackBuckley]

Issue where legends are removed when save_e61 is used with the autoscale functionality on.

[JackBuckley]

Fix issue where save_e61() overrides changes to the chart made in theme (e.g. theme(axis.text.x = element_text(angle = 45, hjust = 1)). At the moment these are overriden and convert back to the defaults.

[JackBuckley]

Fix issue where legends are deleted from multi-panel charts

[aaronw22]

@JackBuckley something you did to the y breaks algo has broken all the tests in save_e61 fixed?

[aaronw22]

Note to self: write some snapshot tests for square legend symbols done

[aaronw22]

@JackBuckley keep_legend in theme_e61 still hasn't been documented (and I still don't know what it's for!)

[aaronw22]

@JackBuckley this is ugly

ggplot(data.frame(x = 1:2, y = c(-24.16, 15.08)),
       aes(x, y)) +
  geom_point()

[JackBuckley]

@JackBuckley keep_legend in theme_e61 still hasn't been documented (and I still don't know what it's for!)

Oh yeah I think it is no longer needed (it was addressing an issue where save_e61 was overriding all user entered theme changes)

[aaronw22]

@JackBuckley keep_legend in theme_e61 still hasn't been documented (and I still don't know what it's for!)

Oh yeah I think it is no longer needed (it was addressing an issue where save_e61 was overriding all user entered theme changes)

Ok cool I'll let you remove it/clean up