Add a vignette about equivalence with tidyverse

[etiennebacher]

Closes #130

I put datawizard and tidyverse chunks side by side to easily compare the syntax, for example:

(I wanted to put different background colors to each chunk to distinguish between the two syntax, but apparently it is not yet supported by pkgdown (r-lib/downlit#149) so I added comments in each chunk).

[codecov-commenter]

Codecov Report

Merging #183 (87d84e0) into main (6dbcdba) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #183   +/-   ##
=======================================
  Coverage   83.79%   83.79%           
=======================================
  Files          52       52           
  Lines        3196     3196           
=======================================
  Hits         2678     2678           
  Misses        518      518           

Help us with your feedback. Take ten seconds to tell us how you rate us.

[strengejacke]

This is ambiguous, because we don't rely on poorman anywhere in the code.

[strengejacke]

No, there are many possibilities, including unquoted column names, but only for a single variable, I think (because we don't use ... to capture comma-separated, unquoted names). But maybe this is too much to explain here, and for now, we can keep it like this.

[DominiqueMakowski]

Passing thought, maybe we should add in our style guide in easystats the fact that we prefer quoted names rather than non-standard eval (in easystats, as opposed to the tidyverse), because it's easier to program with, because it's flexible, it makes more sense to distinguish names from objects, it's a much more standard programming good practice, and that in general NSE is an eldritch invention

[etiennebacher]

(also, quoted names are much easier to work with for debugging)

[strengejacke]

Nice draft! I like the structure of this vignette very much!

[IndrajeetPatil]

This is a fantastic start, @etiennebacher! Thanks for working on this. Lovely use of pandoc's fenced div!

I am going to make some additions and add a few edits, but overall this is a good structure. I think, once finalized, we can also melt this down a bit and convert it to a JOSS paper (#59).

[etiennebacher]

Thanks @IndrajeetPatil and @strengejacke!

Before finishing this vignette, we need to improve reshape_longer and reshape_wider to match more precisely the tidyr implementations and outputs. I'll skim through datawizard later to see if I forgot some major functions, but maybe you already have some suggestions of additional functions to include?

[IndrajeetPatil]

but maybe you already have some suggestions of additional functions to include

I have added a couple of TODOs for this in the vignette. I will keep updating them as I think about more.

[IndrajeetPatil]

@etiennebacher I am changing the targeted issue for this PR to #130 because that's what the PR is targeting.

For #90, I had a very different kind of vignette in mind.
Specifically, one where we start with a messy dataset and achieve the desired result using data wrangling functions. Something similar to this article.

[etiennebacher]

For #90, I had a very different kind of vignette in mind. Specifically, one where we start with a messy dataset and achieve the desired result using data wrangling functions. Something similar to this article.

I see, I think a good example of a messy dataset could be to download a csv file from the World bank catalogue (not using the package WDI because it provides already cleaned data). We could use:

• filtering to keep countries only (remove aggregates)
• pivoting because it's one year per col
• selecting, renaming and relocating for diverse things

[bwiernik]

An example:

colors <- data.frame(
  group = c("a", "b", "c"),
  color = c("black", "forestgreen", "lightblue")
)
scale_color_manual(values = data_extract(colors, "color",  name = group))

[etiennebacher]

I think it's now ok to review this but it should be merged after #189 because the vignette uses some args of reshape_longer that are not yet in main branch.

Also, I prefer using reshape_longer and reshape_wider because to me it's easier to make the connection with pivot_*, but all other functions start with data_ so should we use data_to_long and data_to_wide in the vignette instead?

[IndrajeetPatil]

I don't have strong opinions on that, but I personally prefer showcasing data wrangling functions having the data_*() prefix because it's an easy pattern for the users to remember.

[IndrajeetPatil]

I also feel that we should remove the data_*() prefix from statistical transformation function names.

[IndrajeetPatil]

@etiennebacher I'd like to merge this soon. I think most of the TODOs have been addressed, so this should be ready.

We can keep updating it as we add more tidyverse equivalents (e.g. arrange() <-> data_arrange()).

[etiennebacher]

@IndrajeetPatil if my last commit doesn't break anything, I agree to merge this and update it later, I think it's ready

[IndrajeetPatil]

Merci beaucoup, Étienne!