how can I speed a function in tidyr up

问题

I have a data like this

    n <- 1e5
set.seed(24)
df1 <- data.frame(query_string = sample(sprintf("%06d", 100:1000), 
 n, replace=TRUE), id.x = sample(1:n), 
                  s_val = sample(paste0("F", 400:700), n, 
        replace=TRUE), id.y = sample(100:3000, n, replace=TRUE), 
            ID_col_n = sample(100:1e6, n, replace=TRUE), total_id = 1:n)

I use the spread function to assign common strings using the following function

library(tidyr)

res <- spread(resNik,s_val,value=query_string,fill=NA)

This works perfectly but when the data is huge, it is like never going to end. I don't know if my computer is hanged or it is still running because after two hours still nothing coming up

I am wondering if one can help me to use another function or something else which works faster than spread?

回答1:

Based on the benchmarks on 1e5 rows dcast from data.table is faster

library(data.table)
system.time({res1 <- spread(df1,s_val,value=query_string,fill=NA)})
# user  system elapsed 
#   1.50    0.25    1.75 


system.time({res2 <- dcast(setDT(df1), id.x+id.y + ID_col_n +total_id~s_val,
                                  value.var = "query_string")})
# user  system elapsed 
#   0.61    0.03    0.61 

res11 <- res1 %>%
           arrange(id.x)
res21 <- res2[order(id.x)]  

all.equal(as.data.frame(res11), as.data.frame(res21), check.attributes=FALSE)  
#[1] TRUE

The difference is increased with the increase in the number of rows i.e. from changing 'n' to 1e6

system.time({res1 <- spread(df1,s_val,value=query_string,fill=NA)})
#   user  system elapsed 
# 28.64    3.17   31.91 
system.time({res2 <- dcast(setDT(df1), id.x+id.y + ID_col_n +total_id~s_val,
                                  value.var = "query_string")})
#   user  system elapsed 
#   5.22    1.08    6.21

data

n <- 1e5
set.seed(24)
df1 <- data.frame(query_string = sample(sprintf("%06d", 100:1000), 
 n, replace=TRUE), id.x = sample(1:n), 
                  s_val = sample(paste0("F", 400:700), n, 
        replace=TRUE), id.y = sample(100:3000, n, replace=TRUE), 
            ID_col_n = sample(100:1e6, n, replace=TRUE), total_id = 1:n)

来源：https://stackoverflow.com/questions/41079280/how-can-i-speed-a-function-in-tidyr-up

标签

tidyr