perf: compute accuracy and reconstructed on first access

[FBumann]

First of the three perf drafts (this one is the wrapper-side lever unlocked by tsam v4's lazy metrics).

What

AggregationResult.accuracy and .reconstructed are now populated on first access via cached_property over a stored per-field factory, instead of being materialized during result conversion. All four construction paths (single, slice-concat, slice view, clustering.apply) defer consistently — for the sliced case even the per-slice materialization and concat are deferred. repr() no longer forces the computation (shows accuracy=<not computed> until first read).

Why / measured impact

tsam v4 computes metrics lazily, but the wrapper immediately materialized them — paying v4's deferred cost on every aggregate() call even when nobody reads the metrics. Measured at 730 days × 256 columns, Distribution(scope="global") + extremes replace, metrics unread:

	main	this PR
tsam develop-v4	2.59 s	1.83 s (−29%)
tsam 3.4.2	2.52 s	2.52 s (defers only the xarray conversion — v3 computes metrics inside tsam.aggregate() regardless)

Reading result.accuracy/.reconstructed (or residuals/compare()/to_dataframe(), which touch them) computes once and caches — behavior and values are unchanged, verified by the full suite plus new laziness tests (test/test_lazy_result.py).

Notes

• The two fields moved from eager dataclass fields to kw-only private factories + cached_property; AggregationResult construction is internal-only, so no public API changes.
• CI benchmark note: the head-vs-base comment runs on tsam v3.4.x, where the win is only the deferred conversion (visible in test_wrapper_result_conversion, not e2e).

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 18aae05f-33eb-4e9d-8939-f7a4e58b7b7f

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/lazy-result

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Benchmarks

Δ% vs base, one table sorted by biggest change first
(min walltime and memray peak). Peak memory is
deterministic — any delta there is real; walltime on GitHub
runners is noisy, so only large time deltas are meaningful.
The interactive plot is attached as the benchmark-plot artifact.

Full table

benchmarks/test_bench_aggregate.py

name	metric	base	head
test_user_disaggregate	time	0.01089s	+2.0%
	peak	26.7 MiB	0.0%
test_wrapper_to_dataframe	time	0.0003656s	-1.1%
	peak	8.55 MiB	0.0%
test_e2e_wide[dist_global]	time	1.782s	-2.1%
	peak	104 MiB	-41.4%
test_e2e_default	time	0.06297s	-2.9%
	peak	7.11 MiB	-37.9%
test_e2e_wide[medoid]	time	0.4844s	-7.2%
	peak	113 MiB	-38.4%
test_e2e_full	time	0.1436s	-8.1%
	peak	6.6 MiB	-40.9%
test_config_representation[dist_global_minmax]	time	0.1223s	-8.3%
	peak	1.78 MiB	-42.8%
test_config_representation[dist_global]	time	0.1202s	-8.5%
	peak	1.78 MiB	-42.8%
test_e2e_slices	time	1.168s	-8.9%
	peak	14.3 MiB	+91.0%
test_user_apply	time	0.3882s	-11.1%
	peak	104 MiB	-41.4%
test_e2e_multidim	time	0.1989s	-14.4%
	peak	26.7 MiB	-34.6%
test_config_extremes[append]	time	0.04859s	-22.3%
	peak	1.9 MiB	-40.1%
test_config_extremes[replace]	time	0.04891s	-22.7%
	peak	1.9 MiB	-40.2%
test_config_representation[dist_cluster]	time	0.04627s	-23.1%
	peak	1.78 MiB	-42.8%
test_config_representation[medoid]	time	0.04707s	-23.1%
	peak	1.9 MiB	-40.3%
test_config_representation[mean]	time	0.04498s	-24.3%
	peak	1.78 MiB	-42.8%
test_wrapper_concat_results	time	0.01898s	-52.7%
	peak	8.72 MiB	-49.1%
test_wrapper_result_conversion	time	0.03495s	-93.6%
	peak	60.3 MiB	-97.1%