Failing tests on R 3.3.0

[krlmlr]

See #1507 (comment) and references.

I think the problem boils down to:

> match("l\u00f8penummer", iconv(  "l\u00f8penummer", to  = "latin1" ))
[1] NA
> match(iconv(  "l\u00f8penummer", to  = "latin1" ), "l\u00f8penummer")
[1] NA

I'm taking a closer look.

[krlmlr]

Looks like this problem was introduced in R 3.3.0 -- I'm seeing the same problem here. NB: I think the code used in the tests need to be changed anyway.

> match("l\u00f8penummer", iconv(  "l\u00f8penummer", to  = "latin1" ))
[1] NA
> match(iconv(  "l\u00f8penummer", to  = "latin1" ), "l\u00f8penummer")
[1] NA

I haven't found anything in the NEWS about it, is this a true regression?

[krlmlr]

Unfortunately, yes:

> match(iconv(  c("A", "l\u00f8penummer"), from = "UTF8", to  = "latin1" ), "l\u00f8penummer")
[1] NA  1
> match(iconv(  c("l\u00f8penummer"), from = "UTF8", to  = "latin1" ), "l\u00f8penummer")
[1] NA

This might be related to

match(x, table) is faster (sometimes by an order of magnitude) when x is of length one and incomparables is unchanged (PR#16491).

Posted to R-devel.

[yutannihilation]

FYI: the bug is reported below and the fix is already committed.

16885 – bug in match (I think it's related to PR#16491)
https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=16885