group_by with a data or timestamp grouping variable causes R to crash

[giacecco]

I am using dplyr 0.1.1 on R 3.0.2 on MacOS Mavericks.

What you see below crashes on the 'group_by' command, either that I converted the SCHEDULED_OFF column to a full timestamp or to date only, as in the example.

df <- read.csv(pipe("unzip -p myfile.zip"), header = TRUE, colClasses = "character")
df <- df[, c("SPORTS_ID", "SCHEDULED_OFF", "VOLUME_MATCHED")]
df <- df[!is.na(df$SCHEDULED_OFF),]
df$SCHEDULED_OFF <- strptime(df$SCHEDULED_OFF, "%d-%m-%Y")
class(df$SCHEDULED_OFF)
[1] "POSIXlt" "POSIXt" 
df2 <- group_by(df, SCHEDULED_OFF)

The input dataset is > 600k rows. I am sorry I am not authorised to share it

Giacecco

[romainfrancois]

dplyr does not handle data frames with POSIXlt columns. we need to make this more explicit by giving warnings instead of crashing like that.

@giacecco for future reference, please do the exercise of creating reproducible example. I guess you can easily reproduce the crash without your data set, but with a few lines of generated data.

[giacecco]

Got it and sorry for not having setup a basic example showing the issue.

G.

[romainfrancois]

I've reopened the issue because it is not resolved yet. :)

[romainfrancois]

Done now.