Blueprinter: Render Performance Optimisations

[midhun-thimmapuram]

Benchmark comparison:

Output on Ruby 2.6.9, MacBook Pro (15-inch, 2018, 2.2 GHz 6-Core Intel Core i7)

benchmark code

blueprinter 0.25.3 (current master)

Rehearsal -----------------------------------------------
blueprinter   0.362873   0.004129   0.367002 (  0.367198)
-------------------------------------- total: 0.367002sec

                  user     system      total        real
blueprinter   0.337027   0.001376   0.338403 (  0.338711)
Warming up --------------------------------------
         blueprinter     1.000  i/100ms
Calculating -------------------------------------
         blueprinter      2.724  (± 2.2%) i/s -     28.000  in  10.326712s
                   with 95.0% confidence
Calculating -------------------------------------
         blueprinter    63.991M memsize (     0.000  retained)
                       480.353k objects (     0.000  retained)
                         0.000  strings (     0.000  retained)

With current PR optimisations Run - 1

Rehearsal -----------------------------------------------
blueprinter   0.203004   0.004435   0.207439 (  0.207799)
-------------------------------------- total: 0.207439sec

                  user     system      total        real
blueprinter   0.192406   0.000247   0.192653 (  0.192837)
Warming up --------------------------------------
         blueprinter     1.000  i/100ms
Calculating -------------------------------------
         blueprinter      4.718  (± 3.3%) i/s -     47.000  in  10.099948s
                   with 95.0% confidence
Calculating -------------------------------------
         blueprinter    16.860M memsize (     0.000  retained)
                        80.139k objects (     0.000  retained)
                         0.000  strings (     0.000  retained)

With current PR optimisations Run - 2

Rehearsal -----------------------------------------------
blueprinter   0.219407   0.005306   0.224713 (  0.227365)
-------------------------------------- total: 0.224713sec

                  user     system      total        real
blueprinter   0.208257   0.000257   0.208514 (  0.208746)
Warming up --------------------------------------
         blueprinter     1.000  i/100ms
Calculating -------------------------------------
         blueprinter      4.405  (± 3.8%) i/s -     44.000  in  10.165494s
                   with 95.0% confidence
Calculating -------------------------------------
         blueprinter    16.799M memsize (     0.000  retained)
                        79.889k objects (     0.000  retained)
                         0.000  strings (     0.000  retained)

[midhun-thimmapuram]

@skryukov What's next for PR to be merged?

[skryukov]

We should wait for maintainers, I'm just an outsider 😅

[mcclayton]

Hi @midhun-thimmapuram, I received your message, but I'm no longer a maintainer / part of the Procore org. and therefore am unable to provide an approving review that will enable you to merge.
See: #288 (comment)

[terrence2]

I'm largely unfamiliar with the context, but there appears to be two unrelated changes here.

The performance optimization is coming from the memoization of field sorting in fields_for, which is a nice, simple, and clean improvement.

There's also a mild refactor of date time formatting that appears to not change the operations that occur per format, but does clarify the code somewhat, so is a nice patch as well.

Given that these are unrelated, I'd much prefer to have them land as separate patches, but not enough that I'd block getting this landed at all, given how long it's been open.

[midhun-thimmapuram]

terrence2

Thanks for the approval, I can create another PR moving the date time formatting changes there if it is required so that we can have them as separate patches.