[R] Altrep and translations (was " Description of error is untranslated when ....")

[Ricardo Villalba]

Hello
On the other hand, i find that R_compact_intrange checks whether (r = n2 -
n1 + 1) >= R_XLEN_T_MAX,  while seq_colon checks for (r = fabs(n2 - n1)) >=
R_XLEN_T_MAX. And only then n is defined as r + 1 + epsilon. r is not used
again
That is why seq_colon does not reject that size at first and then altrep
does.
Should not line 101  be  " if (r + 1 + FLT_EPSILON >= R_XLEN_T_MAX) "
instead of r >= R_XLEN_T_MAX?

El mar, 14 may 2024 a las 4:45, Martin Maechler (<maechler using stat.math.ethz.ch>)
escribió:

> >>>>> Ivan Krylov via R-help
> >>>>>     on Tue, 14 May 2024 08:08:58 +0300 writes:
>
>     > Dear Ricardo Villalba, Thank you for spotting this corner
>     > case!
>
>     > В Mon, 13 May 2024 11:37:57 -0300 Ricardo Villalba
>     > <rikivillalba using gmail.com> пишет:
>
>     >> I track the messages to be coded here:
>     >>
> https://github.com/r-devel/r-svn/blob/abe625945c4402cd2bb97b5a64e7469db3e904f0/src/main/altclasses.c#L580
>     >> and here
>     >>
> https://github.com/r-devel/r-svn/blob/abe625945c4402cd2bb97b5a64e7469db3e904f0/src/main/seq.c#L102
>
>     > You may have noticed that one of these places lacks the
>     > underscore that R uses as the translation keyword: it's
>     > _(string) that looks up the translation of the string.
>
>     > Would you like to suggest a patch on R-devel using r-project.org
>     > or https://bugs.r-project.org ?
>
>     > --
>     > Best regards, Ivan
>
> Thank you, but, well... for such small changes -- that would be
> quite a bit of (human, i.e. R-core) overhead, but of course is
> generally a good possibility  (but here, wait! .. and read on : )
>
> Actually -- this now gets more  "R-devel" - appropriate --
> I notice that there are quite a few  error("..." .. )
> and also  warning("..." ..) calls in
> src/main/altclasses.c *and* in
> src/main/altrep.c        and *none* of them asks for translation;
>
> While this could be for historical reasons (code originally much
> by Gabe Becker from outside R-core *and* at first just in a
> separate branch before being merged in to the main (r-devel)
> branch.
> OTOH: There may be good reasons for translations lookup being
> brittle in case of altrep error messages .. and hence left off
> purposely?
>
> Martin
>
> --
> Martin Maechler
> ETH Zurich  and  R Core Team
>
>

-- 
®DV!

	[[alternative HTML version deleted]]