benchmark workflow -> base R

[vincentarelbundock]

No description provided.

[github-actions]

Benchmark results

💥 means that PR is more than 5% slower than main or CRAN
⚡ means that PR is more than 5% faster than main or CRAN

Click to see benchmark results

Expression	PR time (median, seconds)	% change with main	% change with CRAN	PR memory (MB)	Mem. % change with main	Mem. % change with CRAN
slopes(mod, vcov = FALSE, newdata = “mean”)	0.150	-2.6%	-0.66%	208.412	0%	0%
slopes(mod, newdata = “mean”)	0.360	-1.1%	-0.28%	212.616	0%	0%
slopes(mod, vcov = FALSE, variables = “X3”)	0.128	0%	0%	547.487	0%	0%
slopes(mod, variables = “X3”)	0.593	0.85%	0%	1086.331	0%	0%
slopes(mod, vcov = FALSE)	1.949	-1.47%	0.05%	11173.212	0%	0%
slopes(mod)	8.099	0.37%	-0.22%	21786.187	0%	0%
hypotheses(mod, hypothesis = “b3 - b1 = 0”)	0.057	-3.39%	-1.72%	4.454	0%	0%
hypotheses(mod, hypothesis = “b2^2 * exp(b1) = 0”)	0.057	-1.72%	-1.72%	4.054	0%	0%
hypotheses(mod, hypothesis = ~reference)	0.089	-2.2%	-1.11%	9.021	0%	0%
predictions(mod)	0.485	-3.96%	-0.21%	1493.499	0%	0%
predictions(mod, newdata = “mean”)	0.130	-1.52%	-2.26%	82.003	0%	0%
predictions(mod, newdata = datagrid(X2 = unique))	4.276	1.06%	-1.68%	1654.613	-0.05%	-0.02%
comparisons(mod2)	0.649	0.62%	0.31%	1238.396	0%	0%
comparisons(mod2, comparison = “dydxavg”)	0.472	-1.67%	-1.26%	993.770	0%	0%
comparisons(mod2, comparison = “eydxavg”)	1.303	-2.18%	-1.96%	1650.491	0%	0%
comparisons(mod2, comparison = “ratioavg”)	0.481	-2.04%	-2.04%	966.151	0%	0%

Generated via commit 26d114d

Download link for the artifact containing the test results: ↓ benchmark-results.zip

[vincentarelbundock]

@etiennebacher I just re-wrote the benchmark script in mostly base R. No action need. Just a FYI.

[etiennebacher]

Thanks, I had tried with data.table but unnesting was surprisingly unfriendly compared to tidyverse.

[vincentarelbundock]

Yeah, I think the pattern in data.table is a bit unintuitive, but using by= and [[1]] seems to work:

library(data.table)
url <- "https://vincentarelbundock.github.io/Rdatasets/csv/HistData/Guerry.csv"
dat <- na.omit(fread(url))

nested = dat[, list(data=list(.SD)), by=Region]

unnested = nested[, data[[1]], by = Region]

head(nested, 2)
>    Region                data
>    <char>              <list>
> 1:      E <data.table[17x23]>
> 2:      N <data.table[17x23]>

head(unnested, 2)
>    Region rownames  dept   Department Crime_pers Crime_prop Literacy Donations
>    <char>    <int> <int>       <char>      <int>      <int>    <int>     <int>
> 1:      E        1     1          Ain      28870      15890       37      5098
> 2:      E        4     4 Basses-Alpes      12935       7289       46      2733
>    Infants Suicides MainCity Wealth Commerce Clergy Crime_parents Infanticide
>      <int>    <int>   <char>  <int>    <int>  <int>         <int>       <int>
> 1:   33120    35039    2:Med     73       58     11            71          60
> 2:   23018    14238     1:Sm     76       49      5            70          12
>    Donation_clergy Lottery Desertion Instruction Prostitutes Distance  Area
>              <int>   <int>     <int>       <int>       <int>    <num> <int>
> 1:              69      41        55          46          13  218.372  5762
> 2:              37      80        32          29           2  351.399  6925
>    Pop1831
>      <num>
> 1:  346.03
> 2:  155.90