Use | more judiciously

[MichaelChirico]

Re-examined in what situations | can be used.

Till now we've often conservatively used //NODE1[ $BIG_EXPRESSION ] | //NODE2[ $BIG_EXPRESSION ], but this works fine: (//NODE1 | //NODE2)[ $BIG_EXPRESSION ].

Untested as yet but I am assuming this can have some efficiency benefit (or at least not hurt, while hopefully making the code more readable).

Marking as draft for now since infix_spaces_linter() will likely benefit and #2025 is pending, but feel free to review the existing changes. I'll also try and add some timings.

[codecov-commenter]

Codecov Report

Merging #2028 (e3819ce) into main (f9bf506) will increase coverage by 0.00%.
The diff coverage is 100.00%.

❗ Current head e3819ce differs from pull request most recent head 5463bf1. Consider uploading reports for the commit 5463bf1 to get more accurate results

@@           Coverage Diff           @@
##             main    #2028   +/-   ##
=======================================
  Coverage   98.55%   98.55%           
=======================================
  Files         113      113           
  Lines        4994     4995    +1     
=======================================
+ Hits         4922     4923    +1     
  Misses         72       72           

Files Changed	Coverage Δ
R/function_argument_linter.R	100.00% <100.00%> (ø)
R/indentation_linter.R	100.00% <100.00%> (ø)
R/infix_spaces_linter.R	100.00% <100.00%> (ø)
R/pipe_continuation_linter.R	100.00% <100.00%> (ø)
R/redundant_equals_linter.R	100.00% <100.00%> (ø)
R/vector_logic_linter.R	100.00% <100.00%> (ø)

[MichaelChirico]

Very little (if any) improvement on redundant_equals_linter(), so the readability improvement will be what pushes this change. If we think readability is harmed I can revert.

exp <- get_source_expressions("QC.R")
l <- redundant_equals_linter()

# AT HEAD
system.time(replicate(500, lapply(exp$expressions, l)))
#    user  system elapsed 
#  35.225   0.172  35.577 

# AT ae0551b
system.time(replicate(500, lapply(exp$expressions, l)))
#    user  system elapsed 
#  35.147   0.120  35.392 

Checking on indentation_linter() now but there's so much other overhead there it's hard to compare. I'll check infix_spaces_linter() after #2025 is merged.

[MichaelChirico]

Oh interesting, infix_spaces_linter() does wind up faster... 50% so!

# HEAD
system.time(replicate(500, lapply(exp$expressions, l)))
#    user  system elapsed 
# 171.315   0.280 172.313 

# d4e93e1
system.time(replicate(500, lapply(exp$expressions, l)))
#    user  system elapsed 
#  83.125   0.118  83.554

[MichaelChirico]

Lint will go away after merging #2033