fast in r and turbo in c verisons of radius of gyration calculation

[e-kotov]

Hi @caesar0301 ,

I have made two faster versions of the radius of gyration function.

See below:

remotes::install_github("e-kotov/movr@faster-rgy")
library(movr)
data(movement)
lat <- movement$lat
lon <- movement$lon

bench::mark(
  radius.of.gyration(lat, lon),
  fast.radius.of.gyration(lat, lon),
  turbo.radius.of.gyration(lat, lon),
  iterations = 10,
  relative = FALSE,
  check = TRUE,
)

# A tibble: 3 × 13
  expression                           min   median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time result memory     time       gc      
  <bch:expr>                         <bch> <bch:tm>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm> <list> <list>     <list>     <list>  
1 radius.of.gyration(lat, lon)       138ms 155.44ms      6.51    3.89MB     30.0    10    46      1.53s <dbl>  <Rprofmem> <bench_tm> <tibble>
2 fast.radius.of.gyration(lat, lon)  880µs   1.09ms    583.      3.93MB     58.3    10     1    17.15ms <dbl>  <Rprofmem> <bench_tm> <tibble>
3 turbo.radius.of.gyration(lat, lon) 816µs 826.85µs   1193.    212.08KB      0      10     0     8.38ms <dbl>  <Rprofmem> <bench_tm> <tibble>

> round(radius.of.gyration(lat, lon), 6) ==
+   round(fast.radius.of.gyration(lat, lon), 6)
[1] TRUE
> round(radius.of.gyration(lat, lon), 6) ==
+   round(turbo.radius.of.gyration(lat, lon), 6)
[1] TRUE

The results match with the original function, but are 100+ times faster. The C implementation is not only faster, but also consumes just a fraction of memory. This might make a difference when analyzing many trajectories.

I am making a draft PR, as I do not suggest to add these new functions in addition to the existing one, but to replace it.

Let me know if you like any of the implementations. I guess if your package already has C code, then it makes sense to use the C function as it is still a bit faster then the fully vectorized zero-dependency R function, and is also much more memory efficient.

[e-kotov]

@caesar0301 hi, this is a kind reminder for some feedback on this PR.