chore: bump jaxtyping to 0.3.4 and drop numpy workaround

[charlesbmi]

Summary

• Bump jaxtyping from >=0.2.36 to >=0.3.4
• Remove the explicit numpy>=1.25.0 core dependency that was only needed as a workaround for jaxtyping#360

jaxtyping v0.3.4 includes the fix from patrick-kidger/jaxtyping#361, so mach can import without numpy installed in the base environment. Numpy remains available via optional extras (e.g. uff).

Test plan

• CI passes
• uvx --with mach --python 3.12 python -c "import mach" works without numpy in the base install

Made with Cursor

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (1) 📘 Rule violations (0) 📎 Requirement gaps (0) 🎨 UX issues (0) 🔗 Cross-repo conflicts (0)

1. Lockfile dependency drift ✓ Resolved 🐞 Bug ☼ Reliability

Description

pyproject.toml now requires jaxtyping>=0.3.4 and drops core numpy, but uv.lock still records
the old requirements, so uv sync --frozen / uv lock --locked will fail or install the previous
dependency set.

Code

pyproject.toml[R31-35]

dependencies = [
    "array-api-compat>=1.11.0",
    "einops>=0.8.0",
-    "jaxtyping>=0.2.36",
-    # TODO: remove numpy dependency once jaxtyping no longer implicitly depends on numpy
-    # https://github.com/patrick-kidger/jaxtyping/issues/360
-    "numpy>=1.25.0",
+    "jaxtyping>=0.3.4",
]

Evidence

The repository explicitly uses the lockfile for dependency installation and validation; the current
uv.lock still encodes the old base requirements (including core numpy and older jaxtyping
constraint/version), which contradicts the new pyproject.toml requirements.

pyproject.toml[31-35]
uv.lock[1231-1241]
uv.lock[1671-1684]
.github/actions/setup-python-env/action.yml[17-28]
Makefile[38-60]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`pyproject.toml` was updated (bumped `jaxtyping`, removed core `numpy`), but `uv.lock` was not regenerated. This breaks frozen installs (`uv sync --frozen`) and locked checks (`uv lock --locked`), and can cause CI/dev to keep using the old dependency set.

### Issue Context
- CI uses `uv sync --frozen`, which requires `uv.lock` to match `pyproject.toml`.
- The repo Makefile enforces lock consistency via `uv lock --locked`.

### Fix Focus Areas
- uv.lock[1231-1241]
- uv.lock[1671-1684]
- pyproject.toml[31-35]
- .github/actions/setup-python-env/action.yml[17-28]
- Makefile[38-60]

### What to change
1. Run `uv lock` (or `uv lock --upgrade-package jaxtyping` if preferred) to regenerate `uv.lock`.
2. Ensure the regenerated lock:
  - reflects `jaxtyping>=0.3.4`
  - no longer lists `numpy` as an unconditional base requirement
3. Commit the updated `uv.lock`.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. Vis extra missing numpy ✓ Resolved 🐞 Bug ≡ Correctness

Description

After dropping core numpy, mach._vis fails to import because it unconditionally imports numpy,
but the vis optional dependency set does not include numpy (so pip install mach-beamform[vis]
is insufficient).

Code

pyproject.toml[R31-35]

dependencies = [
    "array-api-compat>=1.11.0",
    "einops>=0.8.0",
-    "jaxtyping>=0.2.36",
-    # TODO: remove numpy dependency once jaxtyping no longer implicitly depends on numpy
-    # https://github.com/patrick-kidger/jaxtyping/issues/360
-    "numpy>=1.25.0",
+    "jaxtyping>=0.3.4",
]

Evidence

The vis extra currently installs matplotlib/colorcet only, while mach._vis imports numpy
unconditionally; removing numpy from core dependencies makes mach-beamform[vis] insufficient to
import mach._vis.

pyproject.toml[31-35]
pyproject.toml[40-49]
src/mach/_vis.py[5-12]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`src/mach/_vis.py` imports `numpy` at module import time, but after removing `numpy` from core dependencies, installing `mach-beamform[vis]` does not install `numpy`. This causes `import mach._vis` (and anything that imports it) to fail with `ModuleNotFoundError: numpy`.

### Issue Context
`mach._vis` already provides a helpful install hint for missing `matplotlib`, but missing `numpy` will now be the failure mode in a base environment without numpy.

### Fix Focus Areas
- pyproject.toml[40-49]
- src/mach/_vis.py[5-12]
- pyproject.toml[31-35]

### What to change
- Add `numpy>=1.25.0` (or an appropriate minimum) to `[project.optional-dependencies].vis`.
- (Optional hardening) Wrap the `import numpy as np` in a `try/except ImportError` and raise a targeted message similar to the existing matplotlib guidance.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

3. Examples extra missing numpy 🐞 Bug ≡ Correctness

Description

After removing core numpy, importing mach.io.must (and mach.io.uff) will fail with
ModuleNotFoundError for numpy/scipy unless users also install the uff extra. The
project.optional-dependencies.examples extra does not include numpy/scipy (or
mach-beamform[uff]), so pip install mach-beamform[examples] is no longer sufficient for example
I/O usage.

Code

pyproject.toml[R31-35]

dependencies = [
    "array-api-compat>=1.11.0",
    "einops>=0.8.0",
-    "jaxtyping>=0.2.36",
-    # TODO: remove numpy dependency once jaxtyping no longer implicitly depends on numpy
-    # https://github.com/patrick-kidger/jaxtyping/issues/360
-    "numpy>=1.25.0",
+    "jaxtyping>=0.3.4",
]

Evidence

The PR removes core numpy, but mach.io.must/mach.io.uff still import numpy/scipy
unconditionally; meanwhile the examples extra does not include numpy/scipy whereas the uff
extra does, so installing only [examples] can no longer run example I/O code.

pyproject.toml[31-60]
src/mach/io/must.py[1-8]
src/mach/io/uff.py[8-13]
examples/doppler.py[48-61]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
Core `numpy` was removed, but some example-facing modules (`mach.io.must`, `mach.io.uff`) still import `numpy`/`scipy` at module import time. As a result, users who install `mach-beamform[examples]` can still hit `ModuleNotFoundError` when running examples or using example I/O helpers.

## Issue Context
- `pyproject.toml` core deps no longer include `numpy`.
- `mach.io.must` imports `numpy` and `scipy.io` at import time.
- `mach.io.uff` imports `numpy` and `scipy.signal` at import time.
- The `examples` extra does not install `numpy`/`scipy` (while `uff` does).

## Fix Focus Areas
- pyproject.toml[31-60]
- src/mach/io/must.py[1-10]
- src/mach/io/uff.py[8-15]

### Suggested implementation direction
1) Update `[project.optional-dependencies].examples` to include either:
  - `mach-beamform[uff]` (preferred for consistency), or
  - explicit `numpy` + `scipy` requirements.
2) (Optional but recommended) Add `try/except ImportError` guards around `numpy`/`scipy` imports in `mach/io/must.py` and `mach/io/uff.py`, raising an error that clearly instructs `pip install mach-beamform[uff]` (or `[all]`).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

[qodo-code-review]

2. Vis extra missing numpy 🐞 Bug ≡ Correctness

After dropping core numpy, mach._vis fails to import because it unconditionally imports numpy,
but the vis optional dependency set does not include numpy (so pip install mach-beamform[vis]
is insufficient).

Agent Prompt

### Issue description
`src/mach/_vis.py` imports `numpy` at module import time, but after removing `numpy` from core dependencies, installing `mach-beamform[vis]` does not install `numpy`. This causes `import mach._vis` (and anything that imports it) to fail with `ModuleNotFoundError: numpy`.

### Issue Context
`mach._vis` already provides a helpful install hint for missing `matplotlib`, but missing `numpy` will now be the failure mode in a base environment without numpy.

### Fix Focus Areas
- pyproject.toml[40-49]
- src/mach/_vis.py[5-12]
- pyproject.toml[31-35]

### What to change
- Add `numpy>=1.25.0` (or an appropriate minimum) to `[project.optional-dependencies].vis`.
- (Optional hardening) Wrap the `import numpy as np` in a `try/except ImportError` and raise a targeted message similar to the existing matplotlib guidance.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

[qodo-code-review]

Code review by qodo was updated up to the latest commit 00621ee