filter, slice, arrange dropping attributes

[DavisBrian]

I'm not sure if this is by design or a bug.

I would expect that since filter and slice are subsets of the original data frame that all of the original data frames attributes would go along with it. Similarly since arrange is a reordering of the data I would expect it to keep attributes, much like group_by and select do.

library(dplyr)

dat <- as_data_frame(iris)
attr(dat, "meta") <- c("This is important")

# drops meta attributes
dat %>% filter(Species == "setosa") %>% attributes
dat %>% slice(1:50) %>% attributes
dat %>% arrange(Sepal.Width) %>% attributes


# keeps meta attribute
dat %>% group_by(Species) %>% attributes
dat %>% select(Petal.Length, Petal.Width, Species)  %>% attributes

[hadley]

@romainfrancois can we also preserve all data frame attributes?

@DavisBrian do you expect summarise() to keep attributes too?

[DavisBrian]

@hadley I would not expect summarise() to keep attributes. In my mind summarise() represents different data. The other functions are more or less just subsets and orderings of the same data. Or at least that's my use case.

[eamcvey]

In my experience, this issue often creates problems using lubridate and dplyr together, which is a thing I'm always wanting to do!

[hadley]

Should be fixed the dev version

[hadley]

Oops, wrong issue.

[romainfrancois]

Ah. Might have closed it too soon. So @hadley do we want to keep attributes for summarise too ?

[hadley]

Hmmmmmmmm, for summarise I think we should be able to rely on the summary function doing the right thing. (So that might require preserving attributes in the hybrid handlers)

[romainfrancois]

Sure, but this only concerns attributes of the columns right ?
This is more about some meta attributes of the data frame itself, should they be propagated to the resulting data frame.
Or perhaps I did not understand you.

[hadley]

Oh, sorry yes :/ So summarise should preserve these attributes.