scheduling fixes

[gkennos]

Summary

scheduling roll out fixes

Checklist

• Applied exactly one label (breaking, feature, fix, dependencies, or chore)
• Tests pass locally (uv run pytest -q)
• Lint passes (uv run ruff check .)
• Generated model changes were regenerated and reviewed, if applicable

[Copilot]

🟡 Changes recommended

Fix the rollout end calculation, fallback status prefixing, and shared network ownership behavior.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Adds deterministic treatment schedule rollout across cycles and phases, with elapsed timing and resolution metadata.

Changes:

• Adds cycle/block anchoring and phase rollout logic.
• Integrates timing metadata into administration frames.
• Adds tests, exports, documentation, and shared devcontainer networking.

File summaries

File	Summary
tests/test_schedule_properties.py	Adds rollout and timing tests.
hemonc_alchemy/toolkit/analytics/treatment/scheduling/rollout.py	Implements rollout logic; phase end calculation and fallback status handling require changes.
hemonc_alchemy/toolkit/analytics/treatment/scheduling/properties.py	Integrates elapsed timing into administration frames.
hemonc_alchemy/toolkit/analytics/treatment/scheduling/__init__.py	Exports rollout APIs.
docs/toolkit/scheduling.md	Documents timeline behavior and APIs.
.devcontainer/compose.yaml	Configures shared networking; network ownership must be corrected for sibling devcontainers.

Review details

Suppressed comments (3)

.devcontainer/compose.yaml:68

• A fixed name without external: true leaves this network Compose-owned by the first project. Starting a second checkout/devcontainer with a different Compose project name can then fail on the existing com.docker.compose.project label instead of joining the network, so the stated sibling-devcontainer behavior is not reliable. Use a pre-created external network or keep the project-scoped default and document the setup.

    name: hemonc-alchemy_default

hemonc_alchemy/toolkit/analytics/treatment/scheduling/rollout.py:675

• When _phase_order uses a documented fallback, it returns a non-None order_reason beginning with resolved_via_fallback:. This condition therefore prevents phase_start from advancing between fallback-ordered phases, so every phase is rolled out at the same offset even though the status says the ordering was resolved. Only unresolved ordering should suppress phase chaining.

        if phase_end is not None and order_reason is None:

hemonc_alchemy/toolkit/analytics/treatment/scheduling/rollout.py:710

• order_reason also carries the resolved fallback status, but this branch unconditionally prefixes it with unresolved:. An empty result from a fallback-ordered variant therefore exposes a contradictory frame.attrs["timing_status"]; preserve already-prefixed statuses and only add unresolved: to a bare failure reason.

    if order_reason and frame.empty:
        frame.attrs["timing_status"] = f"unresolved: {order_reason}"

• Files reviewed: 6/6 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[brianna-argall]

I'm happy with this :)

My remaining suggestions are around indefinite handling:

• Ideally (+1) indefinites would be handled and not appear as none in the indefinite column. Currently an appropriate warning is displayed
• Given that indefinite can apply to either days within a cycle or cycle continuation ideally the column label would be more transparent about the nature of the indefinite

Additionally, there is currently no check on being systemic - rt sigs roll out with drug NaN

Edge case variants thought not to have been specifically examined in development were examined including:

• Carboplatin and Paclitaxel (CP) and Tislelizumab var2 : A perioperative variant with a surgery phase and tislelizumab used both in neoadjuvant and adjuvant contexts
• Carboplatin, Osimertinib, Pemetrexed var1: A variant with both PO and IV sigs, with the PO drug delivered on sequential days from compact notation
• Carboplatin and Paclitaxel (CP) and Ipilimumab var3: A variant with a component delivered cycle 3 only whereas other components have cycle 1-6 deliveries.

[Copilot]

Copilot review overview

🟡 Changes recommended

Critical startup and schedule-anchoring issues, along with additional unresolved rollout correctness issues, remain.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 2 · 1

Open (3)

• External network breaks fresh-clone Compose startup · New
• Nonconsecutive initial cycles are anchored at day zero · New
• First aggregation drops indefinite cycle markers · New

Resolved since last review (1)

• This returns only the end of the last sorted block. For overlapping blocks, a later-start block can…

[Copilot]

Copilot review overview

🟡 Changes recommended

Four unresolved review findings include two critical rollout issues and two moderate scheduling/parser issues.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 3 · 2

Open (5)

• Validate all overlapping anchors for conflicting cycle starts · New
• Do not suppress phase-step gap checks with fallback phases · New
• External network breaks fresh-clone Compose startup
• Reject zero-length indefinite intervals · New
• First aggregation drops indefinite cycle markers

Resolved since last review (1)

• Nonconsecutive initial cycles are anchored at day zero

[Copilot]

Copilot review overview

🟡 Changes recommended

Moderate correctness issues remain in status aggregation, overlap anchoring, and ambiguous timing choices.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1 · 1

Open (2)

• Validate all overlapping anchors for conflicting cycle starts
• Select the actual preceding block after overlaps · New

Resolved since last review (4)

• Do not suppress phase-step gap checks with fallback phases
• External network breaks fresh-clone Compose startup
• Reject zero-length indefinite intervals
• First aggregation drops indefinite cycle markers

[Copilot]

Copilot review overview

🟡 Changes recommended

Unresolved rollout correctness issues remain in choice handling, phase consistency validation, and negative-day offsets.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 2

Open (2)

• Do not resolve unspecified Choice timing sequences · New
• Validate all overlapping anchors for conflicting cycle starts

Resolved since last review (1)

• Select the actual preceding block after overlaps

Previously missed (1)

In code that hasn't changed since last review

Correct off-by-one conversion for negative cycle days

hemonc_alchemy/​toolkit/​analytics/​treatment/​scheduling/​rollout.py:529

The day - 1 conversion is off by one for the documented negative-day notation: -14 means 14 days before cycle day 1, but this produces offset -15 in both elapsed and calendar timelines. Apply the 1-based conversion only to positive cycle days and retain negative values as their stated lead-in offsets.