`tbl_df` breaks `xtable()` -- bug?

[aammd]

I swear I used to be able to print tbl_df objects, but suddenly after upgrading to latest dplyr that went away:

library(dplyr)

iris %>% 
  tbl_df %>%
  xtable %>%
  print
## Error in .subset2(x, i, exact = exact) : subscript out of bounds
## doesn't work

iris %>% 
  tbl_df %>%
  as.data.frame %>%
  xtable %>%
  print
## works fine

[romainfrancois]

I've' assigned you @hadley just so that you check I did not break anything else with the commit. Seems straightforward enough.

[happyshows]

Hi, I have encountered a similar but really annoying issue, though I'm not sure whether it's due to xtable in dplyr or ggvis. (In addition, I found current release of dplyr breaks renderTable in shiny too)

> ds2 <- read.csv('test.csv')
> ds2 %>%  ggvis(~Family,~Measure)
Guessing layer_bars()   <- SUCCESS
> ds2 %>% tbl_df %>% ggvis(~Family,~Measure)
Guessing layer_bars()
Error in .subset2(x, i, exact = exact) : subscript out of bounds  <- FAIL

example dataset ('test.csv'):
   Family Measure
1       A    99.9
2       B    97.6
3       C     0.0
4       D    94.1
5       E    96.6
6       F    99.9
7       G    95.4
8       H    81.2
9       I    99.8
10      J    62.8
11      K    99.5
12      L    23.0
13      M     0.0
14      N    51.4
15      O     0.0
16      P    99.9
17      Q    99.9
18      R    98.3
19      S    99.9
20      T    99.9
21      U    38.1
22      V    88.6
23      W    98.9
24      X    72.7

[romainfrancois]

@happyshows can you create a reproducible example instead of using test.csv which we don't have.

[happyshows]

@romainfrancois
Here you go

ds <- data.frame(x=letters[1:10],y=1:10) # --works
ds %>% ggvis(~x,~y) 

# --fails
# --Error in .subset2(x, i, exact = exact) : subscript out of bounds
ds %>% tbl_df %>% ggvis(~x,~y)

However,

cars %>%  ggvis(~speed,~dist)
cars %>% tbl_df %>% ggvis(~speed,~dist)

both works fine, as layer_points() seems to work
if I choose to plot as bar, similar error will happen

cars %>% mutate(speed = as.factor(speed)) %>% ggvis(~speed,~dist)
cars %>% tbl_df %>% mutate(speed = as.factor(speed)) %>% ggvis(~speed,~dist)

[romainfrancois]

It is indeed the same problem as this works fine with the dev_0_3_1 branch. This will be merged into the master branch once dplyr 0.3.0.1 is on CRAN.

BTW @happyshows I've slightly edited your comment so that the code format nicely. embed code chunks with ```.

[happyshows]

@romainfrancois
Thanks, I did try to find a way to turn the codes into a block, but I'm not very familiar with markdown or github.
Do you have a estimated release time for the dev_0_3_1 on cran? I am currently working on a proof of concept model which utilizes shiny,dplyr,ggvis and would definitely want to fix broken ggvis charts before my meeting with vp soon.

[romainfrancois]

You can get it right now with:

install_github( "hadley/dplyr", ref = "dev_0_3_1" )

Now that dplyr 0.3.0.1 is on CRAN, it should not take too long for these to be merged into master. When dplyr 0.3.1 gets released depends on many things incl seriousness of bug reports we get about 0.3.0.1

[happyshows]

@romainfrancois
Could you double check?

Downloading github repo hadley/dplyr@dev_0_3_1
Error in download(dest, src, auth) : client error: (404) Not Found

[hadley]

@happyshows we just merged the branch, so install_github("hadley/dplyr") should be adequate. Also check that you have devtools 1.6 or higher.