fix: align reconstructed coordinate order with input (+ comparison recipe)

[FBumann]

Follow-up to #93 (now merged): fixes an incomplete part of the dim-order work and adds the comparison-plot recipe we opted for over a first-class plotting method.

Fix — coordinate order (not just dim order)

#93 transposed reconstructed to the input's dim order, but tsam still returns it with alphabetically sorted columns, so the variable (etc.) coordinate order diverged from original:

agg.original["variable"].values        # ['wind', 'solar', 'demand']  (input order)
agg.reconstructed["variable"].values   # ['demand', 'solar', 'wind']  (tsam sorted)

Consequences: original.to_dataframe() and reconstructed.to_dataframe() had different row orders, and compare()'s concat emitted a FutureWarning (join='outer' → 'exact') that will become an error in a future xarray.

reindex_like(da) after the transpose realigns coordinate order too. The column set is identical, so it only reorders — no NaNs. original, reconstructed, and residuals now line up exactly in values and row order. Regression test added (test_non_alphabetical_coord_order_preserved).

Docs — comparison recipe (resolves #95 without a plotting method)

Per discussion on #95/#96, we decided not to add a plot_compare() method (keeps the library lightweight; compare()/to_dataframe() already return plot-ready data). Instead, the getting-started notebook now shows a load-duration-curve comparison built directly from the tidy to_dataframe() output (Original dotted / Reconstructed solid) — the one thing the proposed method offered beyond the existing compare().plotly.line(...) example.

Validation

529 passed / 4 skipped; mypy + ruff clean; notebook executes end-to-end. Closes #95.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Preserved input coordinate/dimension order in reconstructed results to prevent mismatches.
  • Improved original/reconstructed alignment so missing values are avoided.
• Documentation
  • Added a getting-started example demonstrating result.to_dataframe() for comparison workflows.
  • Included an example for generating and visualizing load-duration style curve comparisons.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 687d898d-a2a2-4ea7-897b-2dc068ff4e82

📥 Commits

Reviewing files that changed from the base of the PR and between 678869d and 44a2523.

📒 Files selected for processing (4)

• docs/examples/getting-started.ipynb
• src/tsam_xarray/_core.py
• src/tsam_xarray/_result.py
• test/test_aggregate.py

💤 Files with no reviewable changes (1)

• src/tsam_xarray/_result.py

🚧 Files skipped from review as they are similar to previous changes (2)

• docs/examples/getting-started.ipynb
• test/test_aggregate.py

---

📝 Walkthrough

Walkthrough

Aggregation output now aligns reconstructed coordinates with the input order and labels, suppresses a targeted upstream warning, and adds regression coverage. The getting-started notebook demonstrates dataframe-based original-versus-reconstructed load-duration curves.

Changes

Aggregation alignment and comparison

Layer / File(s)	Summary
Realign reconstructed output src/tsam_xarray/_core.py, test/test_aggregate.py	reconstructed is transposed and reindexed to match input dimensions and coordinate labels; regression coverage verifies non-alphabetical coordinate preservation without NaNs.
Filter aggregation warning src/tsam_xarray/_core.py	Aggregation calls run inside a warning filter targeting the specified FutureWarning, while existing aggregation branches remain unchanged.
Document load-duration comparison docs/examples/getting-started.ipynb, src/tsam_xarray/_result.py	The notebook ranks tidy dataframe values per variant and variable and plots original-versus-reconstructed load-duration curves; dataframe behavior is unchanged apart from comment cleanup.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• FBumann/tsam_xarray#93: Both changes address reconstructed coordinate alignment and dataframe-based comparison workflows.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	Issue #95 asked for a built-in compare plot on AggregationResult, but this PR only adds a notebook recipe and no plot_compare helper.	Add AggregationResult.plot_compare() with timeseries/duration_curve modes, variable selection, faceting, and a clear plotly fallback error.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title matches the main change: fixing coordinate order alignment, while the comparison recipe is a secondary part.
Out of Scope Changes check	✅ Passed	The warning suppression, tests, and notebook recipe all support the stated fix and comparison documentation; nothing clearly unrelated stands out.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/reconstructed-coord-align

---

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.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@docs/examples/getting-started.ipynb`:
- Around line 172-188: Update the load-duration curve plotting block after the
`tidy` transformation to use the notebook’s `xarray_plotly` plotting accessor
instead of `plotly.express`, while preserving the existing rank, energy,
variant, variable, line-style, and title configuration.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 50531d44-1c3e-440c-837f-82636e342aea

📥 Commits

Reviewing files that changed from the base of the PR and between b3fd174 and 678869d.

📒 Files selected for processing (3)

• docs/examples/getting-started.ipynb
• src/tsam_xarray/_core.py
• test/test_aggregate.py