feat(bench): add oneTypeof, oneLiner and twoLiner tests

[pi0]

Hi! I've added 3 new variants to the benchmarks: oneTypeof, oneLiner and twoLiner.

Results

All tests was runned on MacBook Pro / 10.14. It seems that twoLiner is the winner in all runs.

Node 8

Run 1:

twoTypeof (current) x 3,550,837 ops/sec ±1.11% (87 runs sampled)
oneTypeof x 3,771,015 ops/sec ±0.72% (96 runs sampled)
oneLiner x 3,616,855 ops/sec ±0.65% (96 runs sampled)
twoLiner x 3,734,022 ops/sec ±1.16% (90 runs sampled)
if x 3,780,377 ops/sec ±0.91% (94 runs sampled)
negation x 3,498,913 ops/sec ±1.84% (91 runs sampled)
equals x 3,223,031 ops/sec ±3.88% (83 runs sampled)
equalsNot x 3,590,390 ops/sec ±1.15% (90 runs sampled)
switch x 3,589,515 ops/sec ±1.29% (90 runs sampled)
object x 2,083,978 ops/sec ±1.04% (89 runs sampled)
object - in x 2,308,644 ops/sec ±2.00% (87 runs sampled)
object - own x 2,481,388 ops/sec ±2.18% (84 runs sampled)
array x 1,343,607 ops/sec ±3.54% (85 runs sampled)
Fastest is if,twoLiner

Run 2:

twoTypeof (current) x 3,223,350 ops/sec ±4.97% (83 runs sampled)
oneTypeof x 3,373,441 ops/sec ±2.44% (84 runs sampled)
oneLiner x 3,530,979 ops/sec ±1.20% (88 runs sampled)
twoLiner x 3,789,479 ops/sec ±0.70% (95 runs sampled)
if x 3,676,408 ops/sec ±2.40% (92 runs sampled)
negation x 3,646,774 ops/sec ±0.57% (92 runs sampled)
equals x 3,672,144 ops/sec ±1.05% (94 runs sampled)
equalsNot x 3,742,456 ops/sec ±1.10% (91 runs sampled)
switch x 3,687,920 ops/sec ±1.17% (91 runs sampled)
object x 2,151,595 ops/sec ±0.98% (91 runs sampled)
object - in x 2,474,402 ops/sec ±0.87% (90 runs sampled)
object - own x 2,733,454 ops/sec ±0.87% (93 runs sampled)
array x 1,494,727 ops/sec ±0.69% (95 runs sampled)
Fastest is twoLiner,if

Node 10

Run 1:

twoTypeof (current) x 117,488,323 ops/sec ±0.93% (88 runs sampled)
oneTypeof x 117,681,424 ops/sec ±1.27% (89 runs sampled)
oneLiner x 117,480,971 ops/sec ±1.35% (92 runs sampled)
twoLiner x 118,184,360 ops/sec ±1.00% (91 runs sampled)
if x 28,237,352 ops/sec ±0.98% (89 runs sampled)
negation x 28,604,420 ops/sec ±0.66% (94 runs sampled)
equals x 33,609,823 ops/sec ±0.94% (91 runs sampled)
equalsNot x 48,137,146 ops/sec ±1.67% (88 runs sampled)
switch x 17,765,326 ops/sec ±1.20% (90 runs sampled)
object x 4,586,132 ops/sec ±0.76% (92 runs sampled)
object - in x 5,112,496 ops/sec ±0.59% (90 runs sampled)
object - own x 6,853,258 ops/sec ±1.64% (93 runs sampled)
array x 6,328,857 ops/sec ±1.70% (92 runs sampled)
Fastest is twoLiner,twoTypeof (current),oneTypeof,oneLiner

Run 2:

twoTypeof (current) x 117,605,753 ops/sec ±1.01% (88 runs sampled)
oneTypeof x 117,868,886 ops/sec ±0.84% (89 runs sampled)
oneLiner x 118,570,718 ops/sec ±0.75% (91 runs sampled)
twoLiner x 117,264,741 ops/sec ±1.00% (89 runs sampled)
if x 28,596,513 ops/sec ±1.04% (92 runs sampled)
negation x 28,212,960 ops/sec ±0.84% (92 runs sampled)
equals x 33,689,471 ops/sec ±0.66% (93 runs sampled)
equalsNot x 48,890,665 ops/sec ±0.83% (90 runs sampled)
switch x 18,495,350 ops/sec ±0.63% (94 runs sampled)
object x 4,184,827 ops/sec ±3.86% (86 runs sampled)
object - in x 4,790,887 ops/sec ±2.08% (88 runs sampled)
object - own x 6,980,963 ops/sec ±0.79% (91 runs sampled)
array x 6,388,749 ops/sec ±1.16% (92 runs sampled)
Fastest is oneLiner,oneTypeof,twoTypeof (current),twoLiner

[pi0]

Node 0.x is failing because of usage of const in tests. Should we keep it?

[jonschlinkert]

Should we keep it?

I don't feel strongly either way, but I think it's okay if we drop support for Node.js 0.x. Thoughts?

Thanks for doing this!

[pi0]

@jonschlinkert Friendly reminder about this PR :D

[jonschlinkert]

Can't believe I forgot about this PR for so long. Sorry @pi0! Before I merge, did you want to update the main export to use the function that was fastest in benchmarks? The current function did pretty well, so I'm fine with either way.

[pi0]

NP @jonschlinkert. Sure will update branch 😊