Multiple condition `rowSums` - r

Question

I would like to perform a rowSums based on specific values for multiple columns (i.e. multiple conditions). I know how to rowSums based on a single condition (see example below) but can't seem to figure out multiple conditions.
# rowSums with single, global condition
set.seed(100)
df <- data.frame(a = sample(0:100,10),
b = sample(0:100,10),
c = sample(0:100,10),
d = sample(0:100,10))
print(df)
a b c d
1 31 63 54 49
2 25 88 71 92
3 54 27 53 34
4 5 39 73 93
5 45 73 40 67
6 46 64 16 85
7 77 19 97 17
8 34 33 82 59
9 50 93 51 99
10 15 100 25 11
Single Condition Works
df$ROWSUMS <- rowSums(df[,1:4] <= 50)
# And produces
a b c d ROWSUMS
1 31 63 54 49 2
2 25 88 71 92 1
3 54 27 53 34 2
4 5 39 73 93 2
5 45 73 40 67 2
6 46 64 16 85 2
7 77 19 97 17 2
8 34 33 82 59 2
9 50 93 51 99 1
10 15 100 25 11 3
Multiple Conditions Don't Work
df$ROWSUMS_Multi <- rowSums(df[,1] <= 50 | df[,2] <= 25 | df[,3] <= 75)
Error in rowSums(df[, 1] <= 50 | df[, 2] <= 25 | df[, 3] <= 75) :
'x' must be an array of at least two dimensions
Desired Output
a b c d ROWSUMS_Multi
1 31 63 54 49 2
2 25 88 71 92 2
3 54 27 53 34 1
4 5 39 73 93 2
5 45 73 40 67 2
6 46 64 16 85 2
7 77 19 97 17 1
8 34 33 82 59 1
9 50 93 51 99 2
10 15 100 25 11 2
I could just be sub-setting incorrectly, but I haven't been able to find a fix.

Answer 1

One problem with [ while having a single row or single column is it coerces the data.frame to a vector. Based on ?Extract
x[i, j, ... , drop = TRUE]
NOTE, drop is TRUE by default
and later in the documentation
drop - For matrices and arrays. If TRUE the result is coerced to the lowest possible dimension (see the examples). This only works for extracting elements, not for the replacement. See drop for further details.
To avoid that either use drop = FALSE or simply drop the , which will return a single column data.frame because by default, the index without any comma is regarded as column index and not row index for data.frame
rowSums(df[1] <= 50 | df[2] <= 25 | df[3] <= 75)
Update
Based on the expected output, the rowSums can be written as
dfROWSUMS <- rowSums(df[1:3] <= c(50, 25, 75)[col(df[1:3])])
df$ROWSUMS
#[1] 2 2 1 2 2 2 1 1 2 2
NOTE: Earlier comment was based on why the rowSums didn't work. Didn't check the expected output earlier. Here, we need to do comparison of 3 columns with different values. When we do
df[1] <= 50
It is a single column of one TRUE/FALSE
When we do | with
df[1] <= 50 | df[2] <= 25
It would be still be a single column of TRUE/FALSE. Only difference is that we have replaced TRUE/FALSE or FALSE/TRUE in a row with TRUE. Similarly, it would be the case when we add n logical comparisons compared with |. Instead of that, do a +, does the elementwise sum
((df[1] <= 50)+ (df[2] <= 25) + (df[3] <= 75))[,1] # note it is a matrix
Here, we can do it with vector i.e. using , as well
((df[, 1] <= 50)+ (df[, 2] <= 25) + (df[, 3] <= 75)) # vector output
The only issue with this would be to repeatedly do the +. If we use rowSums, then make sure the comparison value replicated (col) to the same dimensions of the subset of data.frame. Another option is Map,
Reduce(`+`, Map(`<=`, df[1:3], c(50, 25, 75)))

Answer 2

We can also use cbind to create a matrix from the multiple conditions using column positions or column names then use rowSums like usual, e.g
> rowSums(cbind(df[,'a'] <= 50 ,df[,'b'] <= 25 ,df[,'c'] <= 75), na.rm = TRUE)
[1] 2 2 1 2 2 2 1 1 2 2
> rowSums(cbind(df['a'] <= 50 ,df['b'] <= 25 ,df['c'] <= 75), na.rm = TRUE)
[1] 2 2 1 2 2 2 1 1 2 2
Using dplyr
library(dplyr)
df %>% mutate(ROWSUMS=rowSums(cbind(.['a'] <= 50 ,.['b'] <= 25 ,.['c'] <= 75), na.rm = TRUE))