Skip to content

[R-pkg-devel] Avoiding differences between Examples output on Windows due to tibble

3 messages · Uwe Ligges, Gavin Simpson

#
Dear List,

The latest version of tibble 2.0.1 is printing output (on Linux at
least) that contains a UTF-8 ellipsis ? instead of the previous three
periods ...

This is causing some differences between the reference materials for
my pkg examples that I generate locally on Linux and the output
generated on the Winbuilder system under Windows. In the latter, the
ellipsis is not being printed and three periods are being used
instead. As such I am seeing the following differences in Windows only
un R CMD check:

** checking differences from 'gratia-Ex_i386.Rout' to
'gratia-Ex.Rout.save' ... OK

239c239
< # ... with 790 more rows
---
** running examples for arch 'x64' ... [24s] OK
** checking differences from 'gratia-Ex_x64.Rout' to
'gratia-Ex.Rout.save' ... OK

239c239
< # ... with 790 more rows
---
I'm assuming this is going to be an issue for CRAN. If so, what can i
do about this? Is there some flag or trick I can use to generate the
same output on Windows as I produce on Linux?

The full sources for package in question are on github:
https://github.com/gavinsimpson/gratia should anyone want to check
something there.

Thanks in advance

Gavin

--
Gavin Simpson, PhD
#
On 17.01.2019 00:41, Gavin Simpson wrote:
Please ask the tibble maintainer what this is about.
In my opinion this should be changed back.

Best,
Uwe Ligges
#
Thank Uwe, I also had some discussions on Twitter with various people
about this.

Thanks to Brodie Gaslam (@BrodieG) via Twitter, a solution seems to be
to set the global option `cli.unicode` to `FALSE` for checking, as
this is coming from the cli pkg.

Hence for my purposes in examples I am using

\dontshow{op <- options(cli.unicode = FALSE)}
....
\dontshow{options(op)}

around sections of examples that print tibble-based output.

This is a bit of a pain, but as it is already common to reduce
precision on example output, setting this option is just one more
thing to configure if you want to continue using reference Example
output with R CMD check and tibbles.

All the best

Gavin

On Thu, 17 Jan 2019 at 05:15, Uwe Ligges
<ligges at statistik.tu-dortmund.de> wrote: