BCF-991 optimize JSON unmarshal'ing for EVM block

[krehermann]

Use go/codec code generation to optimize JSON unmarshal'ing
of EVM block.

Benchmark comparison:

TL;DR ~50% runtime reduction, and 30% memory reduction
Benchmark obtained with

/usr/local/go/bin/go test -benchmem -run=^$ -bench="^(BenchmarkBlock_Small_JSONUnmarshal|BenchmarkBlock_Medium_JSONUnmarshal|BenchmarkBlock_Large_JSONUnmarshal|BenchmarkBlock_XL_JSONUnmarshal)$" -count=10 github.com/smartcontractkit/chainlink/core/chains/evm/types > ~/work/jira/BCF-991-json/<tag>.txt

Result:
benchstat ~/work/jira/BCF-991-json/std.txt ~/work/jira/BCF-991-json/codec.txt

benchstat ~/work/jira/BCF-991-json/std-refactor.txt ~/work/jira/BCF-991-json/codec-refactor.txt 
goos: darwin
goarch: arm64
pkg: github.com/smartcontractkit/chainlink/core/chains/evm/types
                              │ /Users/kreherma/work/jira/BCF-991-json/std-refactor.txt │ /Users/kreherma/work/jira/BCF-991-json/codec-refactor.txt │
                              │                         sec/op                          │              sec/op                vs base                │
Block_Small_JSONUnmarshal-10                                               10.177µ ± 1%                         5.740µ ± 2%  -43.60% (p=0.000 n=10)
Block_Medium_JSONUnmarshal-10                                               222.0µ ± 0%                         101.0µ ± 1%  -54.49% (p=0.000 n=10)
Block_Large_JSONUnmarshal-10                                               1741.2µ ± 0%                         786.5µ ± 1%  -54.83% (p=0.000 n=10)
Block_XL_JSONUnmarshal-10                                                  13.931m ± 1%                         6.342m ± 0%  -54.47% (p=0.000 n=10)
geomean                                                                     483.8µ                              231.9µ       -52.07%

                              │ /Users/kreherma/work/jira/BCF-991-json/std-refactor.txt │ /Users/kreherma/work/jira/BCF-991-json/codec-refactor.txt │
                              │                          B/op                           │               B/op                 vs base                │
Block_Small_JSONUnmarshal-10                                               1.820Ki ± 0%                        2.945Ki ± 0%  +61.80% (p=0.000 n=10)
Block_Medium_JSONUnmarshal-10                                              48.04Ki ± 0%                        23.91Ki ± 0%  -50.22% (p=0.000 n=10)
Block_Large_JSONUnmarshal-10                                               383.2Ki ± 0%                        188.2Ki ± 0%  -50.89% (p=0.000 n=10)
Block_XL_JSONUnmarshal-10                                                  2.843Mi ± 0%                        1.922Mi ± 0%  -32.40% (p=0.000 n=10)
geomean                                                                    99.38Ki                             71.47Ki       -28.09%

                              │ /Users/kreherma/work/jira/BCF-991-json/std-refactor.txt │ /Users/kreherma/work/jira/BCF-991-json/codec-refactor.txt │
                              │                        allocs/op                        │             allocs/op              vs base                │
Block_Small_JSONUnmarshal-10                                                 45.00 ± 0%                          41.00 ± 0%   -8.89% (p=0.000 n=10)
Block_Medium_JSONUnmarshal-10                                                867.0 ± 0%                          540.0 ± 0%  -37.72% (p=0.000 n=10)
Block_Large_JSONUnmarshal-10                                                6.701k ± 0%                         4.127k ± 0%  -38.41% (p=0.000 n=10)
Block_XL_JSONUnmarshal-10                                                   53.30k ± 0%                         32.80k ± 0%  -38.46% (p=0.000 n=10)
geomean                                                                     1.932k                              1.316k       -31.90%

[github-actions]

I see that you haven't updated any CHANGELOG files. Would it make sense to do so?

[jmank88]

This file looks pretty gnarly. Is it possible/would it make sense to have it in an internal/ subdirectory?

[krehermann]

maybe?

if so, it will be a bigger change. the generated methods need to be in the same package as the original types.

it almost works by moving blockInternal and transactionInternal to /internal/foo BUT blockInternal depends on Transaction, not transactionInternal.

so i think we would need another translation between []transactionInternal and []Transaction

[jmank88]

nbd in that case

[krehermann]

actually, i already did it.

as a nice side effect, the code is even faster now. my guess is the my original code gen had a bunch of extra unneeded stuff.

i like the code layout more now, too

[cl-sonarqube-production]

SonarQube Quality Gate

81.0% Coverage on New Code (is less than 90%)
1 New Critical Issues (is greater than 0)

See analysis details on SonarQube

Fix issues before they fail your Quality Gate with SonarLint in your IDE.