New nested_pipe_linter

[MichaelChirico]

Part of #884

No hits on {lintr}

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (ff1dc21) 99.38% compared to head (d1dca49) 99.39%.

❗ Current head d1dca49 differs from pull request most recent head b462119. Consider uploading reports for the commit b462119 to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2301   +/-   ##
=======================================
  Coverage   99.38%   99.39%           
=======================================
  Files         119      120    +1     
  Lines        5392     5412   +20     
=======================================
+ Hits         5359     5379   +20     
  Misses         33       33           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[AshesITR]

This one I would carve out an exception at least for select(). But feel free to ignore that.

Tests for metadata are missing.

[MichaelChirico]

This one I would carve out an exception at least for select(). But feel free to ignore that.

Tests for metadata are missing.

Equally common I think is DF1 |> inner_join(DF2 |> filter(...)).

For select() the issue becomes cases like:

DF1 |>
  inner_join(
    DF2 |>
      select(
        a, b, c, d,
        e, f, g, h, i
      )
  )

This was part of my motivation for the whole linter to begin with -- it becomes quite hard to tell at a glance how |> will parse (is it part of the DF1 |> pipeline... no... where does that nested pipeline begin exactly?).

For select(), the advantage of assigning an intermediate variable is not as clear as for filter(). so I see why that's a pretty good default.

But maybe we should add logic around whether the "nested" pipeline takes up >1 line?

[AshesITR]

> 1 line looks like a good heuristic.
Maybe better than exceptions for some functions.

[MichaelChirico]

> 1 line looks like a good heuristic. Maybe better than exceptions for some functions.

SGTM. I am thinking to make this a parameter, allow_inline = TRUE by default. FALSE is the current logic, TRUE only hits when the "inner" pipeline spans >1 line