[BREAKING CHANGE] remove 'SizeLimit' trait

[danieleades]

i've been playing around with simplifying by removing the 'SizeLimit' trait, and using an Option<u64> instead

this removes around 100 lines of code, and reduces the complexity of the trait bounds

[hrektts]

Thanks for the refactoring! This change slows down the benchmark in my environment. I will revisit this after #25 is landed.

[danieleades]

Thanks for the refactoring! This change slows down the benchmark in my environment. I will revisit this after #25 is landed.

interesting...
if it turns out to be statistically significant, i'd be interested in trying to figure out where the slow-down is being introduced

[danieleades]

bit of a mixed bag, but definitely seems to be a net regression-

     Running benches/bench.rs (target/release/deps/bench-90251808a77ebfbb)
Gnuplot not found, using plotters backend
lidar point msg         time:   [658.58 µs 661.62 µs 666.26 µs]
                        change: [+48.713% +51.940% +55.274%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 17 outliers among 100 measurements (17.00%)
  2 (2.00%) low mild
  2 (2.00%) high mild
  13 (13.00%) high severe

lidar point msg without encapsulation
                        time:   [658.35 µs 660.57 µs 663.21 µs]
                        change: [+49.533% +52.255% +54.988%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 16 outliers among 100 measurements (16.00%)
  7 (7.00%) high mild
  9 (9.00%) high severe

lidar point msg bincode time:   [180.26 µs 180.56 µs 180.89 µs]
                        change: [-14.068% -12.981% -11.878%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 19 outliers among 100 measurements (19.00%)
  5 (5.00%) low mild
  4 (4.00%) high mild
  10 (10.00%) high severe

string msg              time:   [485.43 ns 494.67 ns 506.65 ns]
                        change: [+3.6019% +6.4550% +9.5201%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 19 outliers among 100 measurements (19.00%)
  2 (2.00%) high mild
  17 (17.00%) high severe

string msg without encapsulation
                        time:   [482.87 ns 483.84 ns 484.97 ns]
                        change: [+2.1543% +4.6294% +6.9247%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 18 outliers among 100 measurements (18.00%)
  5 (5.00%) high mild
  13 (13.00%) high severe

string msg bincode      time:   [314.16 ns 315.70 ns 318.05 ns]
                        change: [-5.6984% -3.4561% -1.2635%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 15 outliers among 100 measurements (15.00%)
  1 (1.00%) high mild
  14 (14.00%) high severe

[danieleades]

i have to assume that the difference in the benchmarks is due to the fact that my approach is runtime-polymorphic, whereas the current approach is compile-time polymorphic. I guess this can be closed?

[hrektts]

Reducing the amount of code is important, but I would prefer to focus on performance. This policy is also consistent with changes made in the past to improve performance (#2, #6).
I will close this for now, but feel free to reopen it if you have any suggestions for improvement.