v2026.5.16

Author: csaybar

.gitignore

@@ -10,3 +10,4 @@ python/*.egg-info/
 .DS_Store
 
 python/cozip/_lib/cozip.dylib
+julia/Manifest.toml

VERSION

@@ -1 +1 @@
-2026.5.15
+2026.5.16

julia/Artifacts.toml

@@ -2,51 +2,51 @@
 arch = "aarch64"
 os = "linux"
 libc = "glibc"
-git-tree-sha1 = "400981bd94b62948b56c79e90af58d38f8d108c5"
+git-tree-sha1 = "db4487e5495c81e1dd90adcd2d4e8a4afe6d4ce7"
 lazy = true
 
     [[cozip.download]]
-    url = "https://github.com/asterisk-labs/taco/releases/download/v2026.5.13/libcozip-2026.5.13-linux-aarch64.tar.gz"
-    sha256 = "ead9ad66d1b92b2178843f5a2bec355c1c338b8992c5bd0016c4399f6ae7382d"
+    url = "https://github.com/asterisk-labs/cozip/releases/download/v2026.5.15/libcozip-2026.5.15-linux-aarch64.tar.gz"
+    sha256 = "40d41e3c99ab95dfdec929bdde149b3d65d16cd62c352cae74c7d80a2378f449"
 
 [[cozip]]
 arch = "x86_64"
 os = "linux"
 libc = "glibc"
-git-tree-sha1 = "6fff3000efb6dc986004e568585016723cbe60dc"
+git-tree-sha1 = "32989d32fce26d689e604468b9462f6f13d995de"
 lazy = true
 
     [[cozip.download]]
-    url = "https://github.com/asterisk-labs/taco/releases/download/v2026.5.13/libcozip-2026.5.13-linux-x86_64.tar.gz"
-    sha256 = "df07301874a49db58a23eaad2452674b6b5dc78ff0c7684a19aa92f1d82566d2"
+    url = "https://github.com/asterisk-labs/cozip/releases/download/v2026.5.15/libcozip-2026.5.15-linux-x86_64.tar.gz"
+    sha256 = "0793bc68510bd54ed20c14b24953b66515dda7e3936a1146499c3c96dec8023c"
 
 [[cozip]]
 arch = "x86_64"
 os = "macos"
-git-tree-sha1 = "2c1241eca9a121beb77d73d1b7ac8dba7bdb09f7"
+git-tree-sha1 = "b7cfdf3a2a2b1737e48d6560136d4a78d19552fb"
 lazy = true
 
     [[cozip.download]]
-    url = "https://github.com/asterisk-labs/taco/releases/download/v2026.5.13/libcozip-2026.5.13-macos-universal.tar.gz"
-    sha256 = "752b2018947b9f245c1f9c3c3139b51571fb05dc7ddc5ff080df759e4b0c8b97"
+    url = "https://github.com/asterisk-labs/cozip/releases/download/v2026.5.15/libcozip-2026.5.15-macos-universal.tar.gz"
+    sha256 = "93eaa7152b1bb963f49726487d4458ee73eb72b2b5bb3a484a4c6aa4b960361c"
 
 [[cozip]]
 arch = "aarch64"
 os = "macos"
-git-tree-sha1 = "2c1241eca9a121beb77d73d1b7ac8dba7bdb09f7"
+git-tree-sha1 = "b7cfdf3a2a2b1737e48d6560136d4a78d19552fb"
 lazy = true
 
     [[cozip.download]]
-    url = "https://github.com/asterisk-labs/taco/releases/download/v2026.5.13/libcozip-2026.5.13-macos-universal.tar.gz"
-    sha256 = "752b2018947b9f245c1f9c3c3139b51571fb05dc7ddc5ff080df759e4b0c8b97"
+    url = "https://github.com/asterisk-labs/cozip/releases/download/v2026.5.15/libcozip-2026.5.15-macos-universal.tar.gz"
+    sha256 = "93eaa7152b1bb963f49726487d4458ee73eb72b2b5bb3a484a4c6aa4b960361c"
 
 [[cozip]]
 arch = "x86_64"
 os = "windows"
-git-tree-sha1 = "06f58edee9575a5a7e836e3f0503f43ea92602b4"
+git-tree-sha1 = "7c9623e87520b2c471b9ea03ae5c16b97d410bf1"
 lazy = true
 
     [[cozip.download]]
-    url = "https://github.com/asterisk-labs/taco/releases/download/v2026.5.13/libcozip-2026.5.13-windows-x86_64.tar.gz"
-    sha256 = "1da60e6a5099c8ad701d97403f255de033ce5d67eb11f9e7dfab0f944430a3e1"
+    url = "https://github.com/asterisk-labs/cozip/releases/download/v2026.5.15/libcozip-2026.5.15-windows-x86_64.tar.gz"
+    sha256 = "019f1405d81a28b0c0341cf8fcc2400352cd082a44e1019b2d01e466dbb5845d"
 

julia/LICENSE

@@ -1,6 +1,6 @@
 name = "Cozip"
 uuid = "7c91d8e4-3f2a-4b6c-9d18-2e5f7a8b3c4d"
-version = "2026.5.15"
+version = "2026.5.16"
 authors = ["Cesar Aybar <cesar@asterisk.coop>"]
 
 [deps]
@@ -12,6 +12,11 @@ LazyArtifacts = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 
 [compat]
+julia = "1.10"
+DataFrames = "1"
+DuckDB = "1"
+DBInterface = "2"
+Tables = "1"
 GeoInterface = "1.6.1"
 GeoParquet = "0.2.1"
 
@@ -21,4 +26,4 @@ GeoParquet = "e99870d8-ce00-4fdd-aeee-e09192881159"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Test", "GeoParquet", "GeoInterface"]
+test = ["Test", "GeoParquet", "GeoInterface"]

julia/Project.toml

@@ -1,28 +1,22 @@
 # Cozip.jl
 
-Julia binding for [libcozip](https://github.com/asterisk-labs/taco/tree/main/cozip) — pack files into a Cloud-Optimized ZIP archive readable over HTTP range requests.
+Julia binding for libcozip. Open a Cloud-Optimized ZIP archive like a table over HTTP range requests, or write one from a DataFrame.
 
-The native `libcozip` binary is fetched automatically via Julia Artifacts; no C toolchain required.
+The native `libcozip` binary is fetched automatically via Julia Artifacts, no C toolchain required.
 
 ## Install
 
-From the AsteriskRegistry (recommended):
+`Cozip.jl` lives in the AsteriskRegistry.
 
 ```julia
 using Pkg
-Pkg.Registry.add(RegistrySpec(url = "https://github.com/asterisk-labs/AsteriskRegistry"))
+Pkg.Registry.add("https://github.com/asterisk-labs/AsteriskRegistry")
 Pkg.add("Cozip")
 ```
 
-Or directly from the monorepo:
-
-```julia
-Pkg.add(url = "https://github.com/asterisk-labs/taco", subdir = "cozip/julia")
-```
-
 ## Usage
 
-### Simple mode
+### Write
 
 ```julia
 using Cozip, DataFrames
@@ -32,52 +26,59 @@ table = DataFrame(
     path = ["/path/to/a.txt", "/path/to/b.bin"],
 )
 
-Cozip.create("out.cozip", table)
+Cozip.write("out.zip", table)
 ```
 
-### Two-step mode
-
-For inspecting or tuning the `__metadata__` parquet between steps:
+`name` is how each file appears inside the archive. `path` is where it lives on disk, consumed at write time and dropped from the manifest. Any additional columns ride along into `__metadata__` and become queryable on read.
 
 ```julia
-# 1) materialize the __metadata__ parquet
-Cozip.metadata("meta.parquet", table; create_options = "COMPRESSION 'zstd'")
+table = DataFrame(
+    name      = ["a.tif", "b.tif"],
+    path      = ["/path/to/a.tif", "/path/to/b.tif"],
+    cloud_pct = [12.3, 45.1],
+)
 
-# 2) pack the archive
-Cozip.create("out.cozip", "meta.parquet")
+Cozip.write("out.zip", table)
 ```
 
-`create_options` is passed through to DuckDB's `COPY ... TO '...' (FORMAT parquet, <opts>)`, so anything DuckDB accepts works (e.g. `"COMPRESSION 'zstd', ROW_GROUP_SIZE 100000"`).
+### Read
 
-### Extra columns
+```julia
+using Cozip
+
+manifest = Cozip.read("https://example.com/dataset.zip")
+```
 
-Any additional columns in `table` are propagated into `__metadata__`:
+`manifest` is a DataFrame with `name`, `offset`, `size`, plus whatever extras the writer added. Local file or remote URL, same call. Only the byte-0 index and the embedded `__metadata__` Parquet are fetched, never the user payloads.
+
+Filter the manifest like any DataFrame, then use `offset` and `size` to range-request payloads.
 
 ```julia
-table = DataFrame(
-    name      = ["a.tif", "b.tif"],
-    path      = ["/path/to/a.tif", "/path/to/b.tif"],
-    cloud_pct = [12.3, 45.1],
+using DataFrames, Downloads
+
+train = filter(:split => ==("train"), manifest)
+row = train[1, :]
+buf = IOBuffer()
+Downloads.download(
+    "https://example.com/dataset.zip", buf;
+    headers = ["Range" => "bytes=$(row.offset)-$(row.offset + row.size - 1)"],
 )
-
-Cozip.create("out.cozip", table)
+payload = take!(buf)
 ```
 
-The optional `in_index` column (default `true`) controls whether each entry is recorded in the cozip index — entries with `in_index = false` go into the ZIP but aren't exposed as cozip-indexed entries.
-
 ## Versioning
 
-`Cozip.jl` tracks the C library, which uses CalVer with 4 components (e.g. `2026.5.2.6`). The Julia package itself uses 3 components because Julia enforces strict SemVer; the fourth is exposed via:
+`Cozip.jl` tracks the C library. The C side uses 4-component CalVer (e.g. `2026.5.2.6`). The Julia side uses the first three because Julia enforces strict SemVer. The fourth component is exposed at runtime.
 
 ```julia
 using Cozip
-Cozip.LibCozip.cozip_version()  # → "2026.5.2.6"
+Cozip.LibCozip.cozip_version()  # "2026.5.2.6"
 ```
 
 ## Spec
 
-See the [cozip spec](https://github.com/asterisk-labs/taco/tree/main/cozip) for the on-disk format.
+See [SPEC.md](https://github.com/asterisk-labs/cozip/blob/main/SPEC.md) for the on-disk format.
 
 ## License
 
-MIT — see [LICENSE](../LICENSE).
+MIT. See [LICENSE](../LICENSE).

julia/README.md

@@ -2,7 +2,7 @@ using Pkg.Artifacts
 using Downloads
 using SHA
 
-const REPO = "asterisk-labs/taco"
+const REPO = "asterisk-labs/cozip"
 const VERSION = if !isempty(ARGS)
     strip(ARGS[1])
 else

julia/make-artifacts-toml.jl

@@ -43,9 +43,9 @@ classifiers = [
 ]
 
 [project.urls]
-Homepage = "https://asterisk.coop/taco/cozip"
-Repository = "https://github.com/asterisk-labs/taco"
-Issues = "https://github.com/asterisk-labs/taco/issues"
+Homepage = "https://github.com/asterisk-labs/cozip"
+Repository = "https://github.com/asterisk-labs/cozip"
+Issues = "https://github.com/asterisk-labs/cozip/issues"
 
 [project.optional-dependencies]
 test = [
@@ -54,31 +54,21 @@ test = [
     "shapely>=2.0",
 ]
 
-# Single source of truth for the version: cozip/VERSION.
-# Hatchling reads it at build time so wheel metadata and tag stay in sync.
 [tool.hatch.version]
 path = "../VERSION"
 pattern = "^(?P<version>.+)$"
 
 [tool.hatch.build.targets.wheel]
 packages = ["cozip"]
-# Include the prebuilt native lib in the wheel.
 artifacts = ["cozip/_lib/*.dylib", "cozip/_lib/*.so", "cozip/_lib/*.dll"]
 
-# README and LICENSE live one directory up (shared with C/R/Julia bindings).
-# `force_include` copies them into the wheel root so PyPI / pip metadata
-# can find them.
 [tool.hatch.build.targets.wheel.force-include]
 "../README.md" = "README.md"
 "../LICENSE" = "LICENSE"
 
-# Mark the wheel as platform-specific (because of the bundled .dylib),
-# but ABI-agnostic (cffi ABI mode works on any Python 3.x).
 [tool.hatch.build.targets.wheel.hooks.custom]
 path = "hatch_build.py"
 
-# The sdist must be self-contained: it lives at cozip/python/ but
-# needs files from cozip/ (VERSION, README.md, LICENSE).
 [tool.hatch.build.targets.sdist]
 include = [
     "cozip/**",

python/pyproject.toml

@@ -1,17 +1,17 @@
 Package: cozip
 Type: Package
-Title: Cloud-Optimized ZIP — Open a ZIP Like a Table
-Version: 2026.5.15
-Date: 2026-04-30
+Title: Cloud-Optimized ZIP - Open a ZIP Like a Table
+Version: 2026.5.16
+Date: 2026-05-15
 Authors@R: c(
     person("Cesar", "Aybar", email = "cesar@asterisk.coop", role = c("aut", "cre")),
     person("Roy", "Yali", email = "ryali93@gmail.com", role = "aut"),
     person("Julio", "Contreras", email = "julio.contreras@uv.es", role = "aut"))
 Description: R bindings to libcozip, the reference writer for the
-    Cloud-Optimized ZIP (cozip) format. 
+    Cloud-Optimized ZIP (cozip) format.
 License: MIT + file LICENSE
-URL: https://asterisk.coop/taco/cozip, https://github.com/asterisk-labs/taco
-BugReports: https://github.com/asterisk-labs/taco/issues
+URL: https://github.com/asterisk-labs/cozip
+BugReports: https://github.com/asterisk-labs/cozip/issues
 Encoding: UTF-8
 Language: en-US
 Depends:
@@ -27,4 +27,4 @@ Suggests:
 SystemRequirements: GNU make
 Config/testthat/edition: 3
 Roxygen: list(markdown = TRUE)
-RoxygenNote: 7.3.3
+RoxygenNote: 7.3.3