v0.4.0

Highlights

A new algorithm was introduced to sort PackageRecords in a topological order based on their dependencies.
Sorting in this way provides a deterministic way of sorting packages in the order in which they should be installed to avoid clobbering.
The original algorithm was extracted from rattler-server.

Experimental extensions to the conda lock file format have also been introduced to make it possible to completely reproduce the original RepoDataRecords from a lock file.

Fixes were made to the MatchSpec and Version implementation to catch some corner cases and detecting the current shell has become more robust.

Details

Added

• PackageRecord::sort_topologically to perform a topological sort of PackageRecords (#218)
• Experimental fields to be able to reconstruct RepoDataRecord from conda lock files. (#221)
• Methods to manipulate Versions (#229)

Changed

• Refactored shell detection code using $SHELL or parent process name (#219)
• The error message that is thrown when parsing a Platform now includes possible options (#222)
• Completely refactored Version implementation to reduce memory footprint and increase readability (#227)

Fixed

• Issue with parsing matchspecs that contain epochs (#220)
• Zsh activation scripts invoke .sh scripts (#223)
• Detect the proper powershell parent process (#224)

v0.3.0

Highlights

This release contains lots of fixes, small (breaking) changes, and new features.
The biggest highlights are:

JLAP support

JLAP is a file format to incrementally update a cached repodata.json without downloading the entire file.
This can save a huge amount of bandwidth for large repodatas that change often (like those from conda-forge).
If you have a previously cached repodata.json on your system only small JSON patches are downloaded to bring your cache up to date.
The format was initially proposed through a CEP and has been available in conda as an experimental feature since 23.3.0.

When using rattler you get JLAP support out of the box.
No changes are needed.

Support for local file://

file:// based urls are now supported for all functions that use a Url to download certain data.

rattler_networking

A new crate has been added to facilitate authentication when downloading repodata or packages called rattler_networking.

Details

Added

• Support for detecting more platforms (#135)
• RepoData is now clonable (#138)
• RunExportsJson is now clonable (#169)
• file:// urls are now supported for package extraction functions (#157)
• file:// urls are now supported for repodata fetching (#158)
• Getting started with rattler using micromamba (#163)
• Add Platform::arch function to return the architecture of a given platform (#166)
• Extracted Python style JSON formatting into a separate crate (#163)
• Added feature to use rustls with rattler_package_streaming and rattler_repodata_gateway (#179 & #181)
• Expose version_spec module (#183)
• NamelessMatchSpec a variant of MatchSpec that does not include a package name #185)
• ShellEnum - a dynamic shell type for dynamic discovery #187)
• Exposed the python_entry_point_template function (#190)
• Enable deserializing virtual packages (#198)
• Refactored CI to add macOS arm64 (#201)
• Support for JLAP when downloading repodata (#197 & #214)
• Clone, Debug, PartialEq, Eq implementations for conda lock types (#213)
• rattler_networking to enable accessing repodata.json and packages that require authentication (#191)

Changed

• FileMode is now included with prefix_placeholder is set (#136)
• rattler_digest now re-exports commonly used hash types and typed hashes are now used in more placed (instead of strings) [#137 & #153]
• Use Platform in to detect running operating system (#144)
• paths.json is now serialized in a deterministic fashion (#147)
• Determine the subdir for the platform and arch fields when creating a PackageRecord from an index.json (#145 & #152)
• Activator::activation now returns the new PATH in addition to the script (#151)
• Use properly typed chrono::DateTime<chrono::Utc> for timestamps instead of u64 (#157)
• Made ParseError public and reuse ArchiveType (#167)
• Allow setting timestamps when creating package archive (#171)
• about.json and index.json are now serialized in a deterministic fashion (#180)
• SHA256 and MD5 hashes are computed on the fly when extracting packages (#176
• Change blake2 hash to use blake2b instead of blake2s (#192
• LibSolv error messages are now passed through (#202 & #210)
• VersionTree parsing now uses nom instead of a complex regex (#206
• libc version detection now uses lld --version to properly detect the libc version on the host (#209
• Improved version parse error messages (#211
• Parsing of some complex MatchSpecs (#217

Fixed

• MatchSpec bracket list parsing can now handle quoted values (#157)
• Typos and documentation (#164 & #188)
• Allow downloading of repodata.json to fail in some cases (only noarch is a required subdir) (#174)
• Missing feature when using the sparse index (#182)
• Several small issues or missing functionality (#184)
• Loosened strictness of comparing packages in Transactions (#186
• Missing noarch: generic parsing in links.json (#189
• Ignore trailing .0 in version comparison (#196

v0.2.0

Added

• Construction methods for NoArchType (#130)
• Function to create package record from index.json + size and hashes (#126)
• Serialization for repodata (#124)
• Functions to apply patches to repodata (#127)
• More tests and evict removed packages from repodata (#128)
• First version of package writing functions (#112)

Changed

• Removed dependency on clang-sys during build (#131)

v0.1.0

First initial release