Open barracuda156 opened 1 year ago
cc @rok. I'm not sure how locale is handled in the C++ library but looks like something isn't happy on PPC?
P. S. Not sure why R-arrow
has these inactive:
parquet FALSE
json FALSE
arrow
itself has been built with support for these. Complete config is here:
https://github.com/barracuda156/macports-ports/blob/f34cdb879bd40e6d4b5247a19ab4cb630738a443/devel/apache-arrow/Portfile#L120-L153
Hey. It seems like there's two problems here:
export LC_ALL="C"
to see what happens?CCTZ: Invalid timezone of the input vector: "Asia/Kathmandu"
) which you might be able to resolve with instructions here: https://howardhinnant.github.io/date/tz.html#InstallationHey. It seems like there's two problems here:
Locale error is probably thrown here https://github.com/apache/arrow/blob/49631057e9cdbf991e11e0be4b9aa0dadf616850/cpp/src/arrow/compute/kernels/scalar_temporal_unary.cc#L1157
Which seems like a ppc issue. Perhaps try
export LC_ALL="C"
to see what happens?- Timezone not found (
CCTZ: Invalid timezone of the input vector: "Asia/Kathmandu"
) which you might be able to resolve with instructions here: https://howardhinnant.github.io/date/tz.html#Installation
@rok Thank you, I will check these.
Could we also fix recognizing older macOS versions? To get rid of this:
Error: invalid version specification '10.0.0d2'
@barracuda156 I'm not sure where invalid version specification
is thrown. It looks like it's a generic R error (https://github.com/rstudio/reticulate/issues/204#issuecomment-377403623), @nealrichardson any ideas?
@rok Well, here it is actually the OS version, and R
itself recognizes it:
svacchanda$ r
R version 4.2.3 (2023-03-15) -- "Shortstop Beagle"
Copyright (C) 2023 The R Foundation for Statistical Computing
Platform: powerpc-apple-darwin10.0.0d2 (32-bit)
> R.Version()
$platform
[1] "powerpc-apple-darwin10.0.0d2"
$arch
[1] "powerpc"
$os
[1] "darwin10.0.0d2"
$system
[1] "powerpc, darwin10.0.0d2"
$status
[1] ""
$major
[1] "4"
$minor
[1] "2.3"
$year
[1] "2023"
$month
[1] "03"
$day
[1] "15"
$`svn rev`
[1] "83980"
$language
[1] "R"
$version.string
[1] "R version 4.2.3 (2023-03-15)"
$nickname
[1] "Shortstop Beagle"
The version thing is from a test helper we have in the R package, it's an easy remedy, not connected to the locale issues in C++
The invalid version specification
comes from here: https://github.com/apache/arrow/blob/main/r/R/arrow-package.R#L210 because it seems Sys.info()["release"]
on your setup is not parseable as a numeric version. This should work around:
diff --git a/r/R/arrow-package.R b/r/R/arrow-package.R
index a3c860a51c..9fc89b46ea 100644
--- a/r/R/arrow-package.R
+++ b/r/R/arrow-package.R
@@ -207,7 +207,8 @@ on_linux_dev <- function() {
on_macos_10_13_or_lower <- function() {
identical(unname(Sys.info()["sysname"]), "Darwin") &&
- package_version(unname(Sys.info()["release"])) < "18.0.0"
+ # wrap in isTRUE because package_version can return NA
+ isTRUE(package_version(unname(Sys.info()["release"]), strict = FALSE) < "18.0.0")
}
option_use_threads <- function() {
Describe the bug, including details regarding any error messages, version, and platform.
So there are these errors (locale one happens with several tests):
P. S. Version is correct, of course, test it wrong here.
Complete output:
arrow_info()
:Component(s)
R