Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
0.16.0
-
None
-
RHEL 7.6
Custom R build in a non-default location
Description
Hey Team,
I've been using Arrow successfully in python through conda and have been able to write and read parquet files successfully. Now I'm trying to have R users consume some of the parquet files I've produced. They run a shared R build in a custom location, so can't use conda with their setup. We tried installing the C++ libraries system-wide and then the R library in a user's directory, but kept getting Cannot call io_MemoryMappedFileOpen() errors on read_parquet(). I'm not sure if we've missed a step, or where to continue debugging. Does the R package have any known issues on RHEL 7? Below are some details:
Environment:
- RHEL 7.6
- Custom local R environment in a non-default location
Steps taken:
- Installed the C++ libraries first (now live in /usr/lib64) as described (v.0.16.0) in https://arrow.apache.org/install/
- Ran install.packages('arrow') in an interactive R session
- It couldn't find the C++ libraries and said No C++ binaries found for rhel-7
- Couldn't find https://dl.bintray.com/ursalabs/arrow-r/libarrow/bin/rhel-7/arrow-0.16.0.zip when it tried to pull the binaries
- Source download didn't work due to internal github firewall rules
- Installed without errors, but threw Cannot call io_MemoryMappedFileOpen() error on read_parquet()
- Removed the rlib/arrow directory and tried a different route
- Set LIBARROW_BINARY_DISTRO='centos-7'
- Set PKG_CONFIG=/usr/lib64
- Ran install.packages('arrow') in an interactive R session
- Binaries and package seemed to install correctly without complaints
- Still threw Cannot call {{io__MemoryMappedFile_Open{{() on read_parquet}}}}