data/binding: optimized convertion using strconv

[Jacalz]

Description:

This updates the data bindings to use the strconv package instead of fmt for conversions, leading to improved performance (benchmarks to come). This is currently WIP and is just a proof of concept with generations scripts yet to have been updated to take this into effect.

Fixes #1735

Checklist:

• Tests included.
• Lint and formatter run with no errors.
• Tests all pass.

[andydotxyz]

Looks like a good change thanks.
I wonder (random thought) should we also use this version when someone manually specifies the default format?

[Jacalz]

Here are the results of the benchmarks for this change. The benchmarks are adaptations of the corresponding tests with asserts removed. The metric are thus results of 4 .Get() and 3 .Set() calls being made on the underlying types.

name             old time/op    new time/op    delta
BoolToString-8     11.9µs ± 1%     5.7µs ± 2%  -52.02%  (p=0.008 n=5+5)
FloatToString-8    14.6µs ± 1%     7.6µs ± 0%  -47.85%  (p=0.016 n=5+4)
IntToString-8      12.3µs ± 0%     6.1µs ± 1%  -50.52%  (p=0.008 n=5+5)
StringToBool-8     13.5µs ± 2%     7.6µs ± 0%  -43.77%  (p=0.016 n=5+4)
StringToFloat-8    15.4µs ± 1%     8.8µs ± 0%  -42.63%  (p=0.008 n=5+5)
StringToInt-8      14.1µs ± 1%     7.6µs ± 0%  -46.01%  (p=0.008 n=5+5)

name             old alloc/op   new alloc/op   delta
BoolToString-8       486B ± 0%      377B ± 0%  -22.43%  (p=0.008 n=5+5)
FloatToString-8      575B ± 0%      462B ± 0%  -19.65%  (p=0.008 n=5+5)
IntToString-8        511B ± 0%      398B ± 0%  -22.11%  (p=0.008 n=5+5)
StringToBool-8       551B ± 0%      447B ± 0%  -18.82%  (p=0.008 n=5+5)
StringToFloat-8      622B ± 0%      493B ± 0%  -20.79%  (p=0.008 n=5+5)
StringToInt-8        575B ± 0%      461B ± 0%  -19.77%  (p=0.008 n=5+5)

name             old allocs/op  new allocs/op  delta
BoolToString-8       19.0 ± 0%      13.0 ± 0%  -31.58%  (p=0.008 n=5+5)
FloatToString-8      22.0 ± 0%      17.0 ± 0%  -22.73%  (p=0.008 n=5+5)
IntToString-8        18.0 ± 0%      13.0 ± 0%  -27.78%  (p=0.008 n=5+5)
StringToBool-8       20.0 ± 0%      15.0 ± 0%  -25.00%  (p=0.008 n=5+5)
StringToFloat-8      22.0 ± 0%      17.0 ± 0%  -22.73%  (p=0.008 n=5+5)
StringToInt-8        20.0 ± 0%      15.0 ± 0%  -25.00%  (p=0.008 n=5+5)

In summary, conversions are about twice as fast, we are allocating about 20% less each time. I'd say that it is a pretty good improvement 🙂

[Jacalz]

I was unsure if I should leave my quick and dirt benchmarks in or not. They are there now, but I can easily remove them if we think that's a better idea.