Convert column names into row values and find sum - r

Question

I have this data frame-
input_output <- data.frame(ip_op = c('input_0', 'input_2', 'input_9', 'output_1', 'output_2', 'output_3'), a = c(1,32, 12, 246, 901, 837), b = c(284, 23, 19, 284, 9, 12), c = c(12, 8940, 379, 490, 0, 12))
ip_op a b c
1 input_0 1 284 12
2 input_2 32 23 8940
3 input_9 12 19 379
4 output_1 246 284 490
5 output_2 901 9 0
6 output_3 837 12 12
I want to create the following data frame-
input_output
type input output
1 a 45 1984
2 b 326 305
3 c 9331 502
I have tried using transpose but the column names become rownames. How do I transform this data frame?

Answer 1

Does this work:
> library(dplyr)
> library(tidyr)
> input_output %>% pivot_longer(-ip_op) %>% mutate(ip_op = str_extract(ip_op, ('input|output'))) %>% group_by(ip_op, name) %>% summarise(value = sum(value)) %>%
+ pivot_wider(names_from = ip_op, values_from = value) %>% rename(type = name)
`summarise()` regrouping output by 'ip_op' (override with `.groups` argument)
# A tibble: 3 x 3
type input output
<chr> <dbl> <dbl>
1 a 45 1984
2 b 326 305
3 c 9331 502
>

Answer 2

Try reshaping data to long, then separate the variable name to keep the desired suffix. Aggregate the values with respective groups and then reshape to wide. Here the code using tidyverse functions:
library(tidyverse)
#Code
new <- input_output %>% pivot_longer(-1) %>%
separate(ip_op,c('Var1','Var2'),sep='_') %>%
select(-Var2) %>% group_by(Var1,name) %>%
summarise(value=sum(value,na.rm = T)) %>%
pivot_wider(names_from = Var1,values_from=value)
Output:
# A tibble: 3 x 3
name input output
<chr> <dbl> <dbl>
1 a 45 1984
2 b 326 305
3 c 9331 502

Answer 3

Here is a dplyr and tidyr solution.
library(dplyr)
library(tidyr)
input_output %>%
mutate(ip_op = sub("_.*$", "", ip_op)) %>%
group_by(ip_op) %>%
summarise(across(a:c, ~sum(.x, na.rm = TRUE)), .groups = "keep") %>%
pivot_longer(
cols = a:c,
names_to = 'type',
values_to = 'value'
) %>%
pivot_wider(
id_cols = type,
names_from = ip_op,
values_from = value
)
## A tibble: 3 x 3
# type input output
# <chr> <dbl> <dbl>
#1 a 45 1984
#2 b 326 305
#3 c 9331 502

Answer 4

Here is a base R option using aggregate + reshape like below
u <- aggregate(
. ~ ip_op,
transform(
input_output,
ip_op = gsub("_\\d+", "", ip_op)
),
sum
)
reshape(
cbind(stack(u[-1]), type = u$ip_op),
direction = "wide",
idvar = "ind",
timevar = "type"
)
which gives
ind values.input values.output
1 a 45 1984
3 b 326 305
5 c 9331 502

Answer 5

We can use data.table
library(data.table)
dcast(melt(setDT(input_output), id.var = 'ip_op',
variable.name = 'type')[, ip_op := sub("_.*", "", ip_op)],
type ~ ip_op, value.var = 'value', sum)
-output
# type input output
#1: a 45 1984
#2: b 326 305
#3: c 9331 502
Or using transpose
data.table::transpose(setDT(input_output)[, ip_op := sub("_\\d+$", "", ip_op)][,
lapply(.SD, sum), ip_op], make.names = 'ip_op', keep.names = 'type')
-output
# type input output
#1: a 45 1984
#2: b 326 305
#3: c 9331 502
Or with tidyverse and data.table::transpose
library(dplyr)
library(stringr)
input_output %>%
group_by(ip_op = str_remove(ip_op, '_\\d+$')) %>%
summarise(across(everything(), sum, na.rm = TRUE), .groups = 'drop') %>%
data.table::transpose(make.names = 'ip_op', keep.names = 'type')
# type input output
#1 a 45 1984
#2 b 326 305
#3 c 9331 502