[ARROW-240] Update installation instructions for pyarrow#105
[ARROW-240] Update installation instructions for pyarrow#105MechCoder wants to merge 3 commits intoapache:masterfrom
Conversation
python/README.md
Outdated
| These are instructions on how to install PyArrow from scratch on Linux (assuming arrow is not yet installed) | ||
|
|
||
| 1. **g++ and gcc** | ||
|
|
There was a problem hiding this comment.
These instruction sound very general but probably only apply to old Ubuntu versions, 12.04 is nearly end-of-life and 14.04 already has GCC 4.8 which is recent enough for Arrow to built.
There was a problem hiding this comment.
I agree that the instructions are general, but then I recall that the build failed for gcc4.8 which is why I have explicitly mentioned 4.9
|
This seems to be mainly our current Travis build baked into the README. While it may be a convenient approach for Travis, I do not recommend it for the typical user who wants to try out Arrow. We should rather base the simple instructions for users on more recent distribution versions where the installation should be much simpler. If there is a need for official build instructions on older distributions, we could add them in an extra section but not as the primary ones. |
python/README.md
Outdated
| sudo add-apt-repository ppa:george-edison55/cmake-3.x | ||
| sudo apt-get update | ||
| sudo apt-get install cmake | ||
| ``` |
There was a problem hiding this comment.
IMHO it is sufficient to state the minimum cmake and gcc version requirements without providing installation instructions, as this will vary from system to system.
|
I agree with @xhochy -- I think providing some minimal instructions about how to get started should be helpful, but since things are likely to change rapidly as the project grows and becomes more suitable for average users I do not think we should detail the current development environment in detail in the README. In general, we will not be expecting users to build Arrow or PyArrow from source -- they'll get the packages from binary wheels on PyPI or from conda packages. |
MechCoder
force-pushed
the
python_installation
branch
from
a20d751 to
c2533e9
Compare
|
I've updated the instructions and made it minimalistic. Let me know if you need anything else. |
python/README.md
Outdated
| #### PyArrow Dependencies: | ||
| These are the various projects that PyArrow depends on. | ||
|
|
||
| 1. **g++ and gcc Version >= 4.9** |
There was a problem hiding this comment.
Minimal GCC is 4.8, working locally and in the manylinux1 docker image. GCC 4.7 is definitely not working as we're using using features implemented first in 4.8.
|
done |
| The preferred way to install parquet-cpp is to use conda. | ||
| You need to set the ``PARQUET_HOME`` environment variable to where parquet-cpp is installed. | ||
| ```bash | ||
| conda install -y --channel apache/channel/dev parquet-cpp |
There was a problem hiding this comment.
This is OK for now. I'd like to get this into conda-forge as soon as possible to build from a shared compiler toolchain vs. the current gcc 4.9 dev builds
|
+1 |
|
great, thanks! |
Author: Uwe L. Korn <uwelk@xhochy.com> Closes apache#105 from xhochy/parquet-616 and squashes the following commits: eb614c5 [Uwe L. Korn] PARQUET-616: WriteBatch should accept const arrays Change-Id: I306278772f3a2f77d3a3e48941c615c13b3bbb99
* Initial commit * Handle illegal case * Just return the input pointer for illegal cases
… exp) (apache#105) * group_by_returns_TableGroupBy * return_single_type_for_mean_exp * revert table.pyi * compute.mean does not support BinaryScalar or BinaryArray --------- Co-authored-by: ZhengYu, Xu <zen-xu@outlook.com>
3 participants
No description provided.