wehub-resource-sync
87 lines
3.9 KiB
Markdown
87 lines
3.9 KiB
Markdown
# Installing Arrow C++ separately
|
|
|
|
\tableofcontents
|
|
|
|
## Automatically download & build Arrow from source (default)
|
|
|
|
By default, the Rerun C++ SDK's CMake script (which is part of the SDK's zip artifact that can be fetched via `FetchContent`)
|
|
will download a known compatible version of Arrow C++ from GitHub and add it to the build.
|
|
The build configuration is kept to the minimum required by the Rerun C++ SDK.
|
|
|
|
To instead use an existing install of Arrow, disable the CMake build option `RERUN_DOWNLOAD_AND_BUILD_ARROW`.
|
|
(by passing `-DRERUN_DOWNLOAD_AND_BUILD_ARROW=OFF` to your CMake configure step).
|
|
This will cause Rerun to instead use CMake's `find_package` to look for a ready-to-use install of the Arrow C++ library.
|
|
|
|
For more information about CMake config options see [C++ SDK CMake](cmake_setup_in_detail.md).
|
|
|
|
## Install libarrow with Pixi
|
|
|
|
[Pixi](https://pixi.sh/latest/) is a convenient tool for managing cross-platform project dependencies. In
|
|
fact, Rerun uses it for our own internal development dependency management, and you will find `pixi.toml` files in most
|
|
of our external examples.
|
|
|
|
Make sure to use `-DRERUN_DOWNLOAD_AND_BUILD_ARROW=OFF` when building, otherwise Rerun's CMake script
|
|
will download & build Arrow instead, ignoring your Pixi install.
|
|
The advantage of using Pixi is that you can rely on pre-built artifacts rather than adding Arrow's build to your own.
|
|
Also, Pixi is of course also useful for managing other dependencies like Eigen or OpenCV,
|
|
as well as for pinning the version of your build tooling.
|
|
|
|
### Installing Pixi
|
|
On Mac or Linux you can just run:
|
|
```
|
|
curl -fsSL https://pixi.sh/install.sh | bash
|
|
```
|
|
Or on Windows:
|
|
```
|
|
iwr -useb https://pixi.sh/install.ps1 | iex
|
|
```
|
|
|
|
Alternatively if you are already a `cargo` user, you can install `pixi` via:
|
|
```
|
|
cargo install pixi
|
|
```
|
|
|
|
See the [Pixi installation guide](https://pixi.sh/latest/installation/) for other installation options.
|
|
|
|
### Adding Pixi to your own project
|
|
|
|
If you want to use `pixi` to manage dependencies in your own project, you can simply run `pixi init` in the root of your
|
|
project folder. This will create a `pixi.toml` file that manages the project. After that you can run
|
|
`pixi add libarrow==18.0.0` to add Arrow C++ as a dependency to your project.
|
|
|
|
If you install the `rerun-sdk` Python package into the same environment, take care to pick a compatible Arrow version.
|
|
On Linux & Mac you can check the current Arrow version for instance using `curl -s https://pypi.org/pypi/rerun-sdk/json | jq -r | grep pyarrow`.
|
|
|
|
Now, any Pixi tasks added to your project will have access to the `libarrow` library.
|
|
|
|
Even without tasks, you can run `pixi shell` to create a shell environment where all your project dependencies
|
|
(including `libarrow`) will be available. You can use this `pixi shell` to run you project's build commands.
|
|
|
|
Check out the [Pixi docs](https://pixi.sh/latest/#getting-started) for more information on what you can do with Pixi.
|
|
|
|
### Pixi in action
|
|
|
|
The rerun-cpp example: <https://github.com/rerun-io/cpp-example-opencv-eigen> ships with a `pixi.toml` file to manage
|
|
its dependencies, as well as a set of tasks to simplify running it.
|
|
|
|
If you have Pixi installed, all you need to do to run the example is:
|
|
```
|
|
git clone https://github.com/rerun-io/cpp-example-opencv-eigen
|
|
cd cpp-example-opencv-eigen
|
|
pixi run example
|
|
```
|
|
|
|
### Known issues
|
|
|
|
⚠️ On Windows Pixi only downloads release binaries which are **not** compatible with debug builds, causing runtime crashes.
|
|
For debug builds you have to build Arrow yourself, see [Building Arrow C++](https://arrow.apache.org/docs/developers/cpp/building.html)
|
|
or stick with `RERUN_DOWNLOAD_AND_BUILD_ARROW=ON`.
|
|
|
|
## Other ways to install libarrow
|
|
|
|
Rerun will also work with any existing environment install of Arrow that works with `find_package`.
|
|
- Arrow provides pre-built packages for many platforms.
|
|
- See the list at: <https://arrow.apache.org/install/>
|
|
- Conda-forge contains a package for Arrow C++:
|
|
- <https://anaconda.org/conda-forge/libarrow>
|