Advanced debugging

Parallel Programming in R

Nabeel Imam

Data Scientist

The default error behavior

A programmer is frustrated looking at the screen, and has tossed all the papers on the desk.

Toy example

var_list <- list(1:10,
                 1:10,
                 c("a", "b", "c"))


lapply(var_list, sqrt)

Error in FUN(X[[i]], ...) : non-numeric
argument to mathematical function

Numeric
Numeric
Character

Ambiguity

A anthropomorphic computer is confused as it's screen shows an error.

¹ Designed by Freepik

Catching the error

sqrt_custom <- function(var) {

  tryCatch(
    # Expression to evaluate
    sqrt(var),

    # What to do with an error
    error = function (e) return(e)
    )

}

A worker catches and fixes an error using a wrench.

¹ Designed by Freepik

Catching the error

lapply(var_list, sqrt_custom)

[[1]]
 [1] 1.000000 1.414214 1.732051 2.000000 2.236068 2.449490 2.645751 2.828427 3.000000
[10] 3.162278

[[2]]
 [1] 1.000000 1.414214 1.732051 2.000000 2.236068 2.449490 2.645751 2.828427 3.000000
[10] 3.162278

[[3]]
<simpleError in sqrt(var): non-numeric argument to mathematical function>

Catching errors in parallel

cl <- makeCluster(3)
parLapply(cl, var_list, sqrt_custom)
stopCluster(cl)

[[1]]
 [1] 1.000000 1.414214 1.732051 2.000000 2.236068 2.449490 2.645751 2.828427 3.000000
[10] 3.162278

[[2]]
 [1] 1.000000 1.414214 1.732051 2.000000 2.236068 2.449490 2.645751 2.828427 3.000000
[10] 3.162278

[[3]]
<simpleError in sqrt(var): non-numeric argument to mathematical function>

The births data

print(ls_births)

$AK
 month  plurality
     1          1
     2          1
 ...
$AL
   month  plurality
      10          1
       9          1
 ...
...

The summarizing function

summarise_births <- function(df) {


  tryCatch({

    df %>% 
      group_by(month) %>% 
      summarise(total = sum(plurality))

  },

    error = function (e) "Error! Check data")


}

Parallel apply

cl <- makeCluster(4)
clusterEvalQ(cl, library(dplyr))

parLapply(cl, ls_births, summarise_births)

stopCluster(cl)

$AK
    month total
  1     1     7
  2     2    13
  ...

$AL
[1] "Error! Check data"

$AR
    month total
  1     1    28
  2     2    28
  ...
...

Examine the source

head(ls_births[["AL"]], n = 10)

     month plurality
263     10         1
335      9         1
473     12         1
474      6         1
475      9         1
839      9         1
1291    11     Twins
1369     4         1
1609     5         1
1610     5  Triplets

Future map

plan(multisession, workers = 4)
config <- furrr_options(packages = "dplyr")
future_map(ls_births, summarise_births,
           .options = config)
plan(sequential)

$AK
    month total
  1     1     7
  2     2    13
  ...

$AL
[1] "Error! Check data"

$AR
    month total
  1     1    28
  2     2    28
  ...
...

The foreach case

cl <- makeCluster(4)
registerDoParallel(cl)

foreach(df = ls_births,
        .packages = "dplyr"
       ) %dopar% {
  summarise_births(df)
}

stopCluster(cl)

$AK
    month total
  1     1     7
  2     2    13
  ...

$AL
[1] "Error! Check data"

$AR
    month total
  1     1    28
  2     2    28
  ...
...

Let's practice!

Parallel Programming in R