Releases: manzt/juv
v0.3.0
Release Notes
This release adds support for generating lockfiles from Jupyter notebooks using inline metadata, as defined in PEP 723. It is based on uv's recent support for locking standalone scripts, except instead of writing a separate lockfile, juv
embeds (and updates) a lockfile in the Jupyter notebook metadata.
By default, notebooks remain unlocked. To lock a notebook, run juv lock /path/to/notebook.ipynb
, which generates and embeds a lockfile in the notebook's metadata under the "uv.lock"
key. The lockfile is respected and updated automatically when using juv run
, uv add
, or uv remove
.
Additional commands:
juv export
: Outputs an alternative lockfile format (requirements.txt
style) to stdout.uv tree
: Displays the dependency tree for a script.
Both commands work with notebooks, whether locked or unlocked.
This release is considered breaking due to the lockfile support, which requires a minimum uv
0.5.18 and modifies execution.
Breaking changes
Enhancements
- Add
--clear
flag tolock
to clear lockfile metadata (#69) - Add
export
command (#70) - Add
lock
command (#64) - Add
tree
command (#68) - Sync lockfile during
add
command (#65) - Sync lockfile during
remove
command (#66)
Bug fixes
- Require at least one package for
add
andremove
(#73) - Support relative paths in the
run
command (#72)
Contributors
v0.2.28
Other changes
- Add
remove
command (#59)
Contributors
Notes
This release adds juv remove
to remove packages from a notebook or script. Dependencies are removed from the PEP-723 inline metadata. The command follows uv's semantics. See the uv docs for more information.
uvx juv init
uvx juv add Untitled.ipynb 'numpy>=1.0.0' 'polars' # adds 'numpy>=1.0.0' 'polars'
uvx juv remove Untitled.ipynb numpy # removes 'numpy>=1.0.0'
v0.2.27
v0.2.26
v0.2.25
v0.2.24
Enhancements
Notes
This release adds --pin
flag to juv add
to have package specifiers resolved to an exact version at the time of the command, and subsequently pinned in the notebook/script.
uvx juv init
uvx juv add Untitled.ipynb 'numpy>=1.0.0' 'polars' # adds 'numpy>=1.0.0' 'polars'
uvx juv add Untitled.ipynb numpy polars --pin # adds 'numpy==2.1.3' 'polars==1.13.1'
This same behavior can be achieved without juv for regular scripts with a unix pipe:
echo 'numpy\npolars' | uv pip compile --no-deps - | grep '==' | xargs uv add --script foo.py
But alternatively you can use juv add
for the same thing:
uv init --script foo.py
uvx juv add foo.py numpy polars --pin
v0.2.23
Enhancements
- Add
juv stamp
for time-based dependency resolution pinning (#50)
Notes
uv
supports time-based dependency resolution via exclude-newer
, allowing packages to be resolved as they existed at a specific moment in time.
This feature greatly enhances the reproducibility of one-off scripts and notebooks without needing a lockfile. However, exclude-newer
requires a full RFC 3339 timestamp (e.g., 2020-03-05T00:00:00-05:00), which can be tedious to manage manually.
This release introduces juv stamp
, a command that provides a high-level, ergonomic API for pinning and unpinning various relevant timestamps in both standalone Python scripts and Jupyter notebooks:
# Stamp a notebook
juv init foo.ipynb
juv stamp foo.ipynb
# Stamp with a specific time
juv stamp foo.ipynb --time "2020-03-05T00:00:00-05:00"
juv stamp foo.ipynb --date 2022-01-03
# Use Git revisions
juv stamp foo.ipynb --rev e20c99
juv stamp foo.ipynb --latest
# Clear the pinned timestamp
juv stamp foo.ipynb --clear
# For Python scripts
uv init --script foo.py
uv add --script foo.py polars anywidget
uvx juv stamp foo.py