Skip to content

sync#890

Closed
henrydingliu wants to merge 29 commits into
experimentalfrom
main
Closed

sync#890
henrydingliu wants to merge 29 commits into
experimentalfrom
main

Conversation

@henrydingliu
Copy link
Copy Markdown
Collaborator

@henrydingliu henrydingliu commented May 31, 2026

Summary of Changes

Related GitHub Issue(s)

Additional Context for Reviewers

  • I passed tests locally for both code (uv run pytest) and documentation changes (uv run jb build docs --builder=custom --custom-builder=doctest)

Note

Medium Risk
Datetime and global Options behavior affect valuation dates and test fixtures across the core package; removing deprecated option helpers may break downstream callers that still use them.

Overview
Package configuration refactors chainladder’s public init: Pandas 2/3 datetime compatibility moves from options.DT64_* to module-level __dt64_dtype__ / __dt64_unit__, with call sites updated in base, pandas, and utility_functions. Options becomes an instance class with validated get_option / set_option, per-option or full reset_option, and stored defaults; deprecated top-level array_backend / auto_sparse helpers are removed in favor of chainladder.options.

Tests gain shared _sample_fixture in conftest.py (backend setup/teardown), expanded options coverage, and datetime tests keyed off __dt64_unit__.

Documentation adds contributor docs/README.md, a Friedland book section (RST doctests for chapters 6–7, WIP chapter 8 notebook, TOC entry), and large numpydoc expansions with runnable examples on Development, BootstrapODPSample, and Mack correlation classes. Minor doc build tweaks (pyproject.toml trailing comma, Development autosummary stub).

Reviewed by Cursor Bugbot for commit 449b5c1. Bugbot is set up for automated code reviews on this repo. Configure here.

kennethshsu and others added 29 commits May 22, 2026 14:24
Adds Sphinx doctest Examples sections to DevelopmentCorrelation and ValuationCorrelation. Each example opens with the Mack chain-ladder assumption being tested, prints the full decision signal (statistic, confidence band, and boolean) rather than a single boolean, and ties the result back to the chain-ladder workflow.

Refs #704
Adds Sphinx doctest Examples section to the BootstrapODPSample class showing basic fit (resampled_triangles_.shape and scale_), downstream stochastic IBNR via Chainladder, and the effect of drop_high on scale_. Uses random_state=42 and n_sims=100 for deterministic, fast output.

Refs #704
Rewords the paragraph introducing the drop_high=True example to describe it as a leave-one-out sensitivity check on the column maxima rather than outlier removal, since drop_high mechanically removes the column max without any outlier test.

Addresses review feedback on #836.
Per @henrydingliu review on #844: opening paragraph now scoped to the calendar-effect concept; the per-diagonal vs whole-triangle distinction is introduced as a transition between the two testcode blocks.

Refs #704
- Drop the jb build internals + duplicate warnings note (kennethshsu)
- Rename 'tutorial notebooks' to 'onboarding and Quickstart notebooks' (kennethshsu)
- Drop the 'What to edit for which part of the site' table and 'Known issues' section as redundant with the Source files table and #841 (kennethshsu)
docs: add README documenting docs build sources and outputs
Improved docstring and added examples for Development class
[REFACTOR]: Remove repetitive code. Remove dead Python 3.8 code. Add …
…option getters and setters. Deprecate cl.array_backend() and cl.auto_sparse()
first batch of deliverables for friedland reconstruction
[REFACTOR]: Continue work on cl.__init__.py refactor
@henrydingliu henrydingliu requested a review from jbogaardt as a code owner May 31, 2026 07:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants