tidyverse style compliance

[lorenzwalthert]

I have noted a few times that code in articles does not correspond to the tidyverse style guide. I can't find any note on the style guide to use in the README.md either. I think we could improve on coding style in various ways:

• add a note in the README.md to style code before submission with {styler}.
• recommend the use of tidy = 'styler' in R code chunks in .Rmd files or set in the first chunk for all subsequent R chunks, i.g. knitr::opts_chunk$set(tidy = 'styler').
• integrate the above suggestion directly into the index.Rmd template that is created with hugodown::use_tidy_post("short-name") (not sure the issue should be transferred to the {hugodown} repo then).

[hadley]

Do you have any examples?

[lorenzwalthert]

I applied styler to the 2020 posts in https://github.com/lorenzwalthert/tidyverse.org/commit/d861f38d6440c69314abe350726df768d5d0b622, but the diff is not so useful since it mostly contains end of line space formatting and other artifacts.

Reviewing the last 18 posts superficially manually, here is what I found:

I think the most prominent example is with indention and line breaks:

• https://www.tidyverse.org/blog/2020/12/finetune-0-0-1/

model <- 
  nearest_neighbor(neighbors = tune(), weight_func = tune(), 
                   dist_power = tune()

• https://www.tidyverse.org/blog/2020/11/tidymodels-sparse-support/

wide_reviews <- tidy_reviews %>%
  select(product, word, tf_idf) %>%
  pivot_wider(names_from = word, names_prefix = "word_",
              values_from = tf_idf, values_fill = 0)

• https://www.tidyverse.org/blog/2020/09/usemodels-0-0-1/

impairment_spec <- 
  boost_tree(trees = tune(), min_n = tune(), tree_depth = tune(), learn_rate = tune(), 
    loss_reduction = tune(), sample_size = tune())

Or things like

glmnet_grid <- tidyr::crossing(penalty = 10^seq(-6, -1, length.out = 20), mixture = c(0.05, 
    0.2, 0.4, 0.6, 0.8, 1)) 

in https://www.tidyverse.org/blog/2020/09/usemodels-0-0-1/

Then, there is minor ones like

1 %>% identity() %>% invisible() %>% identity()

in https://www.tidyverse.org/blog/2020/11/magrittr-2-0-is-here/ and others.

Obviously the world has bigger problems, but implementing one of my suggestions seem a reasonable cost/benefit ratio. In addition, when authors know styling is taken care of, they can spend their attention on the text and what the code does, not what it looks like.