Standardize names for steps that create dummy variables

[EmilHvitfeldt]

trafficstars

The use of dummy in step names have lead to some confusion, especially with the addition of step_dummy_multi_choice() and step_dummy_extract() which has dummy as a part of their name, while other steps such as step_regex(), step_count(), step_indicate_na(), and step_holiday() which do produce dummies, does not.

Before I go any further I'm going to lay down some terminology.

• Dummy variable: A numeric variable that only takes the value 0 or 1 that indicates a categorical effect. (also known as indicator variable)
• Count variable: A numeric variable that indicate number of occurrences. Can any whole number.

Using the above definition I will say that

• step_dummy() produces a set of dummy variables.
• step_dummy_multi_choice() produces a set of dummy variables.
• step_holiday() produces a set of dummy variables.
• step_dummy_extract() produces a set of count variables.
• step_indicate_na() produces a single dummy variable.
• step_regex() produces a single dummy variable.
• step_count() produces a single count variable (when normalize = FALSE)

A way to standardize the naming would be to turn step_holiday() -> step_dummy_holiday(), step_dummy_regex(), etc, etc.

not all dummy steps can have a related count step, but all count steps can have a related dummy step.

What I'm not sure what to do naming wise for steps that produces counts, as it is only step_count() and step_dummy_extract(). step_dummy_extract() could in theory be changed to return dummies instead of counts, and create another step called step_count_extract() that does what step_dummy_extract() does now.

All the above a using a somehow loose definition of categorical effect.

[juliasilge]

Two that are especially confusingly named right now are step_dummy_extract() and step_regex(). One option could be:

• step_regex() ➡️ step_detect_regex()
• step_dummy_extract() ➡️ step_extract_regex()
• step_count() ➡️ step_count_regex()

I feel like step_dummy_multi_choice() could be better. One option would be step_dummy_coalesce()?

I'd lean toward keeping step_holiday() as is and then adjusting the title to say it makes dummy variables:

Generate dummy variables for holidays

[EmilHvitfeldt]

Another verb we have going on here is indicate that in my mind is very similar to detect. It would be nice to avoid keeping both around

[EmilHvitfeldt]

I feel like step_dummy_multi_choice() could be better. One option would be step_dummy_coalesce()?

I like coalesce!

[EmilHvitfeldt]

overall a good idea. But I don't think the benefit of unifying these function names outweigh the annoyance we would get for changing them.

[github-actions[bot]]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex https://reprex.tidyverse.org) and link to this issue.