fix(sec): A3 slice-2a review — shares fallback, robust context, hardening

1. Multi-class shares: prefer the single dei:EntityCommonStockSharesOutstanding
   cover-page fact; else fall back to us-gaap:CommonStockSharesOutstanding at
   period end (Alphabet has no dei fact). Never sum class facts (companyfacts is
   non-dimensional) and never use weighted-average/diluted; conflicting values ->
   null, counted as an "ambiguous shares outstanding" note in validation. Plan's
   "sum class-specific" wording corrected. Verified live: Alphabet shares now
   populate (12.1B), Apple still uses its dei cover date.
2. Fiscal context is the majority (fy, fp) among facts ending at reportDate, with
   ties rejected — no longer the arbitrary first fact.
3. Hardening: catalog selectors require taxonomy == "us-gaap"; indexing drops
   malformed facts (missing accession/end, non-finite value) so a custom concept
   or bad date can't be selected.

Tests: +8 (dei precedence, us-gaap fallback, conflict->null, no weighted-average,
tie-context skip, foreign-taxonomy/malformed ignored, ambiguous-shares note).
14 passed, 1 skipped.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

app/services/sec_facts_parser.py

@@ -20,7 +20,8 @@ The load-bearing rules (design Decision 2 + review):
 from __future__ import annotations
 
 import logging
-from dataclasses import dataclass, field
+import math
+from dataclasses import dataclass
 from datetime import date, datetime
 from typing import Any, NamedTuple
 
@@ -126,11 +127,13 @@ def parse_snapshots(
         if meta is None or not facts:
             skips.append({"accession": accn, "reason": "no facts or filing metadata"})
             continue
-        row = _parse_one(cik, accn, facts, meta)
+        row, note = _parse_one(cik, accn, facts, meta)
         if row is None:
-            skips.append({"accession": accn, "reason": "no usable period identity"})
+            skips.append({"accession": accn, "reason": note or "unparseable"})
             continue
         rows.append(row)
+        if note:  # row produced, but a field-level issue to count in validation
+            skips.append({"accession": accn, "reason": note})
     return rows, skips
 
 
@@ -142,7 +145,12 @@ def _index_by_accession(companyfacts: dict[str, Any]) -> dict[str, list[Fact]]:
             for unit, facts in body.get("units", {}).items():
                 for f in facts:
                     accn = f.get("accn")
-                    if not accn:
+                    end = _d(f.get("end"))
+                    val = f.get("val")
+                    # Skip malformed facts so they can't be selected accidentally:
+                    # every usable fact needs an accession, an end date, and a
+                    # finite numeric value.
+                    if not accn or end is None or not _finite(val):
                         continue
                     out.setdefault(accn, []).append(
                         Fact(
@@ -150,8 +158,8 @@ def _index_by_accession(companyfacts: dict[str, Any]) -> dict[str, list[Fact]]:
                             concept=concept,
                             unit=unit,
                             start=_d(f.get("start")),
-                            end=_d(f.get("end")),
-                            val=f.get("val"),
+                            end=end,
+                            val=val,
                             fy=f.get("fy"),
                             fp=f.get("fp"),
                         )
@@ -159,10 +167,15 @@ def _index_by_accession(companyfacts: dict[str, Any]) -> dict[str, list[Fact]]:
     return out
 
 
-def _parse_one(cik: str, accn: str, facts: list[Fact], meta: FilingMeta) -> SnapshotRow | None:
-    fy, fp = _fiscal_context(facts)
+def _parse_one(
+    cik: str, accn: str, facts: list[Fact], meta: FilingMeta
+) -> tuple[SnapshotRow | None, str | None]:
+    """Returns (row, note). row is None when there's no usable period identity;
+    note is a validation reason (row-skip reason when row is None, else a
+    field-level issue such as ambiguous shares)."""
+    fy, fp = _fiscal_context(facts, meta.report_date)
     if fy is None or fp not in _EXPECTED_YTD_DAYS:
-        return None
+        return None, "no usable period identity"
 
     row = SnapshotRow(
         cik=cik,
@@ -189,16 +202,26 @@ def _parse_one(cik: str, accn: str, facts: list[Fact], meta: FilingMeta) -> Snap
     # balance-sheet instants at reportDate
     row.cash_and_st_investments = _compose_cash(facts, meta.report_date)
     row.total_debt = _compose_debt(facts, meta.report_date)
-    row.shares_outstanding, row.shares_outstanding_date = _select_shares(facts)
-    return row
+    shares, shares_date, ambiguous = _select_shares(facts, meta.report_date)
+    row.shares_outstanding = shares
+    row.shares_outstanding_date = shares_date
+    return row, ("ambiguous shares outstanding" if ambiguous else None)
 
 
-def _fiscal_context(facts: list[Fact]) -> tuple[int | None, str | None]:
-    """A filing's own (fy, fp) — shared by all its facts; take the first set."""
+def _fiscal_context(facts: list[Fact], report_date: date) -> tuple[int | None, str | None]:
+    """The filing's (fy, fp) taken as the majority context among the facts that
+    end at reportDate (the current-period facts, which share the filing's
+    context). Reject a tie so a conflicting context is never chosen arbitrarily."""
+    counts: dict[tuple[int, str], int] = {}
     for f in facts:
-        if f.fy is not None and f.fp:
-            return f.fy, f.fp
-    return None, None
+        if f.end == report_date and f.fy is not None and f.fp:
+            counts[(f.fy, f.fp)] = counts.get((f.fy, f.fp), 0) + 1
+    if not counts:
+        return None, None
+    ranked = sorted(counts.items(), key=lambda kv: kv[1], reverse=True)
+    if len(ranked) > 1 and ranked[0][1] == ranked[1][1]:
+        return None, None  # tie → conflicting contexts, reject
+    return ranked[0][0]
 
 
 def _select_ytd(
@@ -212,11 +235,11 @@ def _select_ytd(
         best_diff: int | None = None
         for f in facts:
             if (
-                f.concept != concept
+                f.taxonomy != "us-gaap"
+                or f.concept != concept
                 or f.unit != unit
                 or f.start is None
                 or f.end != report_date
-                or f.val is None
             ):
                 continue
             diff = abs((f.end - f.start).days - expected)
@@ -232,11 +255,11 @@ def _select_instant(facts: list[Fact], concepts: list[str], report_date: date) -
     for concept in concepts:
         for f in facts:
             if (
-                f.concept == concept
+                f.taxonomy == "us-gaap"
+                and f.concept == concept
                 and f.unit == "USD"
                 and f.start is None
                 and f.end == report_date
-                and f.val is not None
             ):
                 return float(f.val)
     return None
@@ -262,22 +285,49 @@ def _compose_debt(facts: list[Fact], report_date: date) -> float | None:
     return (long_term or 0.0) + (short_term or 0.0)
 
 
-def _select_shares(facts: list[Fact]) -> tuple[float | None, date | None]:
-    """dei:EntityCommonStockSharesOutstanding — cover-page instant. Store its own
-    end (the cover date, which differs from period_end)."""
-    candidates = [
+def _select_shares(
+    facts: list[Fact], report_date: date
+) -> tuple[float | None, date | None, bool]:
+    """Issuer-wide shares outstanding as a single consolidated value (never a
+    class sum — companyfacts is non-dimensional — and never weighted-average/
+    diluted). Returns (value, shares_date, ambiguous).
+
+    1. Prefer the `dei:EntityCommonStockSharesOutstanding` cover-page instant;
+       its own end is the shares date (cover date != period_end).
+    2. Else fall back to `us-gaap:CommonStockSharesOutstanding` at period end
+       (e.g. Alphabet has no dei fact); shares date = reportDate.
+    Conflicting values within the chosen source → (None, None, True) to be
+    counted in validation.
+    """
+    dei = [
         f
         for f in facts
         if f.taxonomy == "dei"
         and f.concept == "EntityCommonStockSharesOutstanding"
         and f.unit == "shares"
         and f.start is None
-        and f.val is not None
     ]
-    if not candidates:
-        return None, None
-    best = max(candidates, key=lambda f: f.end)
-    return float(best.val), best.end
+    if dei:
+        if len({f.val for f in dei}) > 1:
+            return None, None, True
+        best = max(dei, key=lambda f: f.end)
+        return float(best.val), best.end, False
+
+    gaap = [
+        f
+        for f in facts
+        if f.taxonomy == "us-gaap"
+        and f.concept == "CommonStockSharesOutstanding"
+        and f.unit == "shares"
+        and f.start is None
+        and f.end == report_date
+    ]
+    if gaap:
+        if len({f.val for f in gaap}) > 1:
+            return None, None, True
+        return float(gaap[0].val), report_date, False
+
+    return None, None, False  # simply absent — not a conflict
 
 
 def _d(value: Any) -> date | None:
@@ -287,3 +337,8 @@ def _d(value: Any) -> date | None:
         return date.fromisoformat(str(value)[:10])
     except ValueError:
         return None
+
+
+def _finite(value: Any) -> bool:
+    """True for a finite numeric value (rejects None, bool, strings, NaN/inf)."""
+    return isinstance(value, (int, float)) and not isinstance(value, bool) and math.isfinite(value)

docs/dolt-integration-plan.md

@@ -253,12 +253,14 @@ that scoring already reads, refreshed daily by step (c) after activation.
 | Earnings surprise history | last 4+ from `earnings_events` | query |
 
 **Market cap is an estimate** (issuer-wide shares outstanding × one ticker's price —
-approximate for multi-class issuers). For a multi-class issuer, derive the
-issuer-wide share count **either** from the consolidated cover-page figure **or**
-by summing the class-specific `dei:EntityCommonStockSharesOutstanding` facts
-(GOOG + GOOGL) — **never both**, or the count double-counts. Label it "est." in the
-UI and round aggressively rather than withholding it; false precision is the failure
-mode, not the approximation.
+approximate for multi-class issuers). Share count comes from a single consolidated
+value, not a class sum: prefer the one `dei:EntityCommonStockSharesOutstanding`
+cover-page fact; if absent (e.g. Alphabet) fall back to
+`us-gaap:CommonStockSharesOutstanding` at period end. companyfacts is
+non-dimensional, so class-specific facts can't be summed reliably — never do that,
+and never substitute weighted-average/diluted shares; if conflicting values remain,
+store null. Label it "est." in the UI and round aggressively rather than withholding
+it; false precision is the failure mode, not the approximation.
 
 **Units follow existing app conventions:** percentages are percentage points
 (21.0 = 21%), P/E and net-debt/EBITDA are multiples, market cap and net debt are

tests/unit/test_sec_facts_parser.py

@@ -13,6 +13,8 @@ from app.services.sec_facts_parser import (
     FilingMeta,
     _compose_cash,
     _compose_debt,
+    _fiscal_context,
+    _select_shares,
     parse_snapshots,
 )
 
@@ -135,6 +137,95 @@ def test_cash_picks_one_st_investment_source():
     assert _compose_cash(facts, rd) == 45000
 
 
+RD = date(2026, 3, 28)
+
+
+def _dei(end, val):
+    return Fact("dei", "EntityCommonStockSharesOutstanding", "shares", None, end, val, 2026, "Q2")
+
+
+def _gaap_shares(end, val):
+    return Fact("us-gaap", "CommonStockSharesOutstanding", "shares", None, end, val, 2026, "Q2")
+
+
+def test_shares_prefers_dei_cover_page():
+    facts = [_dei(date(2026, 4, 17), 100), _gaap_shares(RD, 999)]
+    assert _select_shares(facts, RD) == (100.0, date(2026, 4, 17), False)
+
+
+def test_shares_falls_back_to_usgaap_at_report_date():
+    # Alphabet case: no dei fact; us-gaap current + a prior comparative.
+    facts = [_gaap_shares(date(2025, 12, 31), 888), _gaap_shares(RD, 12116)]
+    assert _select_shares(facts, RD) == (12116.0, RD, False)  # comparative excluded
+
+
+def test_shares_conflict_returns_null_ambiguous():
+    facts = [_dei(RD, 100), _dei(RD, 200)]  # two differing consolidated values
+    assert _select_shares(facts, RD) == (None, None, True)
+
+
+def test_shares_never_uses_weighted_average():
+    facts = [Fact("us-gaap", "WeightedAverageNumberOfDilutedSharesOutstanding", "shares", None, RD, 5, 2026, "Q2")]
+    assert _select_shares(facts, RD) == (None, None, False)  # not a shares source
+
+
+def test_conflicting_fiscal_context_is_rejected():
+    facts = [
+        Fact("us-gaap", "Revenues", "USD", date(2025, 9, 28), RD, 1, 2026, "Q2"),
+        Fact("us-gaap", "NetIncomeLoss", "USD", date(2025, 9, 28), RD, 2, 2025, "Q3"),
+    ]  # 1-1 tie between two contexts at reportDate
+    assert _fiscal_context(facts, RD) == (None, None)
+    # a clear majority wins
+    facts.append(Fact("us-gaap", "OperatingIncomeLoss", "USD", date(2025, 9, 28), RD, 3, 2026, "Q2"))
+    assert _fiscal_context(facts, RD) == (2026, "Q2")
+
+
+def test_conflicting_context_skips_row():
+    cf = {"cik": 1, "facts": {"us-gaap": {
+        "Revenues": {"units": {"USD": [_dur("2025-09-28", "2026-03-28", 1, "B", fp="Q2")]}},
+        "NetIncomeLoss": {"units": {"USD": [_dur("2025-09-28", "2026-03-28", 2, "B", fy=2025, fp="Q3")]}},
+    }}}
+    filings = {"B": FilingMeta(RD, date(2026, 5, 1), datetime(2026, 5, 1, tzinfo=UTC), "10-Q")}
+    rows, skips = parse_snapshots(cf, filings, {"B"})
+    assert rows == [] and skips == [{"accession": "B", "reason": "no usable period identity"}]
+
+
+def test_foreign_taxonomy_and_malformed_facts_ignored():
+    cf = {"cik": 1, "facts": {
+        "us-gaap": {
+            "Revenues": {"units": {"USD": [_dur("2025-09-28", "2026-03-28", 500, "B")]}},
+            "NetIncomeLoss": {"units": {"USD": [
+                {"start": "2025-09-28", "end": "2026-03-28", "val": None, "fy": 2026, "fp": "Q2", "accn": "B"},
+            ]}},
+            "OperatingIncomeLoss": {"units": {"USD": [
+                {"start": "2025-09-28", "end": "2026-03-28", "val": float("nan"), "fy": 2026, "fp": "Q2", "accn": "B"},
+            ]}},
+        },
+        "acme": {"RevenueFromContractWithCustomerExcludingAssessedTax": {"units": {"USD": [
+            _dur("2025-09-28", "2026-03-28", 99999, "B"),  # custom taxonomy — must be ignored
+        ]}}},
+    }}
+    filings = {"B": FilingMeta(RD, date(2026, 5, 1), datetime(2026, 5, 1, tzinfo=UTC), "10-Q")}
+    rows, _ = parse_snapshots(cf, filings, {"B"})
+    assert rows[0].revenue == 500          # us-gaap Revenues, not the acme concept
+    assert rows[0].net_income is None      # val None ignored
+    assert rows[0].operating_income is None  # NaN ignored
+
+
+def test_ambiguous_shares_produces_row_plus_note():
+    cf = {"cik": 1, "facts": {
+        "us-gaap": {"Revenues": {"units": {"USD": [_dur("2025-09-28", "2026-03-28", 500, "B")]}}},
+        "dei": {"EntityCommonStockSharesOutstanding": {"units": {"shares": [
+            {"end": "2026-04-17", "val": 100, "fy": 2026, "fp": "Q2", "accn": "B"},
+            {"end": "2026-04-17", "val": 200, "fy": 2026, "fp": "Q2", "accn": "B"},
+        ]}}},
+    }}
+    filings = {"B": FilingMeta(RD, date(2026, 5, 1), datetime(2026, 5, 1, tzinfo=UTC), "10-Q")}
+    rows, skips = parse_snapshots(cf, filings, {"B"})
+    assert len(rows) == 1 and rows[0].shares_outstanding is None  # row kept, shares null
+    assert {"accession": "B", "reason": "ambiguous shares outstanding"} in skips
+
+
 # Opt-in live check against real Apple companyfacts. Skips unless SEC_LIVE=1 and a
 # real SEC_USER_AGENT are set (network + fair-access contact email).
 @pytest.mark.skipif(