Migration Guide⚓︎

Poetry → UV Migration (v6)⚓︎

Calcipy has migrated from Poetry to UV for package management and builds. This migration brings significant improvements in speed, simplicity, and modern Python packaging standards.

What Changed⚓︎

Build System:

✅ Migrated from poetry-core to uv_build
✅ Updated to modern PEP 621 [project] format
✅ Using [dependency-groups] instead of [tool.poetry.group]
✅ Lock file changed from poetry.lock to uv.lock

CI/CD:

✅ All workflows updated to use uv commands
✅ GitHub Actions use astral-sh/setup-uv@v5
✅ Pre-commit hooks configured for uv.lock

Task Automation:

✅ All tasks updated to use uv run, uv sync, uv build, uv publish
✅ Noxfile uses venv_backend='uv'

Migration Path for Users⚓︎

If you’re using Calcipy in your project and want to migrate from Poetry to UV:

1. Update Package Manager⚓︎

# Install uv
curl -LsSf https://astral.sh/uv/install.sh | sh

# Or with brew
brew install uv

2. Convert pyproject.toml⚓︎

Your pyproject.toml needs these changes:

Before (Poetry):

[build-system]
requires = ["poetry-core"]
build-backend = "poetry.core.masonry.api"

[tool.poetry]
name = "my-package"
version = "1.0.0"
dependencies = { python = "^3.10" }

[tool.poetry.group.dev.dependencies]
pytest = "^7.0"

After (UV):

[build-system]
requires = ["uv_build>=0.9.7"]
build-backend = "uv_build"

[project]
name = "my-package"
version = "1.0.0"
requires-python = ">=3.10"
dependencies = []

[dependency-groups]
dev = ["pytest>=7.0"]

3. Migrate Lock File⚓︎

# Remove old poetry files
rm poetry.lock

# Create new uv lock file
uv lock

# Install dependencies
uv sync --all-extras

4. Update CI/CD⚓︎

Replace poetry commands with uv equivalents:

Poetry	UV
`poetry install`	`uv sync --all-extras`
`poetry add package`	`uv add package`
`poetry run command`	`uv run command`
`poetry build`	`uv build`
`poetry publish`	`uv publish`

5. Update GitHub Actions⚓︎

# Before
  - uses: actions/setup-python@v4
    with:
      python-version: '3.11'
  - name: Install Poetry
    run: pipx install poetry
  - name: Install dependencies
    run: poetry install

# After
  - uses: astral-sh/setup-uv@v5
  - uses: actions/setup-python@v4
    with:
      python-version: '3.11'
  - name: Install dependencies
    run: uv sync --all-extras

Python Version Management: mise + uv + nox⚓︎

Calcipy uses a powerful combination of tools for managing Python versions:

mise (Version Manager)⚓︎

mise (formerly rtx) is a polyglot version manager that replaces asdf, pyenv, nvm, etc.

Setup mise:

# Install mise
curl https://mise.run | sh

# Or with brew
brew install mise

# Add to your shell (e.g., ~/.bashrc or ~/.zshrc)
echo 'eval "$(mise activate bash)"' >> ~/.bashrc

Configure Python versions:

Create mise.toml:

[tools]
python = ["3.12.5", "3.10.16"]

[env]
_.python.venv = { path = ".venv" }

Use mise:

# Install Python versions from mise.toml
mise install

# Verify installation
mise which python
python --version

Integration with nox⚓︎

Calcipy’s noxfile automatically reads Python versions from mise.toml:

# From src/calcipy/noxfile/_noxfile.py
def _get_pythons() -> List[str]:
    """Return python versions from supported configuration files."""
    return [*{str(ver) for ver in get_tool_versions()['python']}]


@nox_session(venv_backend='uv', python=_get_pythons(), reuse_venv=True)
def tests(session: NoxSession) -> None:
    """Run tests for all specified Python versions."""
    ...

This reads from:

mise.lock (resolved versions)
mise.toml (specified versions)
.tool-versions (legacy asdf format)

Run nox tests:

# List available sessions
uv run nox -l

# Run tests for all Python versions
uv run nox

# Run tests for specific version
uv run nox -s tests-3.12

Benefits of this Setup⚓︎

Version Consistency: Same Python versions for development, testing, and CI
Multi-version Testing: nox automatically tests against all specified versions
Fast Environment Creation: uv’s venv backend is significantly faster than virtualenv
Per-directory Versions: mise manages Python versions per project
No Global Pollution: Each project has isolated Python environments

Common Issues⚓︎

Problem: nox can’t find Python version

Solution: Ensure mise installed all versions:

mise install
mise which python3.9
mise which python3.12

Problem: Different versions between local and CI

Solution: Use mise.lock to pin exact versions:

mise lock
git add mise.lock

Tool vs Dependency Usage⚓︎

With v6, Calcipy now has clear modes:

Tool Mode (New in v6):

# Minimal installation for linting and code analysis
uv tool install 'calcipy[tool]'
calcipy-lint lint
calcipy-tags tags --base-dir=./my-project

Dependency Mode (Traditional):

# Full development environment
uv add --dev 'calcipy[dev]'
uv run calcipy test

See README for detailed usage examples.

`v5`⚓︎

The breaking changes include removing stale and pack.check_license

Speed Test⚓︎

After further reduction of dependencies, the CLI performance has continued to improve:

> poetry run pip freeze | wc -l
79

> hyperfine -m 20 --warmup 5 ./run
Benchmark 1: ./run
Time (mean ± σ):     397.1 ms ±  12.2 ms    [User: 268.4 ms, System: 57.0 ms]
Range (min … max):   385.9 ms … 421.5 ms    20 runs

The total number of dependencies was reduce even further by replacing flake8, isort, and other tooling with ruff; fewer mkdocs plugins; and fewer steps in the main run task to speed up normal usage.

The only breaking change impacted recipes when write_autoformatted_md_sections was renamed to write_template_formatted_md_sections.

Speed Test⚓︎

Following up on performance checks from the v2 migration. The performance is comparable, but you will see savings in cache size and poetry install and when running main (./run main for Calcipy, currently takes ~20s)

> hyperfine -m 20 --warmup 5 ./run
Benchmark 1: ./run
Time (mean ± σ):     863.9 ms ±  10.0 ms    [User: 550.7 ms, System: 102.3 ms]
Range (min … max):   848.5 ms … 885.3 ms    20 runs
> hyperfine -m 20 --warmup 5 "poetry run calcipy-tags"
Benchmark 1: poetry run calcipy-tags
Time (mean ± σ):     770.5 ms ±   5.7 ms    [User: 470.6 ms, System: 89.5 ms]
Range (min … max):   760.1 ms … 780.3 ms    20 runs

`v3`⚓︎

Replaced features from flake8 and plugins with corresponding checks from ruff, however both are still used in parallel.

`v2`⚓︎

Background⚓︎

calcipy v1 was a complete rewrite to switch from doit to invoke:

with invoke, tasks can be run from anywhere without a dodo.py file
tasks can be loaded lazily, which means that some performance gains are possible
since there is no shared state file, tasks can be more easily run from pre-commit or generally in parallel

doit excelled at clearly delineated task output and run summary, but invoke isn’t designed that way. I would like to improve the CLI output, but the benefits are worth this tradeoff.

calcipy v0 was built on doit and thus required a dodo.py file. I began adding cement to support a separate CLI for calcipy installed with pipx or uvx, but that required a lot of boilerplate code. With doit, the string command needed to be complete at task evaluation rather than runtime, so globbing files couldn’t be resolved lazily.

Migration⚓︎

While refactoring, the global configuration was mostly removed (DoitGlobals) along with a few tasks, but the main functionality is still present. Any project dependent on calcipy will need substantial changes. The easiest way to start migrating is to run copier copy gh:KyleKing/calcipy_template . for calcipy_template

Speed Test⚓︎

It turns out that switching to invoke appears to have only saved 100ms

> hyperfine -m 20 --warmup 5 ./run
Benchmark 1: ./run
Time (mean ± σ):     863.9 ms ±  10.0 ms    [User: 550.7 ms, System: 102.3 ms]
Range (min … max):   848.5 ms … 885.3 ms    20 runs
> hyperfine -m 20 --warmup 5 "poetry run calcipy-tags"
Benchmark 1: poetry run calcipy-tags
Time (mean ± σ):     770.5 ms ±   5.7 ms    [User: 470.6 ms, System: 89.5 ms]
Range (min … max):   760.1 ms … 780.3 ms    20 runs

> hyperfine -m 20 --warmup 5 "poetry run python -c 'print(1)'"
Benchmark 1: poetry run python -c 'print(1)'
Time (mean ± σ):     377.9 ms ±   3.1 ms    [User: 235.0 ms, System: 61.8 ms]
Range (min … max):   372.7 ms … 384.0 ms    20 runs
> hyperfine -m 20 --warmup 5 ./run
Benchmark 1: ./run
Time (mean ± σ):     936.0 ms ±  26.9 ms    [User: 1548.2 ms, System: 1687.7 ms]
Range (min … max):   896.4 ms … 1009.4 ms    20 runs
> hyperfine -m 20 --warmup 5 "poetry run calcipy_tags"
Benchmark 1: poetry run calcipy_tags
Time (mean ± σ):     618.5 ms ±  29.7 ms    [User: 1536.8 ms, System: 1066.2 ms]
Range (min … max):   578.2 ms … 694.9 ms    20 runs
> hyperfine -m 20 --warmup 5 "poetry run doit list"
Benchmark 1: poetry run doit list
Time (mean ± σ):      1.002 s ±  0.015 s    [User: 1.643 s, System: 1.682 s]
Range (min … max):    0.974 s …  1.023 s    20 runs

Additionally, the major decrease in dependencies will make install and update actions much faster. With the recommended extras installed, calcipy-v1 has 124 dependencies (with all extras, 164) vs. calcipy-v0’s 259. Counted with: cat .calcipy_packaging.lock | jq 'keys' | wc -l

Code Comparison⚓︎

Accounting for code extracted to corallium, the overall number of lines decreased from 1772 to 1550 or only 12%, while increasing the CLI and pre-commit capabilities.

~/calcipy-v0 > cloc calcipy
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Python                          26            942           1075           1772
-------------------------------------------------------------------------------
SUM:                            26            942           1075           1772
-------------------------------------------------------------------------------
~/calcipy > cloc calcipy
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Python                          27            454            438           1185
-------------------------------------------------------------------------------
SUM:                            27            454            438           1185
-------------------------------------------------------------------------------
~/corallium > cloc corallium
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Python                           7            176            149            365
-------------------------------------------------------------------------------
SUM:                             7            176            149            365
-------------------------------------------------------------------------------

~/calcipy > cloc tests
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
YAML                             2              0              0            580
Python                          19            176             68            578
JSON                             2              0              0             60
Markdown                         3              9             10              8
Text                             1              0              0              2
-------------------------------------------------------------------------------
SUM:                            27            185             78           1228
-------------------------------------------------------------------------------
~/calcipy-v0 > cloc tests
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
JSON                            30              0              0            762
YAML                             2              0              0            580
Python                          24            314            186            578
Markdown                         3              9             10              8
-------------------------------------------------------------------------------
SUM:                            59            323            196           1928
-------------------------------------------------------------------------------
~/corallium > cloc tests
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Python                           6             36             15             69
Markdown                         1              1              0              2
-------------------------------------------------------------------------------
SUM:                             7             37             15             71
-------------------------------------------------------------------------------

doit output⚓︎

I would like to restore the doit task summary, but invoke’s architecture doesn’t really make this possible. The --continue option was extremely useful, but that also might not be achievable.

> poetry run doit run
.  format_recipes > [
        Python: function format_recipes
]

2023-02-19 10:40:23.954 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/breakfast
2023-02-19 10:40:23.957 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/rice
2023-02-19 10:40:23.959 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/meals
2023-02-19 10:40:23.964 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/seafood
2023-02-19 10:40:23.967 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/pizza
2023-02-19 10:40:23.969 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/poultry
2023-02-19 10:40:23.972 | INFO     | recipes.formatter:_write_toc:287 - Creating TOC for: ./recipes/docs/sushi
.  collect_code_tags > [
        Python: function write_code_tag_file
]

.  cl_write > [
        Cmd: poetry run cz changelog
        Python: function _move_cl
]

.  lock > [
        Cmd: poetry lock --no-update
]

Resolving dependencies...
.  nox_coverage > [
        Cmd: poetry run nox --error-on-missing-interpreters --session coverage
]

...

doit> Summary:
doit> format_recipes was successful
doit> collect_code_tags was successful
doit> cl_write was successful
doit> lock was successful
doit> nox_coverage was successful
doit> auto_format was successful
doit> document was successful
doit> check_for_stale_packages was successful
doit> pre_commit_hooks failed (red)
doit> lint_project was not run
doit> static_checks was not run
doit> security_checks was not run
doit> check_types was not run

Migration Guide⚓︎

Poetry → UV Migration (v6)⚓︎

What Changed⚓︎

Migration Path for Users⚓︎

1. Update Package Manager⚓︎

2. Convert pyproject.toml⚓︎

3. Migrate Lock File⚓︎

4. Update CI/CD⚓︎

5. Update GitHub Actions⚓︎

Python Version Management: mise + uv + nox⚓︎

mise (Version Manager)⚓︎

Integration with nox⚓︎

Benefits of this Setup⚓︎

Common Issues⚓︎

Tool vs Dependency Usage⚓︎

v5⚓︎

Speed Test⚓︎

v4⚓︎

Speed Test⚓︎

v3⚓︎

v2⚓︎

Background⚓︎

Migration⚓︎

Speed Test⚓︎

Code Comparison⚓︎

doit output⚓︎

`v5`⚓︎

`v4`⚓︎

`v3`⚓︎

`v2`⚓︎