https://issues.apache.org/jira/browse/ARROW-13853

@github-actions crossbow submit test-r-linux-as-cran

Revision: f6c29d0379efb31fe4e41c959356bc484c4ca43c

Submitted crossbow builds: ursacomputing/crossbow @ actions-859

Task	Status
test-r-linux-as-cran

Thanks all for the great reviews!

Interesting...the failed tests are because stringr behavior for str_to_title() is different from Python's title() when string begins with an integer. The C++ implementation follows Python's behavior.

# R stringr
> str_to_title("1Foo1")  # "1foo1"
# Python
>>> "1Foo1".title()  # "1Foo1"

I conclude that this also applies for "is_title()" function because it ignores beginning integers...well...stringr does not has an is_title() function.

What to do in such a case? cc @jonkeane @nealrichardson

Hmm, practically speaking I suspect this won't matter to all that many people. We have a few options:

1. We could introduce an argument to the kernel that allows one to change the behavior to one or the other
2. We could accept Python's behavior and note that this will produce slightly different results than {stringr}
3. We could change the behavior to {stringr}'s (and for that matter tools::toTitleCase()) and note that the Python implementation will be slightly different.

I suspect that 2 is probably the easiest right now.

And if you're looking for other ways this could be even more hairy: https://stat.ethz.ch/R-manual/R-devel/library/tools/html/toTitleCase.html

this implementation knows about conventional mixed-case words such as ‘LaTeX’ and ‘OpenBUGS’ and a few technical terms which are not usually capitalized such as ‘jar’ and ‘xls’. However, unknown technical terms will be capitalized unless they are single words enclosed in single quotes: names of packages and libraries should be quoted in titles.

Yeah, when I was implementing the title() kernels in C++, I encountered all sorts of corner cases and hurdles, but we decided to keep it simple and match Python's approach.

For this PR, the R binding calls the C++ kernels directly, so it will produce different results for extreme cases when compared to stringr. If users find issues or request more complex functionality, then this can be revisited in a future PR.