[benjaminpope]

Bumps the version to 1.8.0. Once this merges, pushing tag v1.8.0 publishes to PyPI. Merge #335 (urllib3, which clears the 3 Dependabot alerts) first.

Draft release notes for v1.8.0

Performance

Reduction is substantially faster. All of the changes are pure numpy and add no new dependencies. Outputs agree with 1.7.0 to within floating-point noise: at most 1.3e-14 of each array's largest value, checked by a new test against reference outputs frozen from the previous code.

• Closure-phase covariance (Perf 1/5: vectorise covariance helpers; frozen reference-output test #339): vectorized, along with the other covariance and error helpers and cov2cor. With a 17-hole mask and 200 frames it drops from 2.8 s to 24 ms.
• compute_index_mask (Perf 2/5: vectorise compute_index_mask; build it once per extraction #340): vectorized and built once per extraction instead of twice. With 23 holes it drops from 1.5 s to 7 ms.
• Fourier-plane sampling (Perf 3/5: vectorise _compute_complex_bs; batch phase-slope regressions #341): vectorized, and the phase-slope regressions are batched. extract_bs on the 316-frame NIRISS example drops from 1.08 s to 0.18 s; on a 17-hole, 200-frame cube it drops from 8.2 s to 1.0 s.
• Multiple-triangle bispectrum (Perf 4/5: precompute the multiple-triangle index tensor #342): index tensors are precomputed. extract_bs(bs_multi_tri=True) on 316 frames drops from 2.2 s to 0.31 s.
• clean_data (Speed up clean_data: exact find_max, local bad-pixel interpolation, cached window #343): an exact fast find_max, bad-pixel interpolation done locally at the bad pixels (with a fallback to astropy) and a cached window map make it about 1.4 times faster.

Security

• urllib3 is updated to 2.8.0 in the lockfile (Bump urllib3 from 2.7.0 to 2.8.0 #335).

Known minor issues found during this work are tracked in #338.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.