[udsy19]

Bug

compute_tke_spectrum(field, length=None) (physicsnemo/metrics/cae/cfd.py) only sets the domain lengths when length is omitted:

if length is None:
    lx, ly, lz = 2 * np.pi, 2 * np.pi, 2 * np.pi

so passing the documented length argument crashes a few lines later at kx = np.fft.fftfreq(nx, lx / nx):

compute_tke_spectrum(np.random.rand(3, 8, 8, 8), length=1.0)
# UnboundLocalError: cannot access local variable 'lx' where it is not associated with a value

The existing test only calls it with the default, so this went unnoticed.

Fix

Default length to 2 * np.pi and always set lx = ly = lz = length. Output without length is unchanged.

Tests

test_compute_tke_spectrum_length in test/metrics/test_metrics_cfd.py checks that length=2π reproduces the default output exactly and that wavenumbers scale with 1 / length (length=1.0 → 2π × the default). It fails on main with the UnboundLocalError and passes with the fix; test/metrics/test_metrics_cfd.py passes locally (CPU, macOS).

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• The CHANGELOG.md is up to date with these changes.
• An issue is linked to this pull request. (No existing issue; small self-contained fix.)

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

CODEOWNERS review map

Current for commit 3f16e360a5d6. An approval covers every file listed for that owner; one owner is sufficient for shared files.

⏳ @mnabian — 1 file(s)

• physicsnemo/metrics/cae/cfd.py

⏳ @peterdsharpe — 1 file(s)

• physicsnemo/metrics/cae/cfd.py

No CODEOWNER

• CHANGELOG.md
• test/metrics/test_metrics_cfd.py

Comment /codeowners-info to refresh.

[greptile-apps]

[Low impact] The change is limited to the optional domain-length path of one CFD metric.

The binned wavenumber axis should respect the supplied domain length before merging.

Findings

1.  Binned axis ignores domain size ▶

Summary

Fixes compute_tke_spectrum crashing when callers supply length.

• The spectrum uses the domain length passed by the caller.

_{Reviews (1) · Last reviewed commit: "Fix UnboundLocalError in compute_tke_spe..." · Reviewed by Greptile}

[greptile-apps]

Binned axis ignores domain size

Passing a different length now changes mk, but the returned wave_numbers still uses the fixed spacing 2 * np.pi. Doubling length halves the raw wavenumbers but leaves the binned spectrum’s axis unchanged. Users plotting wave_numbers against tke_spectrum therefore get an axis that ignores their domain length.

Make the binned axis scale with length, and check it in the new test.