Cannot View data frame with integer64 columns

[renkun-ken]

System details

RStudio Edition : both Desktop and Server
RStudio Version : 1.2.812
OS Version      : Ubuntu 16.04, macOS 10.13.6
R Version       : 3.4.4, 3.5.1

Steps to reproduce the problem

data <- data.frame(x = bit64::as.integer64(c(1,2,3)))
View(data)

Error in hist.default(hist_vals, plot = FALSE) : 
  some 'x' not counted; maybe 'breaks' do not span range of 'x'
In addition: Warning message:
In pretty.default(range(x), n = breaks, min.n = 1) :
  Internal(pretty()): very small range.. corrected

[ronblum]

Thanks @renkun-ken for letting us know about this issue.

Reproduced on Windows 10 and MacOS 10.12.6 (Sierra). This is a regression (worked in RStudio 1.1.453).

[ronblum]

This is working (as of 1.2.830) when initially rendered. However:

1. Open a project
2. Repeat the steps above
   • Works well
3. Close the project
4. Reopen the project
   • The error returns, along with an empty data viewer

[jmcphers]

The error's actually changed somewhat; it now references the new code:

Error in hist.default(as.numeric(hist_vals), plot = FALSE) :

The problem is that as.numeric doesn't work on integer64 values until the bit64 package is loaded. That's why it doesn't work on startup, but it does during the session in which you've called bit64::as.integer64 (as doing so loads the package).

We try pretty hard not to have RStudio load spurious packages at startup, and I'm also not sure all of the codepaths that support the histogram are going to handle 64 bit integers gracefully, so I think we should prevent this kind of thing by refusing to histogram-ize any numeric columns which have a special class applied.

[ronblum]

The error is gone. However, we're still displaying values in both cases. First we get (1, 2, 3):

but when the project is reopened, we get this:

[jmcphers]

The bit64 package must be loaded in order for us to get sensible values. We do not load packages on startup, so the raw values are displayed. This behavior is weird, but it isn't a regression, and should be filed separately.

[ronblum]

Done! Verified that error is gone.

[nielsenrechia]

Hi, I'm getting the same error Error in hist(hist_vals, plot = FALSE) : could not find function "hist", but insted of integer64 it is occuring for for double columns.

data <- data.frame(x = as.double(c(1,2,3)))
View(data)

Someone could help me?

thanks