tidy3A

Tidy-style verbs for named multidimensional arrays: point clouds, images, sensor measurements, scientific and complex-valued arrays. Built on xarray and NumPy, in the spirit of tidy3.

Status: version 0.1.0 is prepared for the first release, not yet published. The design decisions are in docs/plan.md; the rules fit on one page in docs/decisions.md.

Requires Python 3.11 or newer. From a checkout, install the package:

python -m pip install -e .

Quick start

import numpy as np
from tidy3a import fields, filter, formula, infer, mutate, read_array

points = np.array([[3, 4, 0], [0, 0, 50], [60, 0, 0]], dtype="float32")
cloud = read_array(
    points,
    dims={"point": infer, "feature": 3},
    labels={"feature": ["x", "y", "z"]},
    field_axis="feature",
)
x, y, z, distance = fields("x", "y", "z", "distance")

nearby = (
    cloud
    >> mutate(distance = formula(r"\sqrt{x^2 + y^2 + z^2}"))
    >> filter(distance <= 50)
)
print(nearby.collect()["distance"].values)  # [ 5. 50.]

The implemented API includes read_array, read_binary, fixed_text, infer, tidy(), tidy-layer verbs (select, filter, slice, mutate, group_by, summarise, joins, pull, bind_cols), array helpers (dot, numeric functions, conj, rename_axes, pack), and formula() for plain mathematical text and a strict LaTeX dialect. Scripts declare names with fields() or write col("x"). collect() returns a computed xarray Dataset; to_xarray() returns a shallow copy sharing the underlying values.

Notebooks

Install IPython separately to use the notebook extension. In a notebook cell:

%load_ext tidy3a.jupyter
from tidy3a import *

The extension enables bare field names inside verbs, so the quick-start example can omit the fields() declaration. Multi-line pipes need parentheses. Sharing a notebook with tidy3 requires its ownership-aware transformer; the prerequisite is recorded in the plan.

Execution and first-release scope

The first release runs eagerly on NumPy. read_array accepts homogeneous raw numeric files, .npy files, and NumPy arrays. read_binary reads mixed fixed-width records into separate variables, preserving numeric types and strictly decoding UTF-8/ASCII text declared with fixed_text. Records can be packed or have explicit field offsets and padding; see the reader contract for supported options. tidy() wraps xarray DataArrays and Datasets; optional file readers stay in xarray's ecosystem.

read_array(..., lazy=True) memory-maps raw and .npy files. It does not make the verb pipeline lazy: arithmetic can allocate results, filtering evaluates its mask, and updating a field label copies its containing array into memory. tidy() preserves a lazy xarray input initially; collect() computes it. Dask execution guarantees remain planned.

Integer arithmetic follows NumPy, including unsigned subtraction wrapping around. Convert before subtracting when negative results are needed, for example nir * 1.0 - red for uint8 image bands. See first-release limitations for the supported subset.

Examples and development

The point-cloud example runs on real nuScenes Mini data and checks itself against plain NumPy:

python examples/nuscenes_sweep.py /path/to/nuscenes-mini

Three public sample datasets (gridded air temperature, ERA-Interim wind, and stars' Landsat scene) run against plain xarray; see examples/sample_data.py for the files, optional reader dependencies, and checks.

To run the local tests:

python -m pip install -e ".[dev]"
python -m pytest -q

Before pushing, run the CI checks in fresh environments with only declared dependencies (requires uv):

scripts/check.sh

The release procedure describes the GitHub Release workflow and the one-time GitHub/PyPI setup.

License

MIT