dplyr 1.0.0 update throwing error in plot_compare_stage

[hdugan]

With the new dplyr update v.1.0.0, get the following error:

plot_compare_stage(nc_file, field_file)
Error: `vec_cast.POSIXct.POSIXct()` is implemented at C level.
This R function is purely indicative and should never be called.

This is traced back to vec_rbind in the vctrs package.
I actually think this might be a dplyr bug... so it might sort itself out.

Until then, keep the old versions of the following packages:

install_version("dplyr", version = "0.8.5", repos = "http://cran.us.r-project.org")
install_version("vctrs", version = "0.2.4", repos = "http://cran.us.r-project.org")
install_version("tidyr", version = "1.0.0", repos = "http://cran.us.r-project.org")

We might have some other dplyr errors in other functions as well

[jordansread]

yucky error. is it in the master version or the ggplot update branch? or both?

[hdugan]

ggplot update branch only. (fyi, the master branch still has dependencies on GLMr)

There have been lots of dplyr issues opened in the last week, so I'm just going to wait on this for a little bit.

[jordansread]

I ran into this on the master branch today too

[jordansread]

I get it when attempting to left_join with DateTime as the "by" argument. Since DateTime here is a POSIXct. I got around it by using date = as.Date(lubridate::ceiling_date(DateTime, 'days')) and joining on date instead, since I was working with daily output anyhow.

[hdugan]

🤕 I'll look into it this week.

[jordansread]

Nevermind me saying it was in the master - the error was actually something I hit after get_var, so I was doing the join outside of glmtools

[hdugan]

Good!
Trying to understand why this is happening is bizarre, and I'm surprised there isn't more stackoverflow discussions about it. POSIXct is throwing the errors. Works if I convert to POSIXlt (which I haven't used in a long time).

[hdugan]

Figured it out (or at least part of it). In the get_time function, hours_since <- ncvar_get(glm_nc, "time") returns an array as opposed to a vector. This adds 'dimensions' to the object which downstream cause a lot of problems. By changing to a numeric vector everything seems to work as previously.

get_time  <-  function(glm_nc){
	hours_since  <- ncvar_get(glm_nc, "time")
	time_info <- get_time_info(glm_nc)

	time <- time_info$startDate + time_info$time_unit * hours_since * 60*60*24

	return(time)
	
}

[padilla410]

This bug is related to the issues in #275. dplyr 1.0.0 outputs data summaries into tibbles. A reproducible example:

library(dplyr)

data(iris)

# a `data.frame` object
class(iris)

out <- iris %>% 
  group_by(Species) %>% 
  summarise(count = n())

# a `tibble` object
out

The output:

> library(dplyr)
> 
> data(iris)
> 
> # a `data.frame` object
> class(iris)
[1] "data.frame"
> 
> out <- iris %>% 
+   group_by(Species) %>% 
+   summarise(count = n())
> 
> # a `tibble` object
> out
# A tibble: 3 x 2
  Species    count
  <fct>      <int>
1 setosa        50
2 versicolor    50
3 virginica     50