feat(offerings): expose the EFFECT notice's effective time

[shahar-arbor]

What this changes

Before:

• Effect.from_xml() read finalEffectivenessDispDate and dropped finalEffectivenessDispTime.
• Nothing said that an EFFECT notice is not public on its filing_date.

After:

• Effect.effective_time returns the time as "HH:MM:SS", or None when the notice has none.
• to_context() shows the time.
• The effective_date docstring and the Effect guide say that SEC accepts and publishes the notice at about 00:15 ET on the next business day, and point point-in-time users to filing.header.acceptance_datetime.

EffectiveData takes the time as a new keyword argument with a default, so existing constructor calls still work.

Fixes #1499

Verification

• Ground truth: the test uses two notices, with XML as SEC serves it, with only blank lines removed.
  • Lyntris Inc.'s S-1 notice 9999999995-26-002694: effective_time == "15:30:00", effective_date == "2026-08-18".
  • Fitness Fanatics' POS AM notice 9999999995-25-003344: effective_time is None.
  • Checked by hand: the Lyntris header gives ACCEPTANCE-DATETIME 20260819001513, and Fitness Fanatics' gives 20251124001520 (Monday) for a 2025-11-21 filing date.
• Regression test: tests/issues/regression/test_issue_1499.py, offline, links the issue. It fails on main and passes with the fix.
• Silence check: a notice without the element gives None, and to_context() leaves the line out, rather than showing an empty time.
• Docs: docs/guides/effect-data-object-guide.md gains the property row and the publication-time note.
• Changelog: changelog.d/1499.added.md. assemble_changelog.py --check passes.

Also run:

• tests/core/test_effect.py: the offline tests pass. The two get_source_filing tests need the network and were not run.
• ruff check: clean on both files.
• check_regression_provenance.py and check_regression_skips.py: OK.

6.0

Not a 6.0 change.

Working context (optional)

We use EFFECT notices to date IPOs. Sample used for the publication lag: 60 EFFECT notices for IPO registrations filed Aug 2025 – Aug 2026. For each, the header ACCEPTANCE-DATETIME was compared with filing_date:

• 52 were accepted at about 00:15 ET the next day;
• 8 were accepted on Monday for a Friday filing date.

I also considered changing RegistrationS1.effective_date to take the latest EFFECT, not the first. I left it alone. On Fitness Fanatics' F-1 file (333-289484), the second EFFECT is for a POS AM, so the first EFFECT is the right answer for "when was the F-1 declared effective".

🤖 Generated with Claude Code

[dgunning]

Thanks, this looks good to merge.

• The new final_effective_time keyword has a default, and Effect.from_xml() is the only place in the library that builds EffectiveData, so existing callers can't break.
• When a notice has no time, the property returns None and to_context() leaves the line out, which is the right behaviour for an optional element.
• The regression tests pass locally against the real Lyntris and Fitness Fanatics notices.
• The note in the guide and docstring that an EFFECT notice isn't public until about 00:15 ET on the next business day will be useful to anyone doing point-in-time work, and pointing them to filing.header.acceptance_datetime is the right steer.

This is additive, so it can go into 5.62.0, the last 5.x release, if it merges by 10-11.