Possible performance tuning. #1
Description
I replaced two lines in my code base from:
var copy = JSON.parse(JSON.stringify(o))
to
var copy = universalCopy(o)
And it increased the runtime of test suite from 4 seconds to 5 seconds.
My code base is a type checker, the main place i use universalCopy() is to copy the type definition files AST, which is a large data structure.
JSON.parse(JSON.stringify()):
$ time node test/index.js | grep -v '^ok' | grep -v '^# '
pid: 28829
TAP version 13
1..12275
real 0m16.841s
user 0m16.800s
sys 0m0.256s
universal-copy:
$ time node test/index.js | grep -v '^ok' | grep -v '^# '
pid: 28711
TAP version 13
1..12275
real 0m31.591s
user 0m31.464s
sys 0m0.420s
I also created two flamegraphs:
- Looking at the wallclock time, in this stress test the performance was 50%
- in my actual test-suite the performance dropped by 20%.
- Looking at the flamegraph in detail, JSON.stringify/JSON.parse is 800% faster in isolation.
Also, from these flames, it looks like the other huge CPU hog is my naively implemented language parser using the parsimmon framework :)
It would be cool if benchmarks can be added and this library can be performance tuned.
Once I start performance tuning my application, I can optimize the copy() function and contribute the code back (it might make simplification trade-offs for performance... and no longer be "universal")