Optimize parsing in primitive.py

[chenzhaiyu]

Performance difference

Benchmarks were run on the provided .vg fixtures using default VertexGroup processing, reporting mean runtime over 15 runs.

Dataset	Before	After	Speedup
test_points.vg	0.013251s	0.011740s	11.41%
test_church.vg	0.043811s	0.030116s	31.26%

Validation data

Behavioral equivalence (before vs after) was validated on all provided fixtures:

• tutorials/data/test_points.vg
• tutorials/data/test_church.vg
• tutorials/data/test_points.bvg
• tutorials/data/test_church.bvg

Closes #26

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 81efa3e68e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Center points before projection to avoid cancellation

Computing points_trans as points @ eig_vec.T - (pca.mean_ @ eig_vec.T) is algebraically equivalent to pca.transform(points), but it is numerically less stable because it subtracts two potentially large projected values; with large absolute coordinates and small local spread (common in geospatial scans), this can lose precision and distort point_min/point_max, producing inaccurate OBBs. Using (points - pca.mean_) @ eig_vec.T keeps the optimization while preserving the stable centering order.

Useful? React with 👍 / 👎.

[Copilot]

Pull request overview

This PR focuses on improving runtime performance in abspy/primitive.py by reducing Python-level overhead during vertex-group parsing and plane fitting, aligning with the broader goal of speeding up VertexGroup processing.

Changes:

• Optimizes point extraction in get_points() by slicing once instead of accumulating lines.
• Optimizes primitive/parameter line discovery and parameter parsing in get_primitives().
• Speeds up PCA-based OBB computation in fit_plane() by avoiding pca.transform() input validation overhead.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

[Copilot]

Parsing group parameters via parameters[i][18:-1] drops the final character. For .bvg inputs (where lines come from split('\n')) and for any vgroup_ascii without a trailing newline, this will silently truncate the last parameter (e.g., 0.12 -> 0.1), yielding incorrect plane parameters. Prefer extracting the substring after the group_parameters: prefix and stripping whitespace/newlines instead of fixed slicing.

[Copilot]

Same issue as the empty-group branch: parameters[i][18:-1] truncates the last character of the last numeric token when the line does not end with a newline (notably .bvg parsing). This can change d (and any other last value) and lead to incorrect planes/bounds. Use prefix-based extraction plus .strip() rather than positional slicing.

[Copilot]

fit_plane now computes points_trans with points @ eig_vec.T ..., which requires points to be a NumPy array supporting @. Previously pca.transform(points) accepted any array-like input (lists, tuples) via scikit-learn validation. Since fit_plane is a documented public API, consider converting points = np.asarray(points) (or falling back to pca.transform) to avoid an unintended API regression.