names(U1)
[1] "username" "review_count" "forum_posts" "age" "avg_interval"
[6] "avg_sim" "class"
So how do I create an empty data frame U1.RN that will have same columns as U1?
-
4Can I ask why you need a 0-row data frame? Depending on what you are going to do with it, it might be more efficient to do things a different way (e.g. I hope you aren't planning on filling this row by row in a loop?)Gavin Simpson– Gavin Simpson2010-11-24 16:40:40 +00:00Commented Nov 24, 2010 at 16:40
-
"e.g. I hope you aren't planning on filling this row by row in a loop?" - yeah, :(. What is the R-y way to do the equiv of [pseudocode] for(i in 1:6000) if (pred.U1.nb.c[i]=='unlabeled') U1.RN[j++,]<-U1[i,] [/pseudocode], where pred.U1.nb.c is a vector I got from a predict(), and want to create a data frame by selecting those rows of U1 that predict spewed out? (... trying hard to be verbose and not confusing simultaneously)Tathagata– Tathagata2010-11-24 16:58:58 +00:00Commented Nov 24, 2010 at 16:58
-
6In R, preallocate your storage! You know you want a 6000-row data frame ahead of the loop, so create one and fill it in row by row. Or even quicker; create a matrix of the correct dimension, fill that row by row, and then convert to a data frame, as matrices are much faster to work with. If you want more help (looks like you might not even need a loop, just some simple indexing and subsetting/insertion), can you start a new Q and provide a proper, small example of what you really want to do? If you do, I'll promise to look at it and give a go at an answer.Gavin Simpson– Gavin Simpson2010-11-24 17:18:59 +00:00Commented Nov 24, 2010 at 17:18
-
Thanks Gavin, here's the Q: tinyurl.com/26ugewvTathagata– Tathagata2010-11-24 17:54:28 +00:00Commented Nov 24, 2010 at 17:54
Add a comment
|
4 Answers
You can do this:
U1.RN <- U1[0,]
2 Comments
Tathagata
Perfect ..... > U1.RN<-U1[0,] > names(U1.RN) [1] "username" "review_count" "forum_posts" "age" "avg_interval" [6] "avg_sim" "class" > nrow(U1) [1] 6000 > nrow(U1.RN) [1] 0
Gavin Simpson
+1 neat!!!!!!!! (the extra ! were to get round the min character limit, oh, wait, ... ;-)
Along the lines of df[0,] you can also use a boolean mask which might make the code more readable:
df[FALSE,]
Comments
Using dplyr, there are a few good options:
slice(U1, 0)
filter(U1, FALSE)
filter(U1, NA)
The slice approach is probably clearest.
2 Comments
Gregor Thomas
filter(FALSE) also works and is maybe a little clear of intent. slice(0) might be even better.
joemienko
Agreed that
slice(0) seems to be a winner on readability/clarityFor completeness. This resets an existing data frame to zero rows.
U1.RN <- U1
attributes(U1.RN)$row.names <- c()
# <0 rows> (or 0-length row.names)
Note that rownames() <- NULL "deletes" the rownames and then fills them in with the default.
