Vectorize add_suffixes()
#6642
Vectorize add_suffixes()
#6642
Conversation
DavisVaughan
force-pushed
the
feature/faster-suffix-adding
branch
from
17d6b92 to
a79a798
Compare
| x <- c(y, x) | ||
|
|
||
| # Never marks the "first" duplicate (i.e. never anything in `y`) | ||
| dup <- duplicated(x) |
There was a problem hiding this comment.
Since we're doing performance optimisations, I think this might be a little faster:
dup <- anyDuplicated(x)
while (length(dup) > 0) {
x[dup] <- paste0(x[dup], suffix)
dup <- anyDuplicated(x)
}There was a problem hiding this comment.
Would it? If there are multiple columns that are duplicated between x and y then my approach tackles them all in one pass. We only ever do a 2nd pass at all in the super rare case where adding a suffix created another duplication.
anyDuplicated() returns exactly 1 value, the location of the first duplicated value or 0 if there are no dups. So we'd have to do (at least) as many passes as there are duplicates in x and y
There was a problem hiding this comment.
Related to r-lib/vctrs#1239 since it would be nice to be able to use vec_detect_duplicate(x, ignore = "first") here instead of duplicated()
DavisVaughan
force-pushed
the
feature/faster-suffix-adding
branch
from
3b71f6c to
2816ada
Compare
Closes #6636
Alternate approach that truly vectorizes the calculation. I'm about 95% sure this is correct. It passes all tests and we do have a decent number of tests that check corner cases of
suffixso I'm fairly confident about it.