Small bugfix to prevent available.versions crashing early

[hughjonesd]

when a version is not recognized.

• Alternative approach would be to have current_version return a 0-row data frame rather than a 1-row dataframe with NAs.
• The underlying issue is that all(NA) returns NA on a one-row dataset.

[goldingn]

Hi David,

Could you please add an issue showing where this causes you an error, so I replicate it?

I can see why that line of code would error if a single row dataframe is returned with NA in the available column, just not what would lead to the NA.

Nick

[hughjonesd]

The one row dataframe is returned from current_version if version is not recognized. Maybe that's fine and you want to throw an error, but it would be nice to have a more comprehensible one; and your warning suggests you want to recover from a single bad version.

> versions::available.versions(c("bad", "testthat"))
Error in if (!all(df$available)) { : 
  missing value where TRUE/FALSE needed
In addition: Warning message:
In current.version(pkgs) :
  The current version and publication date of bad could not
                     be detected

[goldingn]

Gotcha, yep that needs a better error!

[hughjonesd]

OK, this tries to throw an error earlier. The alternative would be to do it in current_version.