Package: altarr 0.1.0.9000

Michael Sumner

altarr: Lazy Chunked Arrays as ALTREP Vectors

A prototype exploring how far base R array indexing can drive planned reads of remote, chunked, encoded arrays when the array is an ALTREP vector with a 'dim' attribute and no class. Chunks come from a user-supplied fetch function, so any store (object storage, 'Zarr', 'GDAL') can sit behind it.

Authors:Michael Sumner [aut, cre]

altarr_0.1.0.9000.tar.gz
altarr_0.1.0.9000.zip(r-4.7-x86_64)altarr_0.1.0.9000.zip(r-4.7-arm64)altarr_0.1.0.9000.zip(r-4.6-x86_64)altarr_0.1.0.9000.zip(r-4.6-arm64)altarr_0.1.0.9000.zip(r-4.5-x86_64)
altarr_0.1.0.9000.tgz(r-4.6-x86_64)altarr_0.1.0.9000.tgz(r-4.6-arm64)altarr_0.1.0.9000.tgz(r-4.5-x86_64)altarr_0.1.0.9000.tgz(r-4.5-arm64)
altarr_0.1.0.9000.tar.gz(r-4.7-arm64)altarr_0.1.0.9000.tar.gz(r-4.7-x86_64)altarr_0.1.0.9000.tar.gz(r-4.6-arm64)altarr_0.1.0.9000.tar.gz(r-4.6-x86_64)
altarr_0.1.0.9000.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION |NEWS
card.svg |card.png
altarr/json (API)

# Install 'altarr' in R:
install.packages('altarr', repos = c('https://hypertidy.r-universe.dev', 'https://cloud.r-project.org'))

Bug tracker:https://github.com/hypertidy/altarr/issues

Pkgdown/docs site:https://hypertidy.github.io

On CRAN:

Conda:

3.30 score 2 stars 10 scripts 8 exports 0 dependencies

Last updated from:8b962d5d80. Checks:15 OK. Indexed: yes.

TargetResultTimeFilesSyslog
linux-devel-arm64OK133
linux-devel-x86_64OK109
source / vignettesOK173
linux-release-arm64OK76
linux-release-x86_64OK97
macos-release-arm64OK87
macos-release-x86_64OK197
macos-oldrel-arm64OK89
macos-oldrel-x86_64OK242
windows-devel-arm64OK171
windows-devel-x86_64OK122
windows-release-arm64OK185
windows-release-x86_64OK77
windows-oldrel-x86_64OK70
wasm-releaseOK73

Exports:altarraltarr_example_zarraltarr_extractaltarr_planaltarr_resetaltarr_statsaltarr_zarr_v2is_altarr

Dependencies:

Walkthrough: what base R asks a lazy array for
The example store | 1. Opening a store reads nothing | 2. One value reads one chunk | 3. The cache: reading it again is free | 4. Points: x[cbind(i, j, k)] is one planned request | 5. A rectangle: x[i, j, k] reads element by element | 6. The same rectangle, planned | 7. Materialization: when R wants all the data | 8. With a lower limit, materializing is refused | 9. Copies stay lazy until changed | 10. Saving stores the recipe, not the data | Where to go next
Rendered fromwalkthrough.Rmdusingknitr::rmarkdown

Last update: 2026-10-03
Started: 2026-10-03

Writing a fetch function
1. The smallest fetch function | 2. Test against a reference | 3. What altarr checks for you | 4. One call, many chunks | 5. Dimension order: reverse, never transpose | 6. Types, missing values and unpacking | 7. Recipes: what saveRDS() keeps | 8. Sources must be pinned | 9. Sketches for real backends | 10. Tools while developing
Rendered fromfetch.Rmdusingknitr::rmarkdown

Last update: 2026-10-03
Started: 2026-10-03