Syllabus extensions, post-launch

[colinleach]

Continues from #504

Well, it's launched (!!!), and students are playing with it (only the first few concepts so far, but it's early days).

I'd be happy to slow down after the intensity of the last couple of weeks. However, there are still some obvious things to do at some point, including stuff that's already in progress.

	Concept	Exercise
complex-numbers	#390	non-urgent
factors	wip	??
classes	wip	??

We'll need factors as a prereq for any further dataframe content.

I was previously dismissive of classes, thinking it would be docs-only at best. Unfortunately, the R practice exercises may push us to take it seriously. Three exercises (robot-name, robot-simulator, dnd-character) include expect_s3_class() in the testing, and several more use S3 classes in the example solutions.

To extend dataframes, these are things we already discussed:

• Basic readr stuff, probably limited to not much more than read_csv().
• Basic tidyr stuff, such as pivot_longer() and pivot_wider().
• group_by operations.
• Simple relational joins between tables.
• Tidy evaluation vs data masking (cf add concept exercise trees-and-tribbleations #515 and [New Concept] Dataframes draft #506).

There's a lot beyond that we could do for data analysis and statistics. The R track is pretty unique within Exercism for having much of that capability already in the test runner. I see the biggest challenge as the lack of graphics, when visualizing results is a huge part of it.

[depial]

I'd be happy to slow down after the intensity of the last couple of weeks.

I may have to anyway since I'm going to be traveling for a couple weeks starting on Monday. I haven't decided whether I'll be taking a laptop with me or not, but I should at least be able to review PRs if I don't.

Unfortunately, the R practice exercises may push us to take it seriously.

I think you're point here is a good one. Just a lazy thought... it's possible to deprecate one of the practice exercises and resurrect it as a concept exercise, right? That said, I'd imagine it wouldn't be too hard to find something to port from an OOP language.

I see the biggest challenge as the lack of graphics.

I agree it would be nice to have graphics. I wish I had more impetus to experiment with this, but it has always been an aspect of data science that I've found uninteresting. Maybe I can find some motivation somewhere? :D

[colinleach]

it's possible to deprecate one of the practice exercises and resurrect it as a concept exercise

Actually, strongly discouraged. I'm hoping we can come up with a simpler equivalent for the concept.

I agree it would be nice to have graphics

There's 2 aspects to this. We'll need a way to put some static graphs in the about.md files (as in #513), but learning data science really needs lots of practice at generating graphs interactively, to see how things change. The diametric opposite of what an Exercism test runner is about!

[colinleach]

I'm going to be traveling for a couple weeks

I hope to PR the occasional draft concept, but nothing that needs urgent review.

[depial]

practice at generating graphs interactively

If you're talking about teaching students how to make them, there might be ways to test student made graphs. There's tons of info in ggplot objects (run str() on one if you haven't) if we can figure out how to extract it.

Here's a quick example

test_plot <- ggplot(mtcars, aes(x=wt, y=mpg)) +
  geom_point() +
  geom_text(
    label=rownames(data), 
    nudge_x = 0.25, nudge_y = 0.25, 
    check_overlap = T
  ) +
 labs(
    title="Test Graph"
 )

test_plot$layers

#> $geom_point
#> geom_point: na.rm = FALSE
#> stat_identity: na.rm = FALSE
#> position_identity

#> $geom_text
#> geom_text: na.rm = FALSE, parse = FALSE, check_overlap = TRUE, size.unit = mm
#> stat_identity: na.rm = FALSE
#> position_nudge

test_plot$labels$title
#> 'Test Graph'

I'm just not sure if we can get more meaningful information than a title out of this.