Use pivot_longer() and pivot_wider() instead of gather() and spread()

[matt-dray]

Lesson: Dataframe Manipulation with tidyr.

Suggestion: replace spread() and gather() with pivot_longer() and pivot_wider().

Reason: these new, more intuitive functions are intended to be the de facto approach to making data 'longer' and 'wider' as of {tidyr} version 1.0.0 (though gather() and spread() continue to work).

[matt-dray]

I've had a look at doing this in a branch of a fork of the repo (link valid at time of posting).

Looks like outputs from the pivot functions differ from gather() and spread(): they have tibble class as well as data.frame and are ordered differently. This means some of the figures need updating as well as the text.

For example, consider the gap_long object from the 'From wide to long format with gather()' section, which was created using gather(). str(gap_long) gives:

'data.frame':	5112 obs. of  4 variables:
 $ continent   : chr  "Africa" "Africa" "Africa" "Africa" ...
 $ country     : chr  "Algeria" "Angola" "Benin" "Botswana" ...
 $ obstype_year: chr  "pop_1952" "pop_1952" "pop_1952" "pop_1952" ...
 $ obs_values  : num  9279525 4232095 1738315 442308 4469979 ...

If the same code is re-run with pivot_longer() instead of gather(), the output of str(gap_long) looks like this:

Classes 'tbl_df', 'tbl' and 'data.frame':    5112 obs. of  4 variables:
$ continent   : chr  "Africa" "Africa" "Africa" "Africa" ...
$ country     : chr  "Algeria" "Algeria" "Algeria" "Algeria" ...
$ obstype_year: chr  "pop_1952" "pop_1957" "pop_1962" "pop_1967" ...
$ obs_values  : num  9279525 10270856 11000948 12760499 14760787 ...

So pivot_longer() yielded the tibble class and was ordered by continent-country.

[jcoliver]

Hmmm...interesting. Any reason why the ordering in the output is different between gather (continent, obstype_year, country) and pivot_longer (continent, country, obstype_year)?

[katrinleinweber]

Related: datacarpentry/r-socialsci#104