[
https://issues.apache.org/jira/browse/ARROW-12017?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Keane updated ARROW-12017:
-----------------------------------
Description:
Right now we have discussions of installation in two places: the readme and a
vignette that is (almost exclusively) focused on Linux. Both of these cover
installation as a user and installation as a developer.
We want to change this to have a better split so that we can focus the topics
to what people actually need:
* The readme should have user-facing how-to-install from CRAN + how to get
nightlies and a link to a developer docs vignette if one wants to build from
source / develop
* The developer docs vignette should include details about compilation,
debugging, etc.
The developer docs should:
[ ] Note how to install the R dependencies for the package. The audience here
is not necessarily R developers, so we should be super explicit and not assume
things like "I have set a CRAN mirror, of course"
[ ] Recommend building + installing libarrow to the same / a similar non-system
directory as the python docs do. (Note: this might take some testing + work to
ensure that when R builds it does not attempt to link against existing system
installs — I've had this happen in the past and it can be incredibly confusing
when it does)
[ ] Include debugging instructions for "Can't load this symbol" "Can't find
this function" type errors (either the library isn't linking at runtime at all,
or it's linking to a stale libarrow and libarrow needs to be rebuilt.
[ ] discuss nightly builds and how to get them / when one won't be able to get
them depending on the method of install
was:
Right now we have discussions of installation in two places: the readme and a
vignette that is (almost exclusively) focused on Linux. Both of these cover
installation as a user and installation as a developer.
We want to change this to have a better split so that we can focus the topics
to what people actually need:
* The readme should have user-facing how-to-install from CRAN + how to get
nightlies and a link to a developer docs vignette if one wants to build from
source / develop
* The developer docs vignette should include details about compilation,
debugging, etc.
The developer docs should:
[ ] Note how to install the R dependencies for the package. The audience here
is not necessarily R developers, so we should be super explicit and not assume
things like "I have set a CRAN mirror, of course"
[ ] Recommend building + installing libarrow to the same / a similar non-system
directory as the python docs do. (Note: this might take some testing + work to
ensure that when R builds it does not attempt to link against existing system
installs — I've had this happen in the past and it can be incredibly confusing
when it does)
[ ] Include debugging instructions for "Can't load this symbol" "Can't find
this function" type errors (either the library isn't linking at runtime at all,
or it's linking to a stale libarrow and libarrow needs to be rebuilt.
> [R] [Documentation] Make proper developing arrow docs
> -----------------------------------------------------
>
> Key: ARROW-12017
> URL: https://issues.apache.org/jira/browse/ARROW-12017
> Project: Apache Arrow
> Issue Type: Improvement
> Components: Documentation, R
> Reporter: Jonathan Keane
> Assignee: Jonathan Keane
> Priority: Major
>
> Right now we have discussions of installation in two places: the readme and a
> vignette that is (almost exclusively) focused on Linux. Both of these cover
> installation as a user and installation as a developer.
> We want to change this to have a better split so that we can focus the topics
> to what people actually need:
> * The readme should have user-facing how-to-install from CRAN + how to get
> nightlies and a link to a developer docs vignette if one wants to build from
> source / develop
> * The developer docs vignette should include details about compilation,
> debugging, etc.
> The developer docs should:
> [ ] Note how to install the R dependencies for the package. The audience here
> is not necessarily R developers, so we should be super explicit and not
> assume things like "I have set a CRAN mirror, of course"
> [ ] Recommend building + installing libarrow to the same / a similar
> non-system directory as the python docs do. (Note: this might take some
> testing + work to ensure that when R builds it does not attempt to link
> against existing system installs — I've had this happen in the past and it
> can be incredibly confusing when it does)
> [ ] Include debugging instructions for "Can't load this symbol" "Can't find
> this function" type errors (either the library isn't linking at runtime at
> all, or it's linking to a stale libarrow and libarrow needs to be rebuilt.
> [ ] discuss nightly builds and how to get them / when one won't be able to
> get them depending on the method of install
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
