spread changes factor order about tidyr

[mrdwab]

This actually seems to also affect the "id" variables. See, for example, here: http://stackoverflow.com/q/29381069/1270695

df = data.frame(name=c("B","B","A","A"),
                group=c("g1","g2","g1","g2"),
                V1=c(10,40,20,30),
                V2=c(6,3,1,7))

gather(df, Var, Val, V1:V2) %>% 
  unite(VarG, Var, group) %>% 
  spread(VarG, Val)

Note the factor levels for "name" in the input and output.

> str(.Last.value)
'data.frame':   2 obs. of  5 variables:
 $ name : Factor w/ 2 levels "A","B": 1 2
 $ V1_g1: num  20 10
 $ V1_g2: num  30 40
 $ V2_g1: num  1 6
 $ V2_g2: num  7 3
> str(df)
'data.frame':   4 obs. of  4 variables:
 $ name : Factor w/ 2 levels "A","B": 2 2 1 1
 $ group: Factor w/ 2 levels "g1","g2": 1 2 1 2
 $ V1   : num  10 40 20 30
 $ V2   : num  6 3 1 7

[dataRangler]

I asked the SO question:-) I am a beginner but when I looked at the spread.R source code, line 79 seems to be the start of the sorting. Is it necessary? I am new to R and github and do not know how to test this yet.

# Add in missing values, if necessary
if (length(overall) < n) {
  overall <- match(seq_len(n), overall, nomatch = NA)
} else {
  overall <- order(overall)
}

[dataRangler]

I've just found dplyr::summarise does sorting as well. Is this a design philosophy?

df %>% 
+ group_by(name) %>% 
+ summarise(n()
+ )
Source: local data frame [2 x 2]

  name n()
1    A   2
2    B   2

[hadley]

Instead of unite() do:

d1 %>% mutate(univar = f1:f2, f1 = NULL, f2 = NULL) %>% spread(univar,val)
# OR
d1 %>% mutate(univar = interaction(f1, f2), f1 = NULL, f2 = NULL) %>% spread(univar,val)