summarizing data frame by columns in R

Question

I have this data frame df:

df <- structure(list(App = structure(c(4L, 4L, 3L, 3L, 2L, 2L, 1L), .Label = c("DB", 
"End", "Mid", "Web"), class = "factor"), Server = structure(c(5L, 
6L, 1L, 2L, 3L, 4L, 7L), .Label = c("GServer101", "Hserver103", 
"JServer100", "Kserver200", "Server101", "Server102", "Xdb101"
), class = "factor"), Process1 = c(1L, 5L, 1L, 1L, 1L, 1L, 1L
), Process2 = c(1L, 1L, 1L, 4L, 1L, 1L, 1L), Process3 = c(NA, 
NA, NA, NA, NA, NA, NA), Process4 = c(NA, NA, NA, NA, NA, NA, 
NA), Process5 = c(NA, NA, NA, 1L, 1L, 1L, 1L)), .Names = c("App", 
"Server", "Process1", "Process2", "Process3", "Process4", "Process5"
), class = "data.frame", row.names = c(NA, -7L))

I would like to be able to summarize df data frame and count and place process by columns as below. I need to know how many process each app has group by column name. How would I do this in R?

end <- structure(list(App = structure(c(4L, 3L, 2L, 1L), .Label = c("DB", 
"End", "Mid", "Web"), class = "factor"), Process1 = c(6L, 2L, 
2L, 1L), Process2 = c(2L, 5L, 2L, 1L), Process3 = c(0L, 0L, 0L, 
0L), Process4 = c(0L, 0L, 0L, 0L), Process5 = c(0L, 1L, 2L, 1L
)), .Names = c("App", "Process1", "Process2", "Process3", "Process4", 
"Process5"), class = "data.frame", row.names = c(NA, -4L))

Answer 1

Here is a method using data.table

library(data.table)
# convert df to data.table
setDT(df)

df[, lapply(.SD, sum, na.rm=TRUE), .SDcols=Process1:Process5, by="App"]
   App Process1 Process2 Process3 Process4 Process5
1: Web        6        2        0        0        0
2: Mid        2        5        0        0        1
3: End        2        2        0        0        2
4:  DB        1        1        0        0        1

Or using column positions instead of column names

df[, lapply(.SD, sum, na.rm=TRUE), .SDcols=3:7, by="App"]
   App Process1 Process2 Process3 Process4 Process5
1: Web        6        2        0        0        0
2: Mid        2        5        0        0        1
3: End        2        2        0        0        2
4:  DB        1        1        0        0        1

In case this is new, here's a quick break-down. lapply(.SD, sum, na.rm=TRUE) says sum with na.rm=TRUE across all columns, .SDcols=3:7 or .SDcols=Process1:Process5 subsets this operation to the desired columns, by=App groups the operation.

Answer 2

You can use dplyr:

library(dplyr)
df %>% 
      group_by(App) %>% 
      summarize_at(vars(starts_with("Process")), funs(sum(., na.rm=TRUE)))

# A tibble: 4 × 6
#     App Process1 Process2 Process3 Process4 Process5
#  <fctr>    <int>    <int>    <int>    <int>    <int>
#1     DB        1        1        0        0        1
#2    End        2        2        0        0        2
#3    Mid        2        5        0        0        1
#4    Web        6        2        0        0        0

Or if column positions are preferred, the positions can be passed to .cols parameter:

df %>% 
       group_by(App) %>% 
       summarize_at(.cols=3:7, funs(sum(., na.rm=TRUE)))

# A tibble: 4 × 6
#     App Process1 Process2 Process3 Process4 Process5
#  <fctr>    <int>    <int>    <int>    <int>    <int>
#1     DB        1        1        0        0        1
#2    End        2        2        0        0        2
#3    Mid        2        5        0        0        1
#4    Web        6        2        0        0        0