Sort results by date (@)

[wleoncio]

I often use pkgsearch to look up package submissions by me and my colleagues for our centre's monthly report. For that reason, it would be very useful for us to have the results of advanced_search() ordered by date instead of the default (score?).

I was about to submit a PR with a suggestion to add a sort_by parameter to the summary.pkg_search_result() method (maybe print() too?), but before putting in the work I'd like to check with you if that's something you're interested in seeing added to the package.

[gaborcsardi]

You can just order according to the date column, no?

❯ pkgsearch::pkg_search("testing") |> dplyr::arrange(desc(date))
- "testing" ---------------------------------- 2323 packages in 0.015 seconds -
  #     package  version by                               @ title
  1  16 vdiffr   1.0.6   Lionel Henry                    1M Visual Regressio...
  2 100 testthat 3.1.10  Hadley Wickham                  3M Unit Testing for R
  3   8 httptest 4.2.1   Neal Richardson                 4M A Test Environme...
  4  57 covr     3.6.2   Jim Hester                      6M Test Coverage fo...
  5  16 tinytest 1.4.1   Mark van der Loo                7M Lightweight and ...
  6  10 moments  0.14.1  Lukasz Komsta                   1y Moments, Cumulan...
  7  15 lmtest   0.9.40  Achim Zeileis                   2y Testing Linear R...
  8   8 lmerTest 3.1.3   Rune Haubo Bojesen Christensen  3y Tests in Linear ...
  9  12 RUnit    0.4.32  Roman Zenka                     5y R Unit Test Fram...
 10   9 nortest  1.0.4   Uwe Ligges                      8y Tests for Normality

Of course this does not mean that you are getting the latest 10 packages. Rather, it is the 10 packages with the highest rank, ordered by date.

[wleoncio]

Yep, that works, but I'd rather avoid the piping and the dplyr dependency. Perhaps it's just me? 😅

[gaborcsardi]

I am pretty sure that you can sort a data frame without piping and dplyr, though, no?

[wleoncio]

Yeeeah, but at least when I do it that wat, i.e. treating the search output as a data.frame object, print formatting gets lost.

There's probably a smarter way to do this (using with(), perhaps?), but this is the most base-R way I know how to do this:

x <- pkgsearch::pkg_search("testing")
x[order(x$date, decreasing = TRUE)

The second operation causes x to lose its pkg_search_result class, which results in bypassing its print() method.

[gaborcsardi]

This seems to work:

x <- pkgsearch::pkg_search("testing")
x[] <- x[order(x$date, decreasing = TRUE), ]
x

- "testing" ---------------------------------- 2323 packages in 0.011 seconds -
  #     package  version by                               @ title
  1  16 vdiffr   1.0.6   Lionel Henry                    1M Visual Regressio...
  2 100 testthat 3.1.10  Hadley Wickham                  3M Unit Testing for R
  3   8 httptest 4.2.1   Neal Richardson                 4M A Test Environme...
  4  57 covr     3.6.2   Jim Hester                      6M Test Coverage fo...
  5  16 tinytest 1.4.1   Mark van der Loo                7M Lightweight and ...
  6  10 moments  0.14.1  Lukasz Komsta                   1y Moments, Cumulan...
  7  15 lmtest   0.9.40  Achim Zeileis                   2y Testing Linear R...
  8   8 lmerTest 3.1.3   Rune Haubo Bojesen Christensen  3y Tests in Linear ...
  9  12 RUnit    0.4.32  Roman Zenka                     5y R Unit Test Fram...
 10   9 nortest  1.0.4   Uwe Ligges                      8y Tests for Normality

[gaborcsardi]

In any case, my point is, it does not seems like a good solution to add a sort_by argument to every function that creates a data frame. It is much better to leave this to other tools, then you can easily sort according to multiple fields, in increasing/decreasing order, using various locales, etc.

Many people use dplyr or data.table for manipulating data frames, so this is very easy for them, and if you need a no-dependency approach, that is also easy enough.

In this particular case a sort_by argument is also not so great because it suggests that pkgsearch would return the last 10 (or size) results according to release date, but that is obviously not true.

Btw. there is probably a better way to query what you want. pkgsearch is good at ranking the results, but that's not what you need here. E.g. you could use tools::CRAN_package_db().

[wleoncio]

Fair enough, thank you for the productive discussion! :)