Merging contents of columns using apply or other vectorized approach

Question

I have a data.frame that is almost entirely blanks, but each row has one value. How can I use a vectorized or other r-vernacular approach to merge the contents of each row into a single vector?

sample data:

raw_data  <- structure(
    list(
        col1 = c("", "", "", "", ""),
        col2 = c("", "", "", "", ""),
        col3 = c("", "", "", "", ""),
        col4 = c("", "", "", "Millburn - Union", ""),
        col5 = c("", "", "Cranston (aka Garden City Center)", "",""),
        col6 = c("", "", "", "", ""),
        col7 = c("", "", "", "", ""),
        col8 = c("", "", "", "", "Colorado Blvd"), 
        col9 = c("", "", "", "", ""),
        col10 = c("", "", "", "", ""),
        col11 = c("Palo Alto",  "Castro (aka Market St)", "", "", "")
    ),
    .Names = c("col1", "col2", "col3", "col4", "col5", "col6", "col7", "col8", "col9", "col10", "col11"),
    row.names = c(5L, 4L, 3L, 2L, 1L),
    class = "data.frame"
)

This is what I tried but it fails, as it returns a 2-dimensional matrix instead of the desired vector:

raw_data$test <-  apply(raw_data, MAR=1, FUN=paste0)

Answer 1

You can do this very simply with a single index operation:

raw_data[raw_data!='']

Demo:

R> raw_data[raw_data!=''];
[1] "Millburn - Union"                  "Cranston (aka Garden City Center)" "Colorado Blvd"                     "Palo Alto"                         "Castro (aka Market St)"

If you care about the vector order being top-to-bottom (as opposed to left-to-right then top-to-bottom, which is what the above operation does), you can transpose the input data.frame:

R> t(raw_data)[t(raw_data)!=''];
[1] "Palo Alto"                         "Castro (aka Market St)"            "Cranston (aka Garden City Center)" "Millburn - Union"                  "Colorado Blvd"