#/ Data:
df <- 
data.frame(order=seq(1,6),
           class=rep(c("a", "b"), each=3),
           value=c(5,6,4,10,7,18))

#/ Base R ::: https://stackoverflow.com/questions/1296646/how-to-sort-a-dataframe-by-multiple-columns
df[with(df, order(class, value)), ]

#/ dplyr ::: https://dplyr.tidyverse.org/reference/arrange.html
library(dplyr)
df %>% arrange(class, value)

As usual the base solution is much faster compared to dplyr (tidyverse). Still, with multiple filtering conditions and with operations up- and downstream of this command the tidyverse syntax / verbs can be more convenient to write and easier to read, so exactly what the tidyverse has been developed for (it has not been developed for speed, this is a common misconception from what I understand).

library(microbenchmark)
base=function(x) df[with(df, order(class, value)),]
tidy=function(x) df %>% arrange(class, value)
microbenchmark(base=base(), tidy=tidy())

Unit: microseconds
 expr      min        lq       mean    median       uq      max neval
 base   59.125   71.5165   96.27901   79.3505   88.886 1680.796   100
 tidy 1611.210 1782.8240 1941.95609 1887.5755 1937.872 7368.565   100

I'm on the the data.table camp which is usually faster and, in my opinion, more readable than base R or dplyr. In this case:

df[order(class, value), ]

dplyr (which I haven't really used) seems to have too many dedicated functions and idioms for common tasks, like arrange, filter and I feel I need to re-learn what I already know. I don;t get that with data.table like in this example.

> library(sqldf)

> df
  order class value
1     1     a     5
2     2     a     6
3     3     a     4
4     4     b    10
5     5     b     7
6     6     b    18

> sqldf("select * from df order by class, value")
  order class value
1     3     a     4
2     1     a     5
3     2     a     6
4     5     b     7
5     4     b    10
6     6     b    18

> sqldf("select * from df order by class asc, value asc")