Merge pattern regex adjustment

[olivroy]

Summary

2 tests are failing which I need to investigate, but opening in case you know how to fix.

Basically, the regexp was wrong in this case which caused an infinite loop. So, I added a condition which makes the case where no match occur error early. Just need to figure out why the test is failing and also how I could simplify the test.

Also needs news.

Seems like this creates problems for more than 2 <<

dplyr::tibble(
      a = 1:4,
      b = c(1, NA, 3,  NA),
      c = c(1, 2,  NA, NA),
      d = c("1", "2", NA_character_, NA_character_),
      e = c(TRUE, FALSE, NA, NA)
    ) |> gt() |>
    cols_merge(columns = c(a, b, c), pattern = "<<{1}<<{2}<<{3}>>>>>>")

Should get c("111", "2", "33", "4"), but we are erasing everything as soon as there is NA in the first column.

dplyr::tibble(
      a = 1:4,
      b = c(1, NA, 3,  NA),
      c = c(1, 2,  NA, NA),
      d = c("1", "2", NA_character_, NA_character_),
      e = c(TRUE, FALSE, NA, NA)
    ) |> gt() |>
    cols_merge(columns = c(a, b, c), pattern = "{1}<<{2}<<{3}>>>>")

Should get c("111", "2", "33", "4") in the first column..

Related GitHub Issues and PRs

• Ref: fixes When pattern manages NAs in fmt_icon() |> cols_merge(), NA gets repeatedly appended to the value. #1880

Checklist

• I understand and agree to the Code of Conduct.
• I have listed any major changes in the NEWS.
• I have added testthat unit tests to tests/testthat for any new functionality.

[rich-iannone]

Oh yes, I discovered problems with pattern and nesting like this quite a while back but didn't get to writing up an issue about it. Thank you for taking this one on (while loops are dangerous).

[olivroy]

Yeah.. I have attempted to fix this before, but I never got it. I am out of idea...

Also, how is pattern = "<<{1}<<{2}<<{3}>>>>>>" different from "<<{1}>><<{2}>><<{3}>>". All other cases seem to render fine, but the regex I am using here attemps to exclude too many consecutive <<

If they are the same, maybe we could attempt to normalize <<{1}<<{2}<<{3}>>>>>> to <<{1}>><<{2}>><<{3}>> with a deprecation warning?
(that seems tedious and not straightforward)

Another idea I had was to change < in pattern to "::opening_secondary::" and > to "::closing_secondary::" but I didn't try this..

[rich-iannone]

Also, how is pattern = "<<{1}<<{2}<<{3}>>>>>>" different from "<<{1}>><<{2}>><<{3}>>"

I believe the idea was that if {2} is missing, both {2} and {3} would be removed. Some examples:

x = 1, y = 2, z = 3

"<<{x}<<{y(NA)}<<{z}>>>>>>" -> "1"
"<<{x}<< [{y}]<<({z(NA)})>>>>>>" -> "1 [2]"
"<<{x(NA)}<<{y}<<{z}>>>>>>" -> ""

A sort of weird idea to make this work is to substitute an uncommon literal character in a run of ">>>>>>" at every second position (making ">>|>>|>>") and then removing that literal character (probably shouldn't be a | though) as the final step.

[olivroy]

This is probably the only part I am satisfied with in this PR. It prevents infinite loops from happening. The rest doesn't work as intended.