---

GeoZL extends OpenZL with codecs designed for Earth observation data. It adds spatial prediction, NoData handling, efficient integer packing and bounded-error quantization for multidimensional numeric arrays. The same core is available from C, Python, R, Julia and JavaScript.

Bindings

Language	Install	Role	Docs
Python	pip install geozl	full API and OpenZL nodes	README
R	install.packages("geozl", repos = "https://asterisk-labs.r-universe.dev")	graph, compress, decompress, profile	README
Julia	Pkg.Registry.add(url="https://github.com/asterisk-labs/AsteriskRegistry"); Pkg.add("GeoZL")	graph, compress, decompress, profile	README
JavaScript	npm install @asterisk-labs/geozl	compress, decompress	README

Quick start

import numpy as np
import geozl

y, x = np.mgrid[0:1024, 0:1024]
tile = (2000 + 8 * y + 5 * x).astype(np.uint16)

best = geozl.profile(tile)[0]["graph"]        # rank recipes on a sample
g = geozl.graph(tile, best)                   # build once, reuse per tile
frame = geozl.compress(tile, graph=g)
back = geozl.decompress(frame).view(np.uint16).reshape(tile.shape)

Codecs

codec	CTid	what it does
delta_w	0x72D701	residual against the west neighbour
delta_n	0x72D702	residual against the north neighbour
planar	0x72D703	predicts each pixel from W + N - NW
deinterleave	0x72D704	separates a two-lane interleaved stream
med	0x72D705	median edge detector predictor
average	0x72D706	floor average of the west and north neighbours
wp_static	0x72D707	weighted predictor with its weights in the frame
nodata	0x72D70C	missing samples to a validity mask, holes filled
pfor	0x72D70D	bit packs blocks of 256 and patches overflows
planar_zigzag	0x72D70F	planar residuals and Zigzag in one pass
planar_zigzag_pfor	0x72D710	planar, Zigzag and PFOR as one codec
med_zigzag	0x72D711	MED residuals and Zigzag in one pass
planar_zigzag_pivco	0x72D712	planar and Zigzag, then PivCo or PFOR per byte lane
quant_linear	0x72D781	absolute bound: LINEAR:MAX_ERROR=V
quant_log	0x72D782	relative bound: LOG:MAX_ERROR=P%
quant_sqrt	0x72D783	noise-scaled bound: SQRT:MAX_ERROR=VN

The codec catalog documents each wire format.

AI agent skill

Install the GeoZL skill so coding agents know its graph recipes, bounded-error rules and codec internals.

npx skills add asterisk-labs/geozl

Development

make submodules
make test       # C tests and pytest
make help       # R, Julia, wasm, sanitizers and fuzzing

License

BSD-3-Clause.

Made with ♥ by