fix(hockey): derive the ESPN season window instead of pinning it to 2025-26

[ChuckBuilds]

Season-readiness pass over football-scoreboard and hockey-scoreboard ahead of the openers (NFL ~Sept 10, NHL ~Oct 7). Most of what I checked is already correct — details at the bottom — but two things were worth changing.

The pinned season window

hockey-scoreboard/data_fetcher.py::_get_season_date_range computed now and then ignored it:

now = datetime.now()                    # never used
season_start = datetime(2025, 10, 1)
season_end   = datetime(2026, 6, 30)

That string goes straight into the request as ?dates={range}, so once the 2026-27 season opened it would have kept asking ESPN for last season. It now derives the year with the same August rollover rule nhl_managers.py already uses, so the two agree on which season is current instead of drifting apart:

date	before	after
2026-07-31	20251001-20260630	20251001-20260630
2026-08-22	20251001-20260630	20261001-20270630
2026-10-07 (opener)	20251001-20260630	20261001-20270630
2027-08-01	20251001-20260630	20271001-20280630

Reachability, stated plainly: data_fetcher.py is imported only by debug_tb_games.py, so this is not on the runtime path today — the live fetchers already derive their windows correctly. I still think it's worth fixing: the committed debug script would pull the wrong season, and anyone wiring the module back in inherits a bug that surfaces once a year.

Writing the test also turned up a function-local from datetime import datetime shadowing the module import. Patching the module attribute had no effect, which is precisely why this couldn't be tested before.

The missing football guard

hockey-scoreboard has test_empty_mode_signals_no_content.py, added after the bug where an empty mode reported success and sat on a blank panel for its full duration. football-scoreboard had no equivalent — and the two ship separate forks of sports.py, so a fix in one lineage never reaches the other.

That failure mode is exactly the state both plugins are in right now: pre-season, with recent and live empty while upcoming fills in. Confirmed the new test catches it — making SportsRecent.display return None fails 3 checks.

Football's fork also overrides display() on SportsLive, which hockey's doesn't. Its only non-delegating paths are return False (disabled) and return True (celebration); the empty case falls through to super().display(). So SportsLive gets a targeted disabled-path check rather than being forced through the stand-in, which can't satisfy super().

What I checked that was already fine

• Season rollover on the live path — both nfl_managers.py and nhl_managers.py derive dynamically and both flipped on 2026-08-01. NFL now requests 20260801-20270301 (covers a ~Feb 2027 Super Bowl); NHL 20260901-20270801 (covers a June 2027 Cup).
• Schemas — both valid Draft-07; a default config generated from each validates cleanly.
• Widgets — both use only checkbox-group and color-picker, both registered in the web UI.
• Team pickers — NFL 32 (LV/LAC/LAR/WSH, no OAK/SD/STL/WAS); NHL 32 (UTAH present, relocated ARI gone).
• Config wiring — every schema option is read. This took three passes to establish: keys reached via f-strings (f"{mode_type}_mode_duration", f"{result}_color"), via helpers taking the key as a parameter (_scroll_card_option), and via hockey's nested→flat resolve_value(["teams","favorite_teams"], ["favorite_teams"]) all look unwired to a naive scan. I chased each one to the call site rather than trusting the counts.
• Display contract — no bare returns or implicit-None fall-throughs in either plugin's display() or its delegates.

One thing left, for you

hockey.defaults.season_cache_duration_seconds (also in lacrosse-scoreboard) is declared with x-advanced, a default of 86400 and min/max bounds — and is read nowhere in either repo. It's a knob users can set that does nothing. Removing it is a user-facing schema change, so I've left it; say the word and I'll drop it from both.

football-scoreboard's version (2.19.1) and versions[0] (2.18.1) are drifted, one of the eight I flagged on #314. Not touched here since nothing shipped changed in that plugin.

Verification

football 16 passed, hockey 12 passed (up from 15 and 11). The season test fails 9 checks against the previous code.

Not verified on hardware — both rigs have been unreachable all day.

🤖 Generated with Claude Code

https://claude.ai/code/session_01STMbQE4YctTacQXfbYqKuW

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 7236086e-f4f2-44df-bd5f-bdd58ca05064

---

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.}

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 25 complexity

Metric	Results
Complexity	25

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[ChuckBuilds]

Force-pushed to correct the branch scope — worth flagging since the history changed.

This branch had picked up four commits that don't belong to it (#322's soccer registry fix and #323's youtube-stats fix). My mistake: an earlier merge-order test ran git checkout -B /tmp/mergetest origin/main 2>/dev/null, and a branch name starting with / is invalid. The 2>/dev/null swallowed that error, so the test merges landed on my local main instead, and every branch cut afterwards inherited them.

Nothing was wrong with the code — all four PRs were green — but merging this one would have silently merged #322 and #323 with it.

The branch is now a single commit on top of origin/main, touching only its own files. No review comments existed on it, so nothing was lost. Tests re-run on the rebuilt branch and still pass.