Repository navigation
Normalize scattering CDFs before inverse angle sampling - #34
Merged
scottprahl merged 2 commits intoOct 28, 2026
Merged
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Follow-ups from review of the CDF normalization fix: - warn when cdf's raw trapezoidal integral is off from 1 by more than 1%, since normalizing silently hid an under-resolved forward peak - size mu_with_uniform_cdf's internal table as max(2000, 5*x**2) so large spheres sample correctly (x=100 mean cosine 0.930 -> 0.869, g=0.868) - replace the inversion loop with np.interp, removing the big_k - 1 index - raise ValueError when the phase function is not finite - point the warning at the caller via a private _trapezoid_cdf helper - rewrite the module, cdf and mu_with_uniform_cdf docstrings, with doctested sampling examples - restore a convergence test, add warning, NaN and asymmetry tests, and drop a redundant arccos check - changelog entries, crediting @antonsoo for the original fix Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Owner
|
thanks for your contribution. I added a bit to your PR since this is the first time I have looked at the CDF code in a while. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The phase sum starts above zero and overshoots one. For
mu_with_uniform_cdf(1.33, .001, 4096), low-probability interpolation crosses between the first and last entries: the table contains cosines below -1, giving NaNs inarccos.Use cumulative trapezoidal integration from zero, normalize its endpoint to one, and require two or more points. Add analytic Rayleigh/physical-angle regressions and replace the tests that pinned the old endpoint error. The fixed 2000-point integration grid and random sampler are unchanged.
Both backend suites pass on Python 3.12 and 3.14; configured lint and combined coverage pass. Droplet sampling agrees with independent native Scattnlay quadrature, preserving the original optical outputs and draws from a fixed supplied table. The installed wheel gives identical results. Other platforms and notebooks were not run.