src/executorlib/standalone/batched.py (1)

19-25: ⚠️ Potential issue | 🟠 Major | 🏗️ Heavy lift

Fix batched_futures: value-deduped skip_set can overestimate n_expected and leave later batches permanently unresolved when duplicate result values occur.

In src/executorlib/standalone/batched.py, n_expected = min(n, len(lst) - len(skip_set)), but eligibility is computed via v.result() not in skip_set (value-based). In src/executorlib/task_scheduler/interactive/dependency.py, skip_set is built as a deduplicated set of prior batch results, so duplicates in lst that were already consumed by an earlier batch still remove all remaining occurrences by value, while n_expected only subtracts once per distinct value.

When there aren’t enough remaining non-skipped values to fill the batch, batched_futures keeps returning [], and _update_waiting_task keeps the batch waiting forever.

Concrete example (single value repeated in the consumed batch):

• lst results: [1, 1, 1, 2], n=3
• Batch 1 returns [1, 1, 1]
• Batch 2: skip_set={1}, so n_expected=min(3, 4-1)=3, but only the single future with result 2 is eligible → returns [] indefinitely.

Add tests with duplicate result values (and/or change the algorithm to track already-assigned counts or per-future identity rather than a value-deduped set).

🤖 Prompt for 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.

In `@src/executorlib/standalone/batched.py` around lines 19 - 25, batched_futures
overestimates n_expected because skip_set is value-deduped; change
batched_futures (and its use of skip_set) to account for duplicate result values
by tracking counts or per-future identity: either accept a skip_counts Counter
(value -> number skipped) or build a skip_futures set of already-assigned
futures, then compute n_expected as min(n, count of futures in lst not excluded
by the skip tracking) and when collecting done results decrement the skip count
or mark that specific future as skipped; update references to skip_set,
n_expected, lst and done_lst accordingly and add tests for duplicate-result
scenarios.