partial_start_coverage | prices.history | First bar is more than the tolerance after the requested start (clipped / newly-listed / source lag). | — |
partial_end_coverage | prices.history | Last bar is more than the tolerance before the requested end (VN trading-calendar aware); series may be stale. | — |
trailing_zero_volume_tail | prices.history / index_history | Trailing run of ≥10 zero-volume flat (O=H=L=C) bars — likely suspended/delisted or forward-filled phantom data (bars kept, not dropped). | #176 |
quarantined_invalid_bars | prices.history / index_history | Isolated corrupt bars dropped (OHLC-invariant, non-positive/non-finite, bad volume, duplicate timestamp); the rest of the window is served. | #186 |
resampled_from_d1 | prices.history / index_history | Series aggregated client-side from D1 to a coarser period (W1/MN1/Q1/Y1), not native to the interval. | #183 |
resample_partial_period | prices.history / index_history | An edge resampled period is incomplete relative to the window (bars kept, marked provisional). | #183 |
deduped_duplicate_daily_index_bars | index_history | Identical same-date duplicate index bar kept once; a conflicting same-date pair drops that date entirely. | #162 |
recovered_midnight_open_placeholder | index_history | A same-date D1 pair identical in high/low/close/volume but differing only in open, where exactly one row is at VN-local 00:00 — that midnight row is a synthetic open placeholder, so it is dropped and the real (non-midnight) row is kept. A recovery (not a quarantine): the date is served and is not charged to the failover threshold. | #187 |
stitched_multi_source | index_history_stitched | History stitched across per-calendar-year segments (each stitched year also emits a stitched_segment provenance line). | #147 |
stitched_segment | index_history_stitched | Per-segment provenance of a stitched series — which source served a given calendar year and how many bars (stitched_segment: <year> <source> (<n> bars)). | #147 |
weights_not_available | index_constituents | Membership only — no per-stock index weights (weight=None); never fabricated. | — |
current_snapshot_only | index_constituents | Always present — the basket is the CURRENT membership as fetched, NOT a point-in-time/historical snapshot (as_of=None, never fabricated); backtests using it inherit survivorship and look-ahead bias. | #175 |
fallback_instrument_served | indices.world | Requested SPY (USD/share) unavailable; served Stooq ^SPX (index points, ~10× different magnitude) — rebase before comparing. | #177 |
proxy_substitution | indices.world | Asked a raw index (^N225/^SSEC/^STI/^KS11/^CSI300/^HSI) but served a USD ETF proxy (EWJ/FXI/EWS/EWY/ASHR/EWH) — embeds USD/local FX, NOT a faithful tracker/raw index point series. Also exposed via the structured PriceHistory.proxy_for field. | #193/#197 |
world_reference_excludes_domestic_premium | gold.world_reference_history_vnd | Always present — world-gold-implied VND, excludes the +10–21% VN domestic (SJC/BTMC) premium; NOT the domestic price. | #178 |
world_reference_annual_basis | gold.world_reference_history_vnd | Always present — one point per calendar year (annual-avg gold × annual USD/VND × 37.5/31.1035), stamped Jan-1; not a daily series. | #178 |
world_reference_partial_year_coverage | gold.world_reference_history_vnd | Some requested years dropped for lack of a paired gold+FX observation (honest intersection). | #178 |
world_reference_trailing_year_incomplete | gold.world_reference_history_vnd | Latest emitted year is the current in-progress year — its mean is year-to-date, not a full-year value. | #178 |
world_reference_gold_source_fallback | gold.world_reference_history_vnd | Primary World Bank CMO annual source failed → daily-averaging (CurrencyApi→Stooq) fallback used. | #185 |
world_reference_gold_leg_* / world_reference_fx_leg_* | gold.world_reference_history_vnd | Family: an upstream gold-/FX-leg warning forwarded, namespaced by leg (e.g. world_reference_gold_leg_partial_coverage). | #178 |
partial_coverage | gold.world() / crypto | Series covers below the warn threshold of expected trading days or the requested range (accepted-but-gappy). | #169 |
mixed_source | fundamentals.metrics | Metric inputs span more than one source. | #157 |
skipped_mismatched_report_rows | fundamentals.get_financials | Provider statement rows whose reportType/modelType did not match the requested period/template were dropped (contract violation, never a silent drop). | #44 |
skipped_period_rows | fundamentals.get_financials | Provider period rows that could not be mapped to the requested cadence were dropped (CafeF; never a silent drop). | #45 |
traded_value_estimated_from_close_x_volume | liquidity | Daily traded value estimated as close × volume (not provider-published turnover). | #146 |
zero_liquidity | liquidity | Average daily traded value is zero over the requested window. | #146 |
series_end_gap | macro.get_indicator | Latest monthly observation lags the series' own cadence (possible staleness/discontinuation). | #179 |
imf_weo | macro.get_indicator | Years ≥ the projection year are WEO forecasts (excluded from latest()). | — |
failover | macro.get_indicator | Result required failover across sources (carries the per-source note). | — |
nav_end_gap | funds.nav_history | Latest fund NAV is older than the fund's own trailing cadence allows (stale / paused / dormant feed). | #172 |
deduped_duplicate_nav_rows | funds.nav_history | Identical-value duplicate navDate rows collapsed to one (kept once + warned); a conflicting same-date NAV is quarantined (see quarantined_conflicting_navdates). | #158 |
quarantined_conflicting_navdates | funds.nav_history | Same-date NAV conflict(s) quarantined — two different NAV values for one navDate; that date is dropped entirely (never picked, never averaged) and the rest of the series is served. Names the dropped dates. A systematically-conflicting feed (>10% of in-window dates, floor 3) still raises InvalidData. | #194 |
partial_universe_coverage | equities.universe | Always present per board — the universe is index-basket-derived (~96% of the full SSC roster), not the complete listing. | #167 |
listing_date_not_available | equities.universe | Always present per board — the provider's firstTradingDate is '0' (unusable), so no listing date is exposed. | #167 |
sector_not_available | equities.universe | Always present per board on the plain universe() path — sector/industry is absent from this payload (not fabricated). Replaced by sector_partial_coverage when sector data is derived (with_sector=True). | #167 |
sector_partial_coverage | equities.profile / by_sector / universe(with_sector) | Derived GICS sector is HOSE-only (~74%); unmapped HOSE + all HNX/UPCoM → null, never fabricated; also flags any multi-basket symbol | #195 |
cross_board_duplicate_symbol | equities.universe | On an exchange=None merge, a symbol seen on more than one board is kept-first (board order HOSE, HNX, UPCOM) and the dropped copy is disclosed (never silent). | #167 |
board_unavailable | equities.universe | On an exchange=None merge, one board's fetch failed (SourceUnavailable/EmptyData/InvalidData); it is skipped, not fatal — the other boards still merge, and the skip is disclosed (board_unavailable: {board} — fetch skipped ({ExcType}): {reason}). If all boards fail the merge re-raises. A single-board universe("HNX") still raises (merge-only skip). | #189 |
fund_nav_stale | funds.list_funds | List-level: ≥1 listed fund's own nav_as_of is older than 7 calendar days (stale NAV feed); enumerates the stale codes@date, capped at 5 + +M more. Funds with unknown nav_as_of are never flagged. | #190 |
fund_missing_fees | funds.list_funds | List-level: ≥1 listed fund has no disclosed management fee (management_fee_pct=None — the provider lists managementFee on equity rows only), so an absent fee is never mistaken for a zero fee; enumerates the affected codes, capped at 5 + +M more. Suppressed when include_metadata=False. | #155 |
fund_partial_holdings | funds.asset_allocation | Detail-doc: the disclosed top holdings (equity + bond) sum to less than 50% of NAV, i.e. substantial portfolio exposure is undisclosed (top-N disclosure, not the full book). A bounded false-positive (a genuinely concentrated fund may trip it) is preferred over a false-negative that hides an opaque book. | #155 |
ex_date_unavailable | corp_actions.dividends | Always present per event — v1 serves the VSDC depository spine, which publishes no ex-date (the VNDirect finfo enrichment leg is held for v2), so ex_date is always None and is never fabricated or derived. | #163 |
corp_action_source_partial | corp_actions.dividends | Always present per result (DividendHistory) — the result is from the VSDC depository spine ALONE; the ex-date enrichment leg is not active in v1 (in v2 it fires when the finfo leg is down). | #163 |
vsdc_parse_degraded | corp_actions.dividends | Per event: a page IS a cash dividend (title/reason) but the parse is not fully trustworthy — ≥1 primary field unparseable (the record date, or both ratio_pct AND cash_per_share); OR the cash↔ratio pairing could not be trusted so ratio_pct was left None rather than mis-paired. This covers an implausible > 100% ratio; a Mệnh giá cross-check that confirms more than one distinct candidate (ambiguous twins, e.g. a gross + a "sau điều chỉnh" adjusted rate); OR a ratio stated on a different <br /> line than the cash anchor (v1 does not pair across lines, so the unpaired ratio is dropped, never silently). Ratio percentages are parsed decimal-aware (8.5%/8,5% → 8.5, never 85). OR the page listed multiple cash tranches — in EITHER phrasing (…được nhận X đồng or …số tiền X đồng/cổ phiếu) — and only the first is surfaced in v1. The affected fields are left None; the event is surfaced, never silently dropped or fabricated; an undated event is windowed by its pay_date. A net-of-tax ratio line is NOT this token — see vsdc_ratio_tax_deferred. | #163 |
vsdc_ratio_tax_deferred | corp_actions.dividends | Per event: the ratio line carries a tax/withholding signal (thuế / TNCN / khấu trừ) so its % is net-vs-gross ambiguous; under the #163 v1 de-scope the library no longer classifies net-vs-gross (an open-ended, silent-wrong-prone problem), so the ratio is withheld (ratio_pct=None) and disclosed via this token (net-vs-gross classification deferred to v2). Distinct from vsdc_parse_degraded (a parse fault): here the line parsed fine but is intentionally withheld. A single-tranche tax line carries this token only; a multi-tranche tax line carries both. | #163 |
coverage_truncated_at_max_fetch | corp_actions.dividends | Per result (DividendHistory): the bounded same-org crawl stopped at max_fetch with announcements still un-fetched, so the returned history is NOT exhaustive — raise max_fetch (or pass a seed_id) for fuller coverage. Absent when the crawl exhausts the frontier. | #163 |
corp_action_fetch_incomplete | corp_actions.dividends | Per result (DividendHistory): ≥1 same-org announcement page failed to fetch (SourceError) or to parse (InvalidData — empty/whitespace body or no resolvable ticker) during the crawl, so its (and any onward-linked) events are absent — the history may be incomplete. The crawl tolerates per-page failures rather than aborting, but discloses the gap (never silent). Suffix: : {n} announcement page(s) skipped (fetch or parse failed). | #163 |
corp_action_seed_not_found | corp_actions.dividends | Per result (DividendHistory): no-seed auto-discovery scanned its recent-ID window WITHOUT finding any announcement page for the ticker, so the empty history is NOT a confirmed never-paid — the discovery window may be too small/recent. Pass a seed_id (or widen the window) for an authoritative result. Absent whenever a seed was found or supplied. | #163 |
