[
https://issues.apache.org/jira/browse/ARROW-5614?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16866840#comment-16866840
]
Thomas Buhrmann commented on ARROW-5614:
----------------------------------------
Yes, using conda (on the conda-forge channel) works like a charm! I was
initially reluctant to manage R packages via conda, because of some negative
comments related to inadvertently mixing conda-managed package libraries and
other (default) R library locations. But in practice, for me at least, that
problem hasn't cropped up, and in any case should be easily solved with
appropriate PATH or .libPaths configuration. So yes, I think it's simply a
matter of documentation: making clear how to correctly link against arrow-cpp
(conda install being the easiest), and updating or removing reference to
arrow::install_arrow().
And thanks for the nudge to re-consider conda installs! Makes the creation of
docker images with arrow installed much easier...
> [R] Error: 'install_arrow' is not an exported object from 'namespace:arrow'
> ---------------------------------------------------------------------------
>
> Key: ARROW-5614
> URL: https://issues.apache.org/jira/browse/ARROW-5614
> Project: Apache Arrow
> Issue Type: Bug
> Components: R
> Reporter: Thomas Buhrmann
> Assignee: Neal Richardson
> Priority: Major
>
> I'm trying to get the R package installed in a Debian docker image that
> already contains R and RStudio (via rocker/rstudio from dockerhub), as well
> as arrow-cpp, parquet-cpp and pyarrow installed via conda. I.e. I should have
> all required arrow dependencies in my conda environment's /lib and /include
> folders.
> I then tried to install the R package in two ways (as stated in the README,
> having devtools, and after managing to get git2r installed)
> 1/ via remotes
> {code:java}
> remotes::install_github("apache/arrow/r",
> ref="76e1bc5dfb9d08e31eddd5cbcc0b1bab934da2c7"){code}
> 2/ from source
> {code:java}
> git clone https://github.com/apache/arrow.git
> cd arrow/r
> R -e 'remotes::install_deps()'
> R CMD INSTALL
> --configure-vars='INCLUDE_DIR=/root/miniconda/envs/my_env/include
> LIB_DIR=/root/miniconda/envs/my_env/lib' .{code}
> In both cases the install seems to work fine:
> {code:java}
> ** building package indices
> ** testing if installed package can be loaded from temporary location
> ** checking absolute paths in shared objects and dynamic libraries
> ** testing if installed package can be loaded from final location
> ** testing if installed package keeps a record of temporary installation path
> * DONE (arrow)
> {code}
> But when I then do the following as prompted:
> {code:java}
> library(arrow)
> arrow::install_arrow()
> {code}
> The result is
> {code:java}
> Error: 'install_arrow' is not an exported object from 'namespace:arrow'
> {code}
> And running the example without calling that non-existing function I get the
> error
> {code:java}
> Error in Table__from_dots(dots, schema) :
> Cannot call Table__from_dots(). Please use arrow::install_arrow() to
> install required runtime libraries.
> {code}
> So I don't know if I'm doing something wrong or if the documentation isn't up
> to date? Specifically, what is the arrow::install_arrow() function supposed
> to install, given that I already have the arrow and parquet libs and headers
> installed, and supposedly they've been used (linked to) when I installed the
> R package?
> In general, is there any way to get this package installed in the above
> context (arrow-cpp etc. installed via conda)?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
