datoviz 0.4.0rc1

Datoviz: high-performance rendering for scientific data visualization

Datoviz

Datoviz is a GPU-powered visualization engine for interactive scientific data. Use it when dense points, images, meshes, volumes, annotations, linked panels, or custom scientific scenes need more control and performance than an ordinary plotting library provides.

Datoviz v0.4 provides a retained scene model, native Vulkan rendering, desktop and offscreen presentation, reproducible capture, and direct use from C or Python. Python users call the generated binding with import datoviz as dvz and pass NumPy arrays directly to supported data-upload APIs.

v0.4 Release Status

The v0.4-dev branch is the first v0.4 release candidate. The native C and Python/NumPy paths are release-facing; experimental and deferred surfaces remain explicitly labeled throughout the documentation.

Surface	v0.4 status
Native C scene/app API	supported, with feature-specific gaps
Python API with NumPy adaptation and exact raw calls	supported
Offscreen rendering and capture	supported
Qt/PyQt hosted rendering	supported, optional provider
Retained visual families	supported/experimental by family
WebGPU/WASM browser path	experimental
Scene compute-to-render integration	experimental
DRP2 command streams and runtime internals	advanced/unstable
High-level plotting API	external/GSP

See the detailed feature status, platform support, and v0.3 capability disposition before relying on an experimental or backend-specific feature.

Install

Once a v0.4 release candidate is published on PyPI, use the exact command from its release notes. A pre-release install will normally look like:

python -m pip install --pre datoviz

After the final v0.4 release, the normal command will be:

python -m pip install datoviz

Until packages are published for your platform, use the source build below. The full installation guide covers macOS, Linux, Windows, Python, and C/C++ integration.

Minimal Python Example

This deterministic example creates the scatter plot shown in the documentation: 10,000 random points with random colors and sizes on a dark background.

import ctypes

import numpy as np
import datoviz as dvz

N = 10_000
rng = np.random.default_rng(12345)
pos = np.zeros((N, 3), dtype=np.float32)
pos[:, :2] = rng.uniform(-1, 1, (N, 2))
color = rng.integers(0, 255, (N, 4), dtype=np.uint8)
color[:, 3] = 200
diameter = rng.uniform(4, 12, N).astype(np.float32)

scene = dvz.dvz_scene()
figure = dvz.dvz_figure(scene, 1280, 720, 0)
panel = dvz.dvz_panel_full(figure)
dvz.dvz_panel_set_background_color(panel, dvz.DvzColor(13, 18, 25, 255))

panzoom = dvz.dvz_panzoom(scene, None)
dvz.dvz_panel_bind_controller(panel, panzoom, dvz.DvzDimMaskFlag.DVZ_DIM_MASK_XY)

points = dvz.dvz_point(scene, 0)
dvz.dvz_visual_set_data(points, "position", pos)
dvz.dvz_visual_set_data(points, "color", color)
dvz.dvz_visual_set_data(points, "diameter_px", diameter)

style = dvz.dvz_point_style_desc()
style.aspect = dvz.DVZ_SHAPE_ASPECT_FILLED
style.stroke_width_px = 0
dvz.dvz_point_set_style(points, ctypes.byref(style))
dvz.dvz_visual_set_depth_test(points, False)
dvz.dvz_visual_set_alpha_mode(points, dvz.DVZ_ALPHA_BLENDED)
dvz.dvz_panel_add_visual(panel, points, None)

dvz.run(scene, figure, title="Datoviz")

Continue with the annotated Quickstart or browse the example gallery.

C And C++

Datoviz is a native C library. Installed packages expose headers and a CMake package for C and C++ applications:

find_package(datoviz CONFIG REQUIRED)
target_link_libraries(my_app PRIVATE datoviz::datoviz)

The public API uses C linkage and can be called directly from C++. See C/C++ integration and the generated C API reference.

Build From Source

Source builds require Git, CMake 3.21+, Ninja, just, Python 3.10+, a supported C/C++ compiler, shader tools, and a Vulkan-capable runtime. Clone the active branch with its submodules, then build and test:

git clone --branch v0.4-dev --recursive https://github.com/datoviz/datoviz.git
cd datoviz
just build
just test
python -m pip install -e .

Platform packages, compiler versions, Vulkan/MoltenVK requirements, and native Windows guidance are maintained in the installation guide. Contributors should also read CONTRIBUTING.md and BUILD.md.

Which Layer Should I Use?

Need	Use
Python with Datoviz visuals and NumPy arrays	import datoviz as dvz
Native application or C/C++ integration	C scene/app API
Exact pointer/count form of the Python binding	datoviz.raw
Browser rendering for promoted examples	experimental WebGPU/WASM subset
High-level functions such as scatter() or imshow()	GSP/VisPy2 when available

Datoviz v0.4 is the explicit engine layer: scenes, visuals, data uploads, controllers, windows, captures, and integration surfaces. The old high-level Datoviz Python plotting API is not part of v0.4; that role belongs to the developing GSP/VisPy2 layer. See Choose Your Layer for the complete comparison.

Documentation

• Get Started
• Examples and Gallery
• How-To Guides
• Feature Status
• Platform Support
• Python API
• C API Reference
• WebGPU/WASM Subset
• Changelog
• Citation

License And Credits

Datoviz is released under the MIT license. It is developed by Cyrille Rossant at the International Brain Laboratory, with support from the Wellcome Trust, Simons Foundation, and Chan Zuckerberg Initiative.

If you use Datoviz in research, follow the current citation guidance and repository metadata in CITATION.cff. The final v0.4.0 release is planned for archival with Zenodo and a version-specific DOI.

Datoviz builds on earlier open-source GPU visualization work including VisPy, Glumpy, Galry, and the Vulkan-based Datoviz releases. Development of v0.4 was assisted by OpenAI Codex for implementation, review, testing, documentation, and release preparation. All changes were directed, reviewed, and validated by the project maintainer.