sweeps the whole labs tree to fix a marimo dataflow bug class that left widgets invisible to the tabs cell in 17 of 33 labs. reported by @asgalon in #1332 for lab_02 + lab_03 (fixed in #1385); this pr does the rest and locks the regression out via a new static test.
the bug
marimo routes widget state strictly through each cell's return tuple. every @app.cell that defined helper widgets (sliders, dropdowns) alongside the terminal partX_prediction returned only the prediction. the helpers were defined but never reached the tabs cell — they were effectively dead allocations. sliders never rendered after the prediction unlocked; chart inputs never re-ran on change. every static and engine test still passed because the cells executed fine in isolation.
the audit script (/tmp/audit_widget_returns.py) found:
scope
before
after
labs scanned
33
33
offending labs
17
0
unreturned widgets
175
0
(the 42 "hits" that remain in the updated audit are all widgets defined inside the tabs cell itself, consumed by its own build_part_* closures via scope — they don't need to flow outward. the new static test excludes those.)
changes
codemod over all 33 labs. every widget cell's return (...) tuple now lists every mo.ui.* it defines, alphabetical. every tabs-cell signature is rewritten to declare every widget it references. no logic, labels, or UI copy touched.
TestWidgetReturnCompleteness in labs/tests/test_static.py. fails on any new cell defining a widget without returning it (render sinks like tabs/_tabs and tabs-cell-local widgets excluded). 33 / 33 labs pass; a regression is blocked at CI time.
test plan
labs/tests/test_static.py + tests/test_engine.py: 825 passed 4 skipped 1 xfailed (up from 792 — 33 new test runs)
marimo check spot-sampled on several labs: exit 0
audit script: 17 → 0 offending labs on the meaningful check
CI green on this pr
closes the main thread of #1332. @asgalon's "updated example values" suggestion for Lab 02 and the tabs-cell widget extraction cleanup left as follow-ups.
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.
## 📋 Pull Request Information
**Original PR:** https://github.com/harvard-edge/cs249r_book/pull/1386
**Author:** [@profvjreddi](https://github.com/profvjreddi)
**Created:** 4/17/2026
**Status:** ✅ Merged
**Merged:** 4/17/2026
**Merged by:** [@profvjreddi](https://github.com/profvjreddi)
**Base:** `dev` ← **Head:** `fix/labs-widget-return-audit`
---
### 📝 Commits (1)
- [`80b916f`](https://github.com/harvard-edge/cs249r_book/commit/80b916f3d6aa62dc7e791bccf4dcbc2958754b71) fix(labs): widget return-tuple sweep + static test for regressions (#1332)
### 📊 Changes
**34 files changed** (+419 additions, -326 deletions)
<details>
<summary>View changed files</summary>
📝 `labs/tests/test_static.py` (+82 -0)
📝 `labs/vol1/lab_00_introduction.py` (+5 -2)
📝 `labs/vol1/lab_01_ml_intro.py` (+11 -15)
📝 `labs/vol1/lab_02_ml_systems.py` (+13 -17)
📝 `labs/vol1/lab_03_ml_workflow.py` (+6 -8)
📝 `labs/vol1/lab_04_data_engr.py` (+8 -4)
📝 `labs/vol1/lab_05_nn_compute.py` (+9 -4)
📝 `labs/vol1/lab_06_nn_arch.py` (+8 -4)
📝 `labs/vol1/lab_07_ml_frameworks.py` (+8 -4)
📝 `labs/vol1/lab_08_model_train.py` (+9 -4)
📝 `labs/vol1/lab_09_data_selection.py` (+9 -4)
📝 `labs/vol1/lab_10_model_compress.py` (+8 -4)
📝 `labs/vol1/lab_11_hw_accel.py` (+8 -4)
📝 `labs/vol1/lab_12_perf_bench.py` (+8 -4)
📝 `labs/vol1/lab_13_model_serving.py` (+9 -4)
📝 `labs/vol1/lab_14_ml_ops.py` (+8 -4)
📝 `labs/vol1/lab_15_responsible_engr.py` (+8 -4)
📝 `labs/vol1/lab_16_ml_conclusion.py` (+10 -4)
📝 `labs/vol2/lab_01_introduction.py` (+11 -5)
📝 `labs/vol2/lab_02_compute_infra.py` (+10 -5)
_...and 14 more files_
</details>
### 📄 Description
## what
sweeps the whole labs tree to fix a marimo dataflow bug class that left widgets invisible to the tabs cell in 17 of 33 labs. reported by @asgalon in #1332 for lab_02 + lab_03 (fixed in #1385); this pr does the rest and locks the regression out via a new static test.
## the bug
marimo routes widget state strictly through each cell's return tuple. every `@app.cell` that defined helper widgets (sliders, dropdowns) alongside the terminal `partX_prediction` returned only the prediction. the helpers were defined but never reached the tabs cell — they were effectively dead allocations. sliders never rendered after the prediction unlocked; chart inputs never re-ran on change. every static and engine test still passed because the cells executed fine in isolation.
the audit script (`/tmp/audit_widget_returns.py`) found:
| scope | before | after |
|-|-|-|
| labs scanned | 33 | 33 |
| offending labs | 17 | 0 |
| unreturned widgets | 175 | 0 |
(the 42 "hits" that remain in the updated audit are all widgets defined inside the tabs cell itself, consumed by its own `build_part_*` closures via scope — they don't need to flow outward. the new static test excludes those.)
## changes
1. **codemod over all 33 labs.** every widget cell's `return (...)` tuple now lists every `mo.ui.*` it defines, alphabetical. every tabs-cell signature is rewritten to declare every widget it references. no logic, labels, or UI copy touched.
2. **`TestWidgetReturnCompleteness` in `labs/tests/test_static.py`.** fails on any new cell defining a widget without returning it (render sinks like `tabs`/`_tabs` and tabs-cell-local widgets excluded). 33 / 33 labs pass; a regression is blocked at CI time.
## test plan
- [x] `labs/tests/test_static.py` + `tests/test_engine.py`: 825 passed 4 skipped 1 xfailed (up from 792 — 33 new test runs)
- [x] `marimo check` spot-sampled on several labs: exit 0
- [x] audit script: 17 → 0 offending labs on the meaningful check
- [ ] CI green on this pr
closes the main thread of #1332. @asgalon's "updated example values" suggestion for Lab 02 and the tabs-cell widget extraction cleanup left as follow-ups.
---
<sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
📋 Pull Request Information
Original PR: https://github.com/harvard-edge/cs249r_book/pull/1386
Author: @profvjreddi
Created: 4/17/2026
Status: ✅ Merged
Merged: 4/17/2026
Merged by: @profvjreddi
Base:
dev← Head:fix/labs-widget-return-audit📝 Commits (1)
80b916ffix(labs): widget return-tuple sweep + static test for regressions (#1332)📊 Changes
34 files changed (+419 additions, -326 deletions)
View changed files
📝
labs/tests/test_static.py(+82 -0)📝
labs/vol1/lab_00_introduction.py(+5 -2)📝
labs/vol1/lab_01_ml_intro.py(+11 -15)📝
labs/vol1/lab_02_ml_systems.py(+13 -17)📝
labs/vol1/lab_03_ml_workflow.py(+6 -8)📝
labs/vol1/lab_04_data_engr.py(+8 -4)📝
labs/vol1/lab_05_nn_compute.py(+9 -4)📝
labs/vol1/lab_06_nn_arch.py(+8 -4)📝
labs/vol1/lab_07_ml_frameworks.py(+8 -4)📝
labs/vol1/lab_08_model_train.py(+9 -4)📝
labs/vol1/lab_09_data_selection.py(+9 -4)📝
labs/vol1/lab_10_model_compress.py(+8 -4)📝
labs/vol1/lab_11_hw_accel.py(+8 -4)📝
labs/vol1/lab_12_perf_bench.py(+8 -4)📝
labs/vol1/lab_13_model_serving.py(+9 -4)📝
labs/vol1/lab_14_ml_ops.py(+8 -4)📝
labs/vol1/lab_15_responsible_engr.py(+8 -4)📝
labs/vol1/lab_16_ml_conclusion.py(+10 -4)📝
labs/vol2/lab_01_introduction.py(+11 -5)📝
labs/vol2/lab_02_compute_infra.py(+10 -5)...and 14 more files
📄 Description
what
sweeps the whole labs tree to fix a marimo dataflow bug class that left widgets invisible to the tabs cell in 17 of 33 labs. reported by @asgalon in #1332 for lab_02 + lab_03 (fixed in #1385); this pr does the rest and locks the regression out via a new static test.
the bug
marimo routes widget state strictly through each cell's return tuple. every
@app.cellthat defined helper widgets (sliders, dropdowns) alongside the terminalpartX_predictionreturned only the prediction. the helpers were defined but never reached the tabs cell — they were effectively dead allocations. sliders never rendered after the prediction unlocked; chart inputs never re-ran on change. every static and engine test still passed because the cells executed fine in isolation.the audit script (
/tmp/audit_widget_returns.py) found:(the 42 "hits" that remain in the updated audit are all widgets defined inside the tabs cell itself, consumed by its own
build_part_*closures via scope — they don't need to flow outward. the new static test excludes those.)changes
return (...)tuple now lists everymo.ui.*it defines, alphabetical. every tabs-cell signature is rewritten to declare every widget it references. no logic, labels, or UI copy touched.TestWidgetReturnCompletenessinlabs/tests/test_static.py. fails on any new cell defining a widget without returning it (render sinks liketabs/_tabsand tabs-cell-local widgets excluded). 33 / 33 labs pass; a regression is blocked at CI time.test plan
labs/tests/test_static.py+tests/test_engine.py: 825 passed 4 skipped 1 xfailed (up from 792 — 33 new test runs)marimo checkspot-sampled on several labs: exit 0closes the main thread of #1332. @asgalon's "updated example values" suggestion for Lab 02 and the tabs-cell widget extraction cleanup left as follow-ups.
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.