PERF: avoid allocating column names where possible

[SamSaffron]

When requesting columns names from database adapters AR:Result
would dup/freeze column names, this prefers using fstrings which
cuts down on repeat allocations

Attributes that are retained keep these fstrings around for the long
term

Note, this has the highest impact on "short" result sets, eg: Topic.first where you can void allocating the number of columns * String.

See also: https://samsaffron.com/archive/2018/02/16/reducing-string-duplication-in-ruby for detailed explanation of uminus

per @rafaelfranca we are fine including uminus in master now. cc @tenderlove @sgrif

[rails-bot]

r? @rafaelfranca

(@rails-bot has picked a reviewer for you, use r? to override)

[bdewater]

Might want to change to columns = @columns.map(&:-@) for a small speed improvement.

[SamSaffron]

@bdewater yes I can confirm this is worth it to change, I will amend this PR tomorrow

https://gist.github.com/SamSaffron/805bf43016cc897c0f0c19078e7ad171

[bdewater]

It's the reason Rubocop has Style/SymbolProc. There's a lot of context in #16833

benchmark.rb for this case:

require 'benchmark/ips'

COLUMNS = %w(id foo bar baz qux test yolo)

def block
  COLUMNS.map { |c| -c }
end

def to_proc
  COLUMNS.map(&:-@)
end

Benchmark.ips do |x|
  x.report("block")   { block }
  x.report("to_proc") { to_proc }
  x.compare!
end

Warming up --------------------------------------
               block    65.000k i/100ms
             to_proc    71.129k i/100ms
Calculating -------------------------------------
               block    774.817k (± 1.4%) i/s -      3.900M in   5.034500s
             to_proc    863.802k (± 1.2%) i/s -      4.339M in   5.023677s

Comparison:
             to_proc:   863801.9 i/s
               block:   774816.7 i/s - 1.11x  slower

[rafaelfranca]

LGTM! Let me know when you update the PR to include the &:-@ change.

[SamSaffron]

@rafaelfranca done! 🎊

[rafaelfranca]

❤️