Make it faster🚀

[pocke]

This pull request improves performance.

I've confirmed it was about 6x faster in a micro benchmark and a real rails application.

Benchmark scripts

Micro benchmark

$LOAD_PATH << './lib'

require 'benchmark'
require 'trace_location'

def noop
end

def heavy
  100000.times {noop}
end

Benchmark.bmbm do |x|
  x.report{1.times{TraceLocation.trace{heavy}}}
end

With rails application

# Execute it with rails runner

require 'benchmark'
require 'trace_location'

User.first # because the first execution is really slow

Benchmark.bmbm do |x|
  x.report{1.times{TraceLocation.trace{User.first}}}
end

Benchmark results

Micro benchmark

first - 3462bd7

Rehearsal ------------------------------------
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062014061592631234.md
 13.271392   0.083000  13.354392 ( 13.374203)
-------------------------- total: 13.354392sec

       user     system      total        real
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062014061592631249.md
 14.247306   0.053228  14.300534 ( 14.321461)

Do not use binding_of_caller - 7cf5769

Rehearsal ------------------------------------
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062014061592632386.md
  9.050748   0.062487   9.113235 (  9.126239)
--------------------------- total: 9.113235sec

       user     system      total        real
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062014061592632395.md
  9.554962   0.046541   9.601503 (  9.615075)

Simplify remove_indent - febf7b0

Rehearsal ------------------------------------
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062015061592633000.md
  8.375018   0.087021   8.462039 (  8.473832)
--------------------------- total: 8.462039sec

       user     system      total        real
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062015061592633009.md
  8.795492   0.033288   8.828780 (  8.841047)

Cache method source - 900202c

Rehearsal ------------------------------------
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062015061592634643.md
  2.478327   0.013296   2.491623 (  2.494657)
--------------------------- total: 2.491623sec

       user     system      total        real
 Created at /home/pocke/ghq/github.com/yhirano55/trace_location/trace_location-2020062015061592634646.md
  2.325903   0.016581   2.342484 (  2.345369)

With rails application

I only execute the benchmark with the base commit and the latest commit for Rails app.

first - 3462bd7

Rehearsal ------------------------------------
 Created at /home/pocke/ghq/github.com/bitjourney/kibela/log/trace_location-2020062015061592635193.md
  2.301371   0.023957   2.325328 (  2.329522)
--------------------------- total: 2.325328sec

       user     system      total        real
 Created at /home/pocke/ghq/github.com/bitjourney/kibela/log/trace_location-2020062015061592635195.md
  2.426445   0.000323   2.426768 (  2.430787)

Cache method source - 900202c

Rehearsal ------------------------------------
 Created at /home/pocke/ghq/github.com/bitjourney/kibela/log/trace_location-2020062015061592635466.md
  0.575040   0.016643   0.591683 (  0.593343)
--------------------------- total: 0.591683sec

       user     system      total        real
 Created at /home/pocke/ghq/github.com/bitjourney/kibela/log/trace_location-2020062015061592635467.md
  0.363144   0.000371   0.363515 (  0.364667)

Note

This pull request contains #23 commit to avoid conflicting.

[pocke]

We can use caller_locations instead of binding.of_caller. It is faster, and no dependency.

[pocke]

UnboundMethod#source is really heavy. So caching them improves performance dramatically.

[yhirano55]

Pretty awesome!!