Current release info

Name	Downloads	Version	Platforms
PyPI — pyramids-core (current)
PyPI — pyramids-gis (deprecated alias)			—
conda-forge — pyramids

pyramids-gis is the deprecated alias of pyramids-core. Install pyramids-core — the import name stays pyramids. pip install pyramids-gis still works (it just depends on the matching pyramids-core and emits a FutureWarning), so existing pins keep resolving. The conda-forge package name is unchanged: pyramids.

conda-forge feedstock

Conda-forge feedstock

pyramids - GIS utility package

pyramids is a GIS utility package built on top of GDAL/OGR for working with raster data (GeoTIFF, NetCDF), vector data (shapefiles, GeoJSON), and multi-temporal datacubes.

pyramids-core is licensed under GPLv3 (see LICENSE.md). The platform wheels published on PyPI bundle GDAL and its native dependencies (PROJ, GEOS, libtiff, NetCDF-C, HDF5, libcurl, …) — each under its own MIT, BSD, LGPL, or Apache license. The full attribution list and shipped license texts are documented in THIRD_PARTY_LICENSES.md; if you use pyramids-core in publications please also cite GDAL itself per gdal.org/cite_gdal.html.

graph LR
    GeoTIFF & NetCDF & Shapefile & UGRID -->|read| pyramids
    subgraph pyramids
        direction TB
        Dataset
        NetCDF_class[NetCDF]
        UgridDataset
        DatasetCollection
        FeatureCollection
        subgraph Engines["Dataset engines (ds.io · ds.spatial · ds.bands · ds.analysis · ds.cell · ds.vectorize · ds.cog)"]
        end
        subgraph Plotting["Plotting layer — _plot_helpers.render_array · mesh_render · NetCDFPlot · Selectors / ColourOpts / FacetSpec · basemap"]
        end
    end
    Dataset -->|crop · reproject · align| Dataset
    Dataset --- Engines
    FeatureCollection -->|rasterize| Dataset
    UgridDataset -->|interpolate| Dataset
    Dataset -->|vectorize| FeatureCollection
    DatasetCollection -->|lazy temporal stack| Dataset
    NetCDF_class -->|extends| Dataset
    Dataset & NetCDF_class & DatasetCollection & UgridDataset -->|plot| Plotting
    Plotting -->|delegates| cleopatra(["cleopatra<br/>ArrayGlyph · MeshGlyph · tiles"])

For the class relationships, internal layers, and detailed architecture diagrams, see docs/overview/architecture.md.

Main Features

• Dataset - Read, write, crop, reproject, and align single-band and multi-band rasters (GeoTIFF) with full no-data handling and coordinate reference system support. Public API is organized into seven engine collaborators (ds.io, ds.spatial, ds.bands, ds.analysis, ds.cell, ds.vectorize, ds.cog); same-named facade methods on the Dataset itself keep the short form working — ds.crop(mask) and ds.spatial.crop(mask) are equivalent.
• NetCDF - Extends Dataset for NetCDF files with time/variable dimensions and CF conventions metadata. Optional labeled-array interoperability. NetCDF.plot exposes a labeled-array-style plotting API (variable= + grouped Selectors / ColourOpts / FacetSpec dataclasses, curvilinear coords=, kind=, animate=, lazy chunks=).
• UgridDataset - Read and visualize UGRID-1.0 unstructured meshes (triangles, quads, mixed). Supports mesh-to-raster interpolation and mesh-to-vector export.
• DatasetCollection - Manage time-series of co-registered rasters as a lazy temporal stack (per-timestep gdal handles open on demand; the full cube is never materialised in RAM) with optional dask-backed reductions and groupby.
• FeatureCollection - Work with vector data (shapefiles, GeoJSON) through a unified GeoDataFrame and OGR DataSource interface, including rasterization and geometry operations.
• Plotting - Dataset / NetCDF / DatasetCollection / UgridDataset all expose a plot method backed by cleopatra (the [viz] extra), routed through a shared pyramids.dataset._plot_helpers core. Optional web-tile basemap underlays via pyramids.basemap.add_basemap (a thin wrapper over cleopatra.tiles.add_tiles).
• Cloud-Optimized GeoTIFF (COG) - First-class read/write/validate support via ds.to_cog, ds.is_cog, and ds.validate_cog.
• Spatial operations - Align rasters to a reference grid, reproject between coordinate systems, crop to vector boundaries, and convert between raster, NetCDF, and vector formats.

Installing pyramids

Installing pyramids from the conda-forge channel can be achieved by:

conda install -c conda-forge pyramids

It is possible to list all the versions of pyramids available on your platform with:

conda search pyramids --channel conda-forge

Install from GitHub (development)

To install the latest development version, you can install the library from GitHub:

pip install git+https://github.com/serapeum-org/pyramids

Note: installing from GitHub uses the sdist and requires a pre-installed system GDAL. See the full installation guide and troubleshooting for details.

pip

To install the latest release from PyPI:

pip install pyramids-core

Linux + pixi: no glibc pin needed anymore

pyramids-core ships its Linux wheels tagged manylinux_2_28 (GDAL and its native stack are compiled from source with the manylinux toolchain). pixi's default Linux baseline is glibc 2.28 (it tracks conda-forge's floor), so the wheel resolves out of the box — no [tool.pixi.system-requirements] entry required (verified with pixi 0.65 defaults; newer versions, including the 0.68.1 this repo pins in CI, share the same baseline).

Two cases still need a pin in the consuming project's pyproject.toml / pixi.toml:

[tool.pixi.system-requirements]
libc = "2.39"    # only for the older releases that shipped manylinux_2_39 wheels (0.2x-0.39.x)

or libc = "2.28" if you run a pixi version old enough that its default baseline is still below 2.28. On Linux with glibc < 2.28, install from conda-forge instead. See the full installation guide and troubleshooting for the other cases.

Optional extras

pip install pyramids-core[viz]      # cleopatra plotting support
pip install xarray                 # to_xarray / from_xarray / to_netcdf interop (peer dep, not an extra)

Quick start

from pyramids.dataset import Dataset

# Open a raster file
src = Dataset.read_file("path/to/raster.tif")
print(src.epsg)        # coordinate reference system EPSG code
print(src.cell_size)   # pixel resolution
print(src.shape)       # (bands, rows, columns)

# Read the raster data as a NumPy array
arr = src.read_array()                  # all bands
band0 = src.read_array(band=0)          # one band

# Spatial ops route through the spatial engine; the facade stays short
reprojected = src.to_crs(to_epsg=3857)  # same as src.spatial.to_crs(...)

from pyramids.netcdf import NetCDF
from pyramids import Selectors, ColourOpts, FacetSpec   # grouped plot options

# Open a NetCDF file
nc = NetCDF.read_file("path/to/data.nc")
print(nc.variables)

# labeled-array-style plotting (needs the [viz] extra)
nc.plot("t2m", selectors=Selectors(time="2020-07-01", level=850),
        colour=ColourOpts(cmap="coolwarm", robust=True))
nc.plot("t2m", facet=FacetSpec(col="time", col_wrap=4))   # small multiples
nc.plot("t2m", animate="time", chunks={"time": 1})        # lazy per-frame animation

from pyramids.feature import FeatureCollection

# Open a vector file
vector = FeatureCollection.read_file("path/to/shapefile.shp")
print(vector.epsg)            # CRS EPSG code
print(vector.total_bounds)    # (minx, miny, maxx, maxy)

from pyramids.dataset import DatasetCollection

# Build a lazy stack of co-registered rasters (no pixels read yet)
cube = DatasetCollection.from_files(["a.tif", "b.tif", "c.tif"])
print(cube.time_length, cube.shape)

# Reductions over the time axis use dask under the hood
mean = cube.mean()                       # nan-aware by default

Testing

This project uses pixi as the environment and task manager.

# Install dependencies and create dev environment
pixi install -e dev

# Run all tests (excluding plot tests)
pixi run -e dev main

# Run plot tests only
pixi run -e dev plot

# Run a specific test file
pixi run -e dev pytest tests/netcdf/test_dimensions.py -v

# Run a single test by node id
pixi run -e dev pytest tests/netcdf/test_dimensions.py::TestStripBraces::test_with_braces -q

Docker

A Dockerfile is provided to run pyramids-core in a controlled environment with the correct GDAL stack preinstalled via conda-forge. The image uses a multi-stage pixi build for a minimal production container.

Build the image:

docker build -t pyramids-core:latest .

Run the container (mount your current folder as /workspace):

docker run --rm -it -v ${PWD}:/workspace pyramids-core:latest bash

Inside the container you can verify the package is installed:

python -c "import pyramids; print('pyramids', pyramids.__version__)"