NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI
PyPI, #4016 most downloaded on PyPI
Python library to access and analyze SEC Edgar filings, XBRL financial statements, 10-K, 10-Q, and 8-K reports
Latest stable
6.2.0
Last release 7 days ago
15 Nov 2026
Ships on a steady schedule
a new release about every 1 weeks
Nearly every release is documented
notes for 60 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
4 years old
455 releases · first in 2022
Drag across the ruler to compare, or tap a release to open it.
ParserConfig(extract_images=...) is deprecated; see the 6.1 upgrade notes .
pip install --upgrade edgartools6.2 is a minor release. Two fixes return different values than 6.1 did: get_operating_cash_flow() returns None instead of the continuing-operations figure when discontinued operations also report operating cash, and a reported zero shows as 0 where it showed blank. ParserConfig(extract_images=...) is deprecated; see the 6.1 upgrade notes.
Please report anything that breaks at https://github.com/dgunning/edgartools/issues.
viewer.compare_statements(xbrl) checks the statements edgartools builds against the SEC's statement pages. Each page row and cell is compared with what xbrl.statements (and so get_financials()) shows, in the DataFrame and the rendered table, with dropped rows, wrong values, signs and crashes counted separately. On Brixmor's FY2020 cash flow it reports beginning cash 360,073,000 where the page shows 21,507. (bead edgartools-auvk1.10)NetRevenueAfterInterestExpense.report.section(item) says how reliably an item was located: it returns the Section that report[item] reads, with .confidence, .warnings and .detection_method; strict=True raises the new LowConfidenceSectionError instead of returning a doubtful one. On 10-K, 10-Q, 8-K, 20-F and 40-F. NVIDIA's FY2025 Item 8, a 207-character pointer to Item 15, reports confidence 0.5 with an incorporation-by-reference warning.XBRL.element_metadata(concept).enumeration gives an enumeration concept's declared domain, linkrole, headUsable flag and set/single kind, from the filing schema or the FASB release the filing imports, and XBRL.enumeration_members(value) splits a fact value into namespace, local name and concept id. All 237 enumeration facts in 15 tracked fixture filings now resolve, with all 260 members mapped to a concept.XBRL.element_metadata() says what a concept is. It returns a concept's type, balance, period type, abstract flag, FASB definition and ASC references from the filing's schema, the bundled FASB table and the filing's MetaLinks.json, which it downloads lazily. Oracle's 10-Q 0001193125-26-389274 gives IncreaseDecreaseInContractWithCustomerLiability as debit, with ASC 912-310-45-11 and 230-10-45-28. (GH #1426)viewer.compare() pairs every numeric cell of the SEC statement pages with exactly one XBRL fact. Sign, dimensions, units and equity statements are checked, the tolerance is half the last displayed digit, and a cell the page does not identify is not_comparable with a reason, with coverage reported beside match_rate. Airbnb's FY2024 10-K went from 141 of 190 values matched to 278 of 278 cells. (bead edgartools-auvk1.8)ParserConfig(extract_images=...) is deprecated. The field defaulted to False and was documented as switching image extraction, but nothing read it, so every <img> became an ImageNode regardless. Passing it now raises a DeprecationWarning; it is removed in 7.0 and parsing output is unchanged. (bead edgartools-ru9x)to_dataframe(); the SEC page shows 0 for both. On an 82-filing sample, blank zeros fell from 840 cells to 0 and negative zeros from 185 to 0.doc.sections['part_i'] on a 20-F returned only its divider line ('PART I') when the Part and its first item header scored the same heading level; a Part now runs to the next Part divider or SIGNATURES. Nine of fifteen corpus 20-Fs were affected, e.g. 0001213900-21-018959 part_i grows from 6 to 604,618 characters.Statement.to_dataframe() crashed under pandas 3 on a statement where no row has a display sign. Applying signs assigned an empty object slice into a float column, raising Invalid value '[]' for dtype 'float64', so walking Keysight's FY2020 10-K notes failed on its land-sale details role. Signs are now read as numbers; the filing's cash flow still shows accounts receivable as 75 (millions). (GH #1960)balance_sheet() returned a one-row stub for some filers in 6.1.0. A filing's statement-typed role holding only a text block was accepted before the real balance sheet was reached, so Mettler-Toledo's FY2021-2023 10-Ks returned the Schedule II text block and no values. The real statement now wins: MTD's FY2022 total assets read 3,492,395K again. (GH #1970)or a heading), and context alone could promote No or a bold word inside a sentence. Over the 128-filing parser corpus about 4,100 such headings are gone and no section moved. (bead edgartools-s4cg)Item 16. A. returned none of them as sections. Section patterns expected the letter attached (Item 16A), so 20-F 0001062993-16-008650 lost all eight of Items 16A-16H; all eight now resolve, and the same spelling is accepted for every lettered 10-K, 10-Q and 20-F item. (bead edgartools-ga4a)PortfolioInvestments.from_statement() disagreed with from_xbrl() on the same filing. It re-parsed the statement's labels without member candidates or from_xbrl's exclusions, so on 38 BDC 10-K and 10-Q filings it returned no rows on 11 (Hercules, Sixth Street, Crescent among them) and differed in rows or names on the other 27. It now reads the statement's filing through from_xbrl. (GH #1612)Line one<br>line two broke its row over three lines and left the value after it in no column. Both formats now collapse a cell to one line, as the pipe format already did. (bead edgartools-grn9e)<th> and <td> came out with their cells reordered. Every td was emitted before every th, so a row written Label | Revenue | 100 parsed as Revenue | 100 | Label, moving the row label to the last column. Cells now keep document order. (bead edgartools-j99a)<tfoot> table without <tbody> read its total rows twice (once as body, once as footer); one with <tbody> lost them from markdown, text() and to_dataframe(), which never read the footer. Footer rows now appear exactly once, after the body, in every output. (bead edgartools-niru)tenk.notes["Leases"].details was empty and one lease table was typed as a cover page. A disclosure role with value line items and no text block is now read as a detail; FY2021 Leases has its four tables. (GH #1947)PROPERT<a>IES</a> heading read as PROPERT IES, and markdown space-joined every inline run: 2,519 of 17,394 multi-child paragraphs across the HTML fixtures had extra word breaks, now 139. Markdown now places gaps where text() does. (bead edgartools-8n9an)ViewerReport.currency_scaling returned a scale the SEC page does not use. It was inferred from XBRL decimals, which states rounding. It is now the scale each R page declares, for any currency symbol, and 1 when the page declares none: Airbnb's FY2024 statement of operations, shown in whole dollars, read 1,000,000 and now reads 1. parse_numeric also reads values such as ₨ 12,724 and R$ 88,242.5. (GH #807, bead edgartools-auvk1.8)get_concept('revenue') and get_concept('stockholders_equity') agree with the getters. The synonym groups lacked the IFRS, bank and utility totals that get_revenue() and get_shareholders_equity() read: TD's equity came back as its FY2019 figure (87.7B CAD against 125.6B), Novo Nordisk's equity and Goldman Sachs's revenue as None, and Duke's revenue as its contract-revenue slice.get_concept('net_income') returns profit attributable to owners for IFRS filers. Its synonym group reached ProfitLoss, which includes non-controlling interests, before ProfitLossAttributableToOwnersOfParent, so it disagreed with get_net_income() once that was fixed: TSM's 1,157.52B against 1,158.38B attributable to owners.EntityFacts getters read IFRS filers' own totals. get_shareholders_equity() and get_operating_income() now read the ifrs-full totals, and get_net_income() returns profit attributable to owners. Of 16 IFRS filers measured, 14 got None for equity and the other two (Toyota, Sony) got their last US-GAAP figure, years old; BHP's net income was 24% high. get_total_assets() no longer returns current assets.get_operating_cash_flow() returned operating cash of continuing operations as the total when discontinued operations also had operating cash. It now returns None for those filings (64 of the 95 CY2025 filers reporting continuing without a total had nonzero discontinued operating cash); get_standard_concept() now uses the section or statement for DividendsReceivable, DividendsPreferredStock and ImpairmentOfInvestments. (GH #1452)calculation_linkbase() read the bundled element table before guessing from the name, so headings the name guess missed are now abstract, like Microsoft's FY2024 AwardTypeAxis, which never reached its element catalog. Across the 24,439 bundled elements the guess was wrong for 17. (GH #1730)Your coding agent can read these notes before it upgrades. Set up the MCP server →
pip install --upgrade edgartools
pip install --upgrade edgartools6.1 is a minor release. A few fixes return different values than 6.0 did (for example str(ValidationResult) now prints INCONCLUSIVE when nothing was compared, and reporting_end_date is the document period end); see the 6.1 upgrade notes.
Please report anything that breaks at https://github.com/dgunning/edgartools/issues.
ValidationResult adds checks_completed, validation_complete and status ("valid", "inconclusive", "invalid"), and a skipped section check leaves a SECTION_NOT_EVALUATED note. str() now prints INCONCLUSIVE where it printed VALID: for every income and cash flow statement, and for 29 of 33 tracked balance sheets at level="sections". is_valid is unchanged. (GH #1210)parse_inline_xbrl_set() resolves inline XBRL facts across a filing's documents. An exhibit without its own ix:header, like IBM's FY2024 inline EX-13, now resolves contexts and units the primary document defines: all 3,983 of its facts get a context, where parse_html() on the exhibit alone resolves none. (GH #1753)get_financial_metrics(include_periods=True) reports the dates behind each metric. On a 10-Q, income covers the quarter while cash flow is year-to-date: Apple's Q3 FY2023 revenue is for 2023-04-02 to 2023-07-01, its free cash flow for 2022-09-25 to 2023-07-01. The keyword-only option adds a periods mapping; the default result is unchanged. (GH #1656)NetChangeInCashContinuingOps, NetChangeInCashDiscontinuedOps and their ExFX variants, components that never become NetChangeInCash: Conagra's FY2025 $-9.7M continuing and $-1.3M held-for-sale rows had no concept before, and its $-11.0M total keeps NetChangeInCash. (GH #1448)PortfolioInvestments.reported_industry_breakdown() returns the industry subtotals a BDC files even when no holding is tagged; Ares Capital's FY2025 10-K gives 23 industries summing to its $29,484.8M total. to_context() states "Industry known for X of N positions", and an empty industry filter warns. (GH #1784)get_ttm(), get_ttm_revenue() and get_ttm_net_income() take a filed_by cutoff. Only facts from filings made on or before that date are used, before any quarter is derived, and only a gap-free window qualifies: Apple's TTM revenue with filed_by="2023-08-03" is $385.1B to 2023-04-01, because the Q3 10-Q came a day later. as_of is unchanged. (GH #1699)IncomeTaxesPaidFederal, IncomeTaxesPaidStateLocalGross and IncomeTaxRefundsForeign, components of cash taxes paid and never tax expense, and EntityFacts cash flow statements now show them: ServiceNow's FY2025 $46M federal, $83M state and $154M foreign rows had no concept before. (GH #1448)XBRLFact gains fact_id and footnotes (footnote ids), and Document.xbrl_footnotes maps each id to the Footnote model XBRL.footnotes uses; over 35 tracked 10-K/10-Q fixtures 575 footnotes now reach 38,353 facts, where none did, matching the XBRL path. (GH #1756)xsi:nil. Fact.is_nil is carried into FactQuery/FactsView rows and frames (the last core column, nullable boolean) and XBRL.to_pandas()['facts'], so a nil item no longer looks like an empty one; Apple's FY2023 10-K files three, among them both CommitmentsAndContingencies facts. A nil/empty pair is now two DataFrame rows. (GH #1254)get_concept() reads cash interest and income taxes paid. Four groups, interest_paid, income_taxes_paid and their _gross variants, map one tag each and report basis ('net' or 'gross') in the metadata, with no fallback between bases. In the CY2025 frames 4,150 filers tag InterestPaidNet and 169 tag only InterestPaid, for whom interest_paid is None.CryptoAssetsNoncurrent and CryptoPurchases, each excluding crypto held for platform users, and EntityFacts statements now show them: Tesla's FY2025 $1,008M digital assets had no concept before. (GH #1447)discrete_quarters=True, Apple's FY2023 Q4 operating cash flow of $21,598M was keyed duration_2022-09-25_2023-09-30 and rendered "(YTD)"; it is now keyed by its own dates, duration_2023-07-02_2023-09-30, and merges with any three-month facts the filing reports for that quarter instead of showing it twice. (GH #1697)EntityFiling.acceptance_datetime is renamed reported_acceptance_datetime. The old name warns and goes in 7.0; it is SEC's submissions-JSON value, which was 4 h late for Apple's 10-Q 0000320193-26-000020 in October 2026, so use filing.accepted_at where the exact moment matters. (GH #1501)entity_info['reporting_end_date'] was the latest context date, not the period end. It is now dei:DocumentPeriodEndDate as a date, and the old value moves to max_instant_context_date; it changed on 39 of 43 tracked instances, e.g. American Airlines' Q2 2026 10-Q read 2027-07-17, its mistyped share-count date, and now reads 2026-06-30. (GH #1208)entity_info['ticker'] named the cover's last security, often a note or ETN. It is now the filer's first dei:TradingSymbol, and entity_info['trading_symbols'] / TenK.trading_symbols list every symbol with its class-of-stock member: JPMorgan's FY2023 10-K reads JPM, not AMJB, with all ten securities. (GH #1252)balance. Balances for standard concepts now come from a bundled table of the FASB us-gaap 2011-2026 and srt 2018-2026 element declarations instead of a hand-written list of 124 concepts; in Comcast's 10-Q 0001628280-26-049360, USD us-gaap concepts without a balance drop from 87 of 135 to 1. (GH #1728)0.00, and its dilutive shares rendered 68.32; each cell is now formatted from its own XBRL unit, so they show 0.147 and 68,316. DataFrame values are unchanged. (GH #1737)EarningsRelease misread footnoted EPS and table units. TSM's 3Q2026 6-K (0001046179-26-000726) EPS now parses as NT$29.43 from its raised letter footnote. Bare EPS rows and % change columns stay unscaled for every filer. Tables carry their declared currency (Toyota's "Yen in millions" is JPY); a scale-only note keeps USD and an unmapped currency is None, never USD. Ambiguous plain suffixes like 1.2m remain text. (GH #1824)discrete_quarters=True, a three-month value a filing presents now wins over the derived one, and a derived quarter pairs the latest-filed cumulative value with the shorter period from the same or an earlier filing, so Microsoft's Q2 FY2026 investment gains read the filed $9,931M instead of $5,017M.check_period_data_quality() found no fact by its us-gaap:Assets-style id and read 0% essential coverage for every filing (0 of 13,058 lookups hit on 4 test filings). Every catalog lookup now uses one rule in edgar.xbrl.core. (bead edgartools-gtmsn, GH #1711)XBRL.to_pandas() omitted catalog metadata from the facts frame. Colon-form ids such as aubn:EmployeeRetentionTaxCredit missed the underscore-keyed catalog, so Auburn National's FY2023 10-K (0001193125-24-067944) exported 1,844 facts with none of the five element_* columns. That credit stays the filed 0 and -1,569,000, now with its label. The lookup tries the exact key, then only the namespace separator. (GH #1711)clean_industry_label now returns every member joined with '; ' (any URI scheme), such a holding has sector None and never donates its industry to a peer. Measured on DERA 2025Q3/Q4: 0 of 2,752 distinct industry values change. (GH #1712)xsi:nil read as the element name {…}QName.domain. The member is now '' with the axis kept, Context.nil_dimensions names the axis, and labels show (nil), as the SEC's rendering does; BXP's Q2 2026 10-Q row reads "Unconsolidated Joint Ventures - (nil)" (-$900,000), apart from the unqualified $29,447,000. (GH #1751)xsi:nil state, and a nil fact could be given a value. XBRLFact.is_nil is read by namespace, so AAR's xs:nil preferred-stock facts count; 837 facts across 129 tracked inline fixtures are now marked nil. A nil fact is no longer transformed, scaled or signed: one with ixt:fixed-zero and sign="-" read as -0. (GH #1757)get_revenue() returned None for BDCs. Total investment income is now Revenue, total expenses the new InvestmentExpenses (was nonoperating income) and net investment income the new NetInvestmentIncome; Main Street's FY2025 revenue is $566.4M instead of None, with its $366.7M net investment income no longer called revenue. Insurers' revenue is unchanged. (bead edgartools-k7mi8.6, GH #1842)extract_nonaccrual() dropped a non-accrual flag linked to a holding's rate or principal instead of its fair value. The flag now reads that holding's fair value at the same date: New Mountain Finance's 10-Q for 2026-09-30 reads 9 holdings, 1.75%, where it read none. (bead edgartools-jv8b8)extract_nonaccrual() read a BDC that tags zero non-accruals as unknown. A zero amount or count of non-accrual investments, with no non-accrual footnote, now gives extraction_method="tagged_zero", a rate of 0.0 and num_nonaccrual 0; 8 of 179 active BDCs, Stone Point Credit among them, read this way. (bead edgartools-rol2k)extract_nonaccrual() read no non-accruals from BlackRock's three BDCs. Their schedules mark holdings "Non-accruing debt investment.", which the non-accrual pattern did not match, so every filing read extraction_method="none". BlackRock TCP Capital's 10-Q for 2026-09-30 now reads 17 holdings, 3.27% at fair value. (bead edgartools-tfhpl)to_dataframe(include_unit=True) labelled Microsoft's FY2024 basic EPS (USD per share) as shares and revenue as u_usd; they are now usdPerShare and usd, a row whose periods differ shows mixed, and an unresolved unit shows unit unavailable. Parsed units gain a definition of expanded-QName measures. 1,514 of 13,747 row labels across the tracked fixtures are corrected. (GH #1746)footnoteArcs with another arcrole are no longer read as ordinary footnotes. XBRL.from_files() attached every arc that reached a link:footnote, so a custom-arcrole association showed up in Fact.footnotes; only fact-footnote arcs attach now, as the inline readers already required. 0 of 49,267 arcs in the tracked fixtures change; fact-to-fact arcs are still not modelled. (GH #1767)edgar.xbrl.core.format_value() no longer returns an invisible control character for zero. A zero is now formatted like any other number ($0, or 0.00 when decimals call for it) instead of U+0014; statement cells keep showing a filed zero as blank.to_dataframe() keeps each occurrence's presentation parent — a concept presented under two parents no longer reports the first one on both rows, and its dimension rows follow the line they break down. Cato's FY2023 income statement now places the second net loss row (−$23.9M) under the comprehensive income section; 350 rows in 24 of 2,793 fixture statements change.Rollforward spelling and every ...Member, so 1,674 of the 5,369 abstract elements in us-gaap-2025 read abstract=False; it now misses 10, with no false positives. abstract="1" is read as true, and a declaration the parser read is no longer overridden by the name. (GH #1730)us-gaap:GrossInvestmentIncomeOperating mapped to NonoperatingIncomeExpense, so Barings BDC's $65.2M Q2 2026 total investment income was labelled nonoperating; the mapping is removed and the row keeps its filed label and amounts with no standard concept. (GH #1842)ifrs-full:DividendsPaid was classed as a cash flow, so SAP's FY2025 row read EUR 2,746M instead of the filed EUR 2,743M. The row now carries the parent-shareholder cash line, dividends to non-controlling interests get their own row, and the equity total stays out of the cash-flow view. (GH #1679)2023-07-01. to_dataframe(period_columns="interval") now names each column by its exact dates (2022-09-25 to 2023-07-01); the default names are unchanged. (GH #1741)edgar on one thread while another imported a submodule failed in both threads. import edgar now imports no submodule, and each top-level name such as edgar.Company loads its module on first use, so the reporter's script, which failed in 3 of 3 runs, now succeeds. (GH #1325)filing.accepted_at gives the exact acceptance time; the submissions-JSON value could be hours off. get_filings() read Apple's 10-Q 0000320193-26-000020 as 14:01:02 UTC for weeks, against a true 10:01:02. accepted_at reads the filing's 3-20 KB index header as a tz-aware America/New_York datetime, and filings.accepted_at() reads a whole set. (GH #1501)search_filings() results carried SEC's display decoration in company. NVIDIA's 10-K 0001045810-24-000029 came back as "NVIDIA CORP (NVDA) (CIK 0001045810)"; it is now "NVIDIA CORP", the same name find() resolves through the accession lookup, with the CIK still in cik. (GH #1708)MultiFinancials.income_statement() and cash_flow_statement() now accept include_quarterly. The keyword raised TypeError although the stitch underneath supported it; with include_quarterly=True, Apple's 10-Q 0000320193-23-000077 adds a $81,797M quarter column next to its $293,787M year-to-date revenue. The default stays year-to-date only. (GH #1783)xbrldi:typedMember now reports is_typed_dimension=True. Only a typed dimension may carry a typedMember, so the instance settles the kind of an axis whose declaration sits in an unloaded US-GAAP schema; 12 such axes across the tracked fixtures flip from False. typed_domain_ref stays empty until imported schemas are read.axes_for_role(role)[axis].default_member_id read the default only from the hypercube's own extended link, although a dimension-default arc is global; Microsoft's FY2024 10-K had 48 axis-role pairs reporting None that now agree with xbrl.axes.xbrli:measure children form a product (feet x feet) was stored with only its first measure, so USD x shares read as a plain dollar amount; it is now {'type': 'product', 'measures': [...]} and is no longer classed as a currency. None of the 269 units in the tracked fixtures is a product, so their parsed units are unchanged.xbrldt:targetRole keep their nested members. A cross-role domain copy took only the root's direct members, and a domain-member arc carrying targetRole was not followed, so a branch's children in the target role were missing from domains_for_role(); the issue's branch now lists its 1 leaf member instead of none.10:30 was rewritten to 10_30 as if it were a concept name; nested content is now kept as filed and non-QName typed values are no longer rewritten. Dimension keys in the 15 tracked fixture directories with typed members are unchanged.Middle, SCSP and Nordic Ferry Infrastructure AS for five holdings ($52.3M of fair value); they now read None, and across 39 BDC 10-Ks only those and two OBDC SCSp rows change. (GH #1669)Prospectus497K.tickers reads the cover when the SGML header omits the ticker. Ocean Park High Income ETF (0001580642-24-004234) now returns ['DUKH'] instead of []; in 250 sampled 2024Q3 497Ks the cover ticker matched the header on all 64 single-class filings that carry one, and filled 5 of 29 that do not. (GH #912)fiscal_period. xbrl().facts left it empty on every instant fact, so get_facts_by_fiscal_period() returned no balance-sheet values. A fiscal year end is now FY and another quarter end is that quarter; Microsoft's Q1 FY2027 10-Q query returns 701 facts instead of 290. Dates that close no period stay empty.td class="pl" label plus a value) were treated as column headers when they read like periods, so Apple's 10-Q cover page (R1.htm) lost "Document Period End Date | Mar. 29, 2025" and two other facts from its body and Markdown. They now stay body rows; only that 1 of 381 Apple R-report tables changes. (GH #1477)display:none cells counted toward a row's width, so JPMorgan's FY2024 10-K reported its 2023 managed-basis overhead ratio (54%) under 2022. They now take no grid slot, which corrects period labels in 220 of 12,948 tables across the HTML test corpus. (GH #1847)ixt-sec:exchnameen returned exchange codes outside the dei:SecurityExchangeName enumeration. "NYSE Texas" gave NYSETXSE and "NYSE Chicago" NYSECHX; they now give the dei 2026 codes NYSETX and CHX, and the Long-Term Stock Exchange and MEMX, which have no dei code, are rejected instead of mapped to invented ones. All 330 exchange facts in the fixture corpus are unchanged. (GH #1726)format names now resolve in the registry their prefix is declared for. A format matched its local name under any namespace, so TR3's nonexistent fixed-zero counted as known, and TR4's date-month-year/date-year-month, TR5's num-*-decimal-apos and five ixt-sec rules were missing; ixt:date-month-year over 12/2025 now gives 2025-12. Non-English date rules remain unimplemented. (GH #1743)ix:continuation, splitting a token the filer had split across elements. Cooper's address line came out as 6 101 Bollinger Canyon Road; the chain is now joined with nothing inserted, as Inline XBRL 1.1 and the SEC's extracted instances do. 5 of 247,583 facts in the fixture corpus change, each now equal to the SEC instance. (GH #1748)P9.5Y, which are not valid xs:duration. ixt-sec:duryear and its siblings now spill the fraction into smaller units as the SEC does (9.5 years is P9Y6M), and durwordsen no longer skips words it cannot read, which turned Amazon's "twenty-five years" into P5Y. 570 of 2,441 duration facts in the fixture corpus change. (GH #1754)ix: tags and ix:exclude page furniture. Unwrapping and exclusion applied only to a fact's direct children; they now apply at every depth, on a copy of the tree. In the fixture corpus 5,140 escaped facts carried <ix: markup and none do now, and all 92 of Microsoft's escaped facts match the SEC instance word for word. (GH #1770)link:definition; calculation and definition roles were named by the role URI's last segment. NIKE's FY2026 income-statement calculation role now reads "9952151 - Statement - CONSOLIDATED STATEMENTS OF INCOME" instead of "CONSOLIDATEDSTATEMENTSOFINCOME"; a role with no definition still falls back to the URI. (GH #1731)use and priority, so a use="prohibited" arc added the relationship it was meant to remove; they now apply XBRL 2.1 section 3.5.3.9.7 through one shared resolver. Travelers' Q2 2009 10-Q/A prohibits 1,466 template presentation arcs, and its balance sheet is now its own 36 rows instead of a 245-row template. (GH #1736)CalculationArc.arcrole records which one each edge belongs to; NIKE's FY2026 10-K files all 161 of its calculation arcs under the 2023 arcrole. (GH #1742)PresentationTree.relationships holds every effective relationship as filed, while statement rows stay one per concept. Coca-Cola's FY2011 debt details present long-term debt at order 21 (total label) and order 30 (verbose label), and both are now kept. (GH #1758)us-gaap:StatementTable, so its catalog type was None, Statements.get("IncomeStatement") returned a segment disclosure and get_statement_facts returned 0 rows; it now returns 110, including revenues of $1,492,700,000. (GH #1838)EntityFacts.time_series() kept duration_days only when rows matched. An absent concept or periods=0 returned five columns, so reading duration_days raised KeyError. Empty results now have the same six documented columns as a populated series, with zero rows. (GH #1709)business_category no longer calls Netflix an ETF. The ETF name check was a substring test, so NETFLIX, JETFLEET and MARKETFIELD were labelled ETFs; "ETF" now has to start or end a word. 60 of the 391 SEC company names containing those letters change, all out of ETF.validate_balance_sheet() now finds Assets, Liabilities, equity and their combined total by XBRL concept before label. It had returned INCOMPLETE_DATA for 18 of 33 tracked fixtures, among them Apple, Microsoft and JPMorgan; now 32 balance and one reports a genuine gap.Statement.validate(level="detalied") ran only the fundamental equation check and reported success. It, validate_statement() and validate_balance_sheet() now raise ValidationError naming the 3 valid levels; the levels stay case-insensitive and accept ValidationLevel members.Statement.validate(level="detailed") was documented as line-item rollup validation it never ran. The docstrings now say detailed runs the same two checks as sections (on IBM's FY2024 10-K both levels return the identical result, $137,175M of assets balancing); the argument and results are unchanged. (GH #1739)viewer.validate() skipped a total when one calculation child had no value. A nil or unreported child is now left out of the sum, as XBRL 2.1 calculations do, and listed in the new missing_children key, so Apple's 10-Q balance-sheet identity ($331,233M) is checked where it used to be missing from the results. (GH #1740)validate() compared totals from different dates. Each operand took its row's first non-null value, so National Presto's Q1 2026 10-Q compared April 5 assets with December 31 liabilities and equity and reported a $31,418,000 imbalance. Every operand now comes from one column, named in metadata["period"], and an equation that column cannot complete is reported as INCOMPLETE_DATA. (GH #1852)extract_filing_fees_xbrl ignored inline-XBRL transformations on totals and the final-prospectus flag. Welltower's ixt:fixed-zero totals came back as an em dash instead of '0', and NextEra's ixt:booleantrue flag (accession 0001193125-25-338333) read False. Digit-free amounts now take the transformed value and the flag is read as an xs:boolean; amounts with digits keep their display text. (GH #1681)AuditorInfo.icfr_attestation read a valid 1 as False. The DEI flag is an xs:boolean with two true forms, true and 1, and only true was recognized. Both now read True, and false/0 False, through a shared edgar.xmltools.parse_xml_boolean. (GH #1724)Company.income_statement(), balance_sheet() and cash_flow_statement() returned None for an invalid argument. period="quarter" gave None from all 3 methods while company.facts raised. They now raise ValidationError (IS-A ValueError), including for the balance sheet, and let statement-builder errors propagate; None still means SEC has no company facts. (GH #1732)Document.to_dataframe() raised InvalidIndexError when a duplicate-header suffix collided with a literal header. Headers Revenue, Revenue, Revenue.1 gave two Revenue.1 columns; suffixes now skip every name already present (Revenue, Revenue.2, Revenue.1), and a header named like a metadata column is kept. Column names are unchanged on 1,169 tables from 4 real 10-Ks. (GH #1734)2024-06-30 and 2025-01-01T00:00:00 were keyed as raw text and dropped from reporting_periods; both are now read as the dates they equal, so IBM's FY2024 instance respelled this way still shows total assets of $137,175M at 2024-12-31. (GH #1762)Fact.precision, and a fact of unknown accuracy ranks below any stated one; IBM's FY2024 total assets restated with a two-digit duplicate showed $140,000M instead of the filed $137,175M. (GH #1765)decimal.localcontext(prec=8), parse_html() read IBM's filed 927,264,332 shares outstanding (10-K, 0000051143-25-000015) as 927,264,330; scaling now moves the exponent and negation flips the sign, so neither consults the context. (GH #1854)use_optimal_periods=False, discrete_quarters=True, Apple's Q3 2026 operating cash flow of $34,369M was headed Q3 Nine Months: September 28, 2025 to June 27, 2026; it is now Q3 Quarterly: March 29, 2026 to June 27, 2026. (GH #1697)include_quarterly=True, period_sources counted selected periods instead of filings, so Oracle's six-month revenue of $27,366M was credited to the prior 10-Q and its Q1 to an index past the end of a two-filing list. Each period now names the filing it came from. (GH #1837)cashflow_statement() and cash_flow() removed. The 5.x deprecated spellings are gone from Company , Financials , EntityFacts , Statements and CurrentPe…
pip install --upgrade edgartools6.0 is a major release: Python 3.11 or newer is required, pandas 2.2.2 or newer is required, edgar.files is removed, and a number of fixes return different values. Read the 6.0 upgrade guide before upgrading.
Please report anything that breaks at https://github.com/dgunning/edgartools/issues.
get_facts() now includes the predecessor CIK's history, and Company.predecessor reaches it directly. Company("XOM") annual income goes from empty to FY2016–FY2025 (FY2025 revenue $332,238 million) (GH #1821).EFTSResult.accession_no and EFTSResult.filing_date. Full-text search results now answer to the names Filing uses, as read-only aliases of accession_number and a datetime.date parsed from filed. In the README agent eval, every API error on the full-text task (6 of 6) was one of these two names.to_pandas() on EntityFacts, FactsView and both FactQuery classes. It is an alias of to_dataframe() with the same arguments, matching Filings and XBRL. company.get_facts().to_pandas() was the most-invented call in the README agent eval (8 of 19 failing scripts), and the quick guide documented it.edgar_read gained an opt-in limit for 13F-HR holdings. The section stopped at 30 rows and a 6,000-character cap, so a full portfolio could not be read or reconciled to its filed total. limit sets the row count, limit=0 returns every row without the cap, and the default stays 30. (GH #1416)unidecode, or textdistance. Company and BDC name search now transliterate with anyascii (ISC), and SimilaritySearchIndex scores with rapidfuzz's Jaro. Search output is unchanged: the two transliterators agreed on all 141,949 bundled company and ticker names, and the Jaro scores matched exactly on 220,005 pairs. (GH #934, bead edgartools-07lk.7)company_tickers.parquet goes from 10,365 to 10,443 rows, and 10 tickers that moved to a new registrant now resolve to it offline too, among them XOM (ExxonMobil Holdings Corp, CIK 2115436, was 34088). Online lookups already read SEC's live file first (#1588).pyrate-limiter>=4.0. It already did in practice, through httpxthrottlecache>=0.6.1; the declared floor was still 3.0, and the 3.x compatibility shim behind it is removed. A burst past SEC's 9 requests per second still waits rather than fails.None, not 0. ReportingPerson's ownership figures and power totals, total_shares/total_percent, and OwnershipComparison.shares_change, percent_change and get_summary() no longer read a missing figure as 0, so a 13D/A omitting the Aadi Bioscience share counts is no longer a 2,435,000-share sale; a reported 0 stays 0, and the 5.x FutureWarning is gone. (bead edgartools-u6tkw)CurrentReport fixture docstring did not explain the SEC-injected script. The 8-K fixture for accession 0000710782-24-000005 retains one bot-defense script alongside its Item 5.02 text; the docstring now explains this response content. The fixture, assertions and library behavior are unchanged. (GH #1589)NaN, not 0.0. Report.to_dataframe() reported every blank cell as zero, which was 45% of parsed cells across six filers' 10-K statements. A blank, N/A or NM is now NaN, and a dash still reads as 0.0. Use .fillna(0.0) for the old frame.FactQuery.to_dataframe() now gives preferred_sign and fiscal_year the same dtype on every query. They are nullable Int64, so an absent value is pd.NA, not NaN. In 5.x they were inferred per query: on Foot Locker's FY2024 10-K, preferred_sign read float64 on the bare query and int64 on the income statement. is_dimensioned is now boolean. Use pd.isna(), not np.isnan(). (GH #1100, bead edgartools-rsyt.2)fiscal_year, is_audited and is_estimated nullable dtypes. company.get_facts().query().to_dataframe() and EntityFacts.to_dataframe() now return Int64 and boolean, the same on every query, where 5.x returned int64/bool but switched to float64/boolean when no returned row had a value. Use pd.isna() to test absence. (bead edgartools-7wtj.1)None (breaking). report["Item 99"], find() on a malformed accession, obj() on a Form 3/4/5/D/EFFECT with no XML and a failed 10-K parse raise SectionNotFoundError, ValidationError, DataObjectError and ParsingError; network failures raise TransportError. get_financials() on a filing without XBRL returns None, not an empty Financials. EDGARTOOLS_STRICT_ERRORS is gone. (bead 07lk.10)search_filings() returned 0 results for a query EFTS rejected. EFTS answers a failed query with HTTP 200 and an errorMessage body, which was read as an empty search: search_filings("apple", start_date="2024-13-01") showed 0 hits. A malformed date now raises InvalidDateError before any request; any other rejected query raises TransportError with SEC's message, also in resolve_accession(). next() and fetch_more() stop at the 10,000-hit window EFTS pages through. (GH #1497)httpx2, the maintained fork, instead of the dormant httpx. Unwrapped network errors are httpx2 classes, so except httpx.ReadTimeout: stops catching them; catch TransportError. Certificates verify against the OS trust store, not certifi, and the logger is httpx2. All 263 cassette tests replay unchanged. See the 6.0 upgrade guide. (GH #842, bead edgartools-q2iz)Requires-Python: >=3.11, so pip install -U edgartools on 3.10 resolves to the last 5.x release instead of installing 6.0. The pandas floor is unchanged. See docs/upgrade/6.0.md. (GH #938, bead edgartools-89ra)edgar.offerings.* module paths are removed. The 16 stubs left when the offerings parsers moved into crowdfunding, exempt and prospectus in 5.40.0 never warned, and import edgar.offerings.formd now raises ModuleNotFoundError. from edgar.offerings import FormD and filing.obj() are unaffected; docs/upgrade/6.0.md has the path table and a one-line rewrite. (edgartools-07lk.22.2)edgar.documents.migration and edgar.documents.extractors.heading_section_detector are removed. Nothing in the library used either: the first imitated the old SECHTMLParser on top of the new parser, and the second was a section detector no code called. Use edgar.documents.HTMLParser or parse_html(); the 6.0 upgrade guide lists both.edgar.__all__ are private. They include set_data_directory, get_obj_info and stdlib names like Optional that leaked in through annotation imports. from edgar import X now raises ImportError, and edgar.X names the replacement. Import them from their modules (edgar.paths, edgar.entity, typing) or use Filing.obj_type.edgar.core no longer re-exports the settings and identity names. get_identity, set_identity, EdgarSettings, get_edgar_data_directory and the rest of edgar.settings import only from there (or from edgar for the names it exposes); from edgar.core import get_identity raises ImportError. (bead edgartools-07lk.22.1)include_dimensions is removed from statement rendering. The Financials and xbrl.statements statement methods, Statement.render() and Statement.to_dataframe() warned about it since 5.8.0 and now raise TypeError; use view='detailed' for True and view='standard' for False. CurrentPeriodView and stitched statements keep theirs. (GH #574, bead edgartools-07lk.22.1)cashflow_statement() and cash_flow() removed. The 5.x deprecated spellings are gone from Company, Financials, EntityFacts, Statements and CurrentPeriodView; calling them raises AttributeError. Use cash_flow_statement(), see the 6.0 upgrade guide. (bead edgartools-07lk.22.1)edgar.core.Result is removed. The success/failure wrapper was never returned by anything in edgartools, and from edgar.core import Result now raises ImportError. It was documented as deprecated but never warned. (bead edgartools-07lk.22.1)NORMAL, CAUTION, CRAWL and edgar_mode now raise ImportError, and EDGAR_ACCESS_MODE is no longer read or warned about; it never affected the HTTP client. Use EDGAR_RATE_LIMIT_PER_SEC and EDGAR_HTTP_TIMEOUT. (GH #1326, bead edgartools-07lk.22.1)period_length removed from the EntityFacts statements. income_statement() and cash_flow_statement() no longer accept period_length, by keyword or by position; passing it raises TypeError. Use period='quarterly' or period='annual' instead (bead edgartools-07lk.22.1).edgar.core no longer re-exports the date helpers. filing_date_to_year_quarters, current_year_and_quarter, is_start_of_quarter and parse_acceptance_datetime now import only from edgar.dates; the old path raises ImportError. (bead edgartools-07lk.22.1)edgar.effect and edgar.form144 are removed. They forwarded to edgar.offerings.effect and edgar.ownership.form144 with a DeprecationWarning since 5.55.0, and now raise ModuleNotFoundError. from edgar import Form144 and filing.obj() are unaffected. (bead edgartools-07lk.22.1)StatementNotFound, NoCompanyFactsFound, SECFilingNotFoundError, InvalidDateException, IdentityNotSetException, TooManyRequestsException and DataObjectException warned throughout 5.x and now raise ImportError. The replacements in edgar.exceptions have existed since 5.48.0. (GH #933, bead edgartools-07lk.10)beautifulsoup4 is no longer installed. Every parser runs on lxml, so 6.0 drops beautifulsoup4 and soupsieve from the install; no edgar module imports bs4. edgar.xmltools returns lxml elements only, and passing a BeautifulSoup to FilingHomepage or Attachments.load (deprecated through 5.x) now raises TypeError naming parse_homepage_html. See the 6.0 upgrade guide. (GH #931)edgar.files parser, ChunkedDocument and the page-break options are removed. 7,844 lines go; Attachment.text() and 8-K exhibit text, which still reached the old parser silently, now read the same as Filing.text(), losing no words over 82 real filings except page headers and long table-cell tails. See the 6.0 upgrade guide. (GH #930, #870, bead edgartools-07lk.3)extract_nonaccrual() labelled a result with another period than the filing's. With no investment facts at the period, it fell back to other dates in the filing: Ares Capital's cover-page 10-K/A for 2025-12-31 came back as 2025-06-30, its public-float date. A period given, or the filing's own period, is now read as given, and a filing with no data there returns no rate. (bead edgartools-a3m66)total_source='identifier_total'. (bead edgartools-cpd1b)PortfolioInvestments already applies. (bead edgartools-wg5n8)xbrli:fractionItemType) parsed as value='13' and numeric_value=13.0. Terminating quotients are now exact, and a recurring one such as one third is rounded half-even to 50 significant digits, while decimals stays INF (the source precision). (GH #1766)InvestmentIdentifierAxis. extract_nonaccrual() returned no rate for filers that name a holding by several axes; TriplePoint's FY2024 10-K now reads 8 non-accrual loans worth $20.6M, matching its footnote. Roll-ups linked to the same footnote are dropped, and a balance sheet split into affiliation lines gives the total. (bead edgartools-jv32z)xbrl.footnotes was empty when SEC's extracted instance dropped the footnote links. Footnotes are now read from the inline document's ix:footnote and ix:relationship elements, on first use and only when the instance has none. Great Elm Capital's FY2024 10-K (0000950170-25-036517) goes from 0 to 56 footnotes, and extract_nonaccrual now finds New Wilkie Energy's non-accrual loan ($1,322,000 fair value).extract_nonaccrual() added up every schedule row, including the "Total Investments" and lien or country subtotals many BDCs tag there; Main Street's FY2024 rate read 0.58% instead of 0.92%. The rate now uses the balance-sheet total, and NonAccrualResult.total_source says where the total came from. (bead edgartools-7vl3h)by_fiscal_year(2021) missed its $244.2M net income; Apple, Intel and Costco comparatives drifted the same way. Annual EntityFacts statements also lost SFM's FY2022, which ended 2023-01-01. (GH #1802)ProfitLoss to NetIncomeLoss and its lease lines from sfm: to us-gaap tags, so five years of net income ($244.2M to $523.7M) split across two rows. Rows with the same label and no period in common now merge when a filing's comparatives show both tags carrying the same nonzero amount. (GH #1802)edgar_ownership insiders dropped option exercises behind a run of sales. Each Form 4 entry kept its first five trades, and sales are listed first, so Tesla director Kathleen Wilson-Thompson's 0001104659-26-055079 lost its exercise of 40,948 options at $14.99 behind 16 sale lines. Entries now keep up to 25 trades and report transaction_total, transactions_truncated and form (4 or 4/A). (GH #1817)edgar_ownership insiders returned no insider, role or trades, and ignored days. Every entry held only a date and accession because the tool read Form 4 attributes that do not exist; it now reports the insider, title, officer/director flags and transactions (NVIDIA CFO Colette Kress, 0001588670-26-000014: open-market sale of 20,369 shares at $219.0172), and days bounds the filing date. (GH #1817)Company("EQR") returns CIK 906107 (now Vivmark Residential) instead of raising CompanyNotFoundError. A reviewed table answers only after every current ticker map misses, so a reused ticker still goes to its new owner (GH #1829).PortfolioInvestments.from_xbrl() kept subtotals that repeat their one holding. Since #1506 a costed row over a single extending row was always kept, so PhenixFIN's company subtotals counted as holdings (+$289.5M on its Q2 2026 10-Q) and New Mountain's FY2025 10-K counted ACI Parent's preferred twice. Such a row is now excluded when the row under it states the same cost, unless that row adds only a counter. (GH #1583)Revenue, not "Electric Utility Revenue". RegulatedAndUnregulatedOperatingRevenue is total operating revenue across electric, gas and water operations; it now maps to Revenue and is a listed member of that total. MGE Energy's Q2 2026 Total Operating Revenues of $161,195K, which includes $29,601K of gas revenue, was ElectricUtilityRevenue and is now Revenue. (#1844)InterestExpense, not "Operating Lease Expense". IFRS 16 makes lease-liability interest a finance cost, so InterestExpenseOnLeaseLiabilities now maps like the us-gaap FinanceLeaseInterestExpense. Opera's FY2025 20-F reports $461K of it inside finance expense; it was OperatingLeaseExpense and is now InterestExpense. (#1845)GrossPropertyPlantEquipment and accumulated depreciation to AccumulatedDepreciation. American Airlines' FY2025 balance sheet had three net-PP&E rows ($57,732M cost, $(25,192)M accumulated, $32,540M net); only the $32,540M row is net now. (#1856)IncomeTaxReconciliation* elements and four EffectiveIncomeTaxRateReconciliation*Amount items no longer map to IncomeTaxes; the provision keeps it. Apple's FY2023 reconciliation had six IncomeTaxes rows beside the $16,741M provision, including the $23,885M tax at the statutory rate; only the provision is IncomeTaxes now. (#1857)0001193125-13-060317) returned Sole 0 for American Express and filed Washington Post's None 148,311 under Sole. Both now land in the right column; 197 dash rows across 14 filings change and no other row does.compare_holdings() and holding_history() merged on CUSIP alone, so Bank of America's AFLAC (Q2 vs Q3 2024) returned 9 rows instead of 3, and 3M's new Call read as DECREASED. They now key on (Cusip, PutCall) and carry a PutCall column.PortfolioInvestments.to_dataframe() exported a disclosed 0 fair value and principal as NaN; GBDC's 10-Q for 2025-03-31 (0001476765-25-000040) holding "PPW Aero Buyer, Inc., One stop 3" now reads 0.0. An amount the filing omits is still NaN.% to the raw ratio, so Apple's FY2025 effective tax rate of 0.156 (10-K 0000320193-25-000079) displayed as 0.2%; it now displays as 15.6%. Only rows whose unit is pure are scaled, so a dollar row captioned "Gross margin" is unchanged. (GH #1850)pivot_by_period() dropped 13-week quarters of 52/53-week filers. Period labels counted the calendar months a period touched, so Apple's Q2 and Q3 FY2025 were both 6M 2025 and Q3 net income of $23,434,000,000 (10-Q 0000320193-25-000073) vanished. Durations are now classified by days, and two periods that still share a label get separate dated columns. (GH #1851)configure_http() coverage. Cleanup now restores the manager's prior proxy and closes its client. A real-manager probe restored state in 0/3 cases before the fix and 3/3 after, covering default/custom proxies and failed assertions. (GH #1827)TableNode.to_dataframe() and Cell.numeric_value turned "not meaningful" cells into a zero rate; the JPM FY2024 10-K Overhead ratio row now shows NaN for its NM adjustment. Dashes still read as 0.ixt-sec:exchnameen accepts all 21 exchange names in the SEC registry. Ten names, including Cboe BZX Exchange, Inc. in the Fidelity Wise Origin Bitcoin Fund FY2025 10-K, raised TransformError and left display text in Document.xbrl_facts; that fact now returns CboeBZX. NYSE Arca and MIAX PEARL now return the dei codes NYSEArca and PEARL.CostsAndExpenses as debit; they were the only 3 of its 124 entries that contradicted it. Apple's FY2023 facts no longer show a balance on shares outstanding or operating cash flow. Reported amounts and signs are unchanged. (GH #1728)ProfitLoss, leaving the noncontrolling interest as the only net income row. Mastercard, Caterpillar, Ford, American Tower and Simon Property Group (5 of 50 companies measured) now show it; Mastercard's FY2025 is the 10-K's $14,968M instead of nothing beside a $21M minority interest. (GH #1823)income_statement(period="ttm") raised "found 1 quarters" for e.l.f. Beauty although its revenue has dozens of quarters. Choosing the period columns tried every line item and let the first sparse one (2 facts) fail the whole statement; it now skips a concept it cannot use, as the line-item loop already did. ELF's TTM statement builds, and its fiscal Q4 2026 cost of sales TTM is the 10-K's 479,125,000. (GH #1822)edgar_read read no ownership or purpose section from a Schedule 13D or 13G filed since December 2024. The reader matched only the old SC 13D and SC 13G form names, and purpose returned the object summary instead of Item 4. On BML Investment Partners' SCHEDULE 13D (0001373604-24-000118) purpose now returns Item 4 and ownership reports 2,435,000 shares (9.9%). (GH #1820)XBRL.from_directory() was slow on filings with many footnote relationships. GBDC's March 2025 10-Q (0001476765-25-000040) has 48,629 footnote relationships and 21,393 stored facts. Attachment now indexes fact IDs once per extraction instead of scanning the facts for each relationship, preserving note order and associations. (GH #1771)FactsView.clear_cache() left retained query DataFrames stale. In staged loading of GBDC's March 2025 10-Q (0001476765-25-000040), a cached query omitted the $71,205,000 comparative interest-paid observation after a clear. Existing queries now discard obsolete frames when their own view is cleared. (GH #1811)StitchedFactQuery.to_dataframe() reused rows after period-filter changes. With original observations from GBDC's March 2025 10-Q (0001476765-25-000040), raising across_periods() from two to three retained two rows instead of none. Cache keys now include minimum-period and completeness settings. (GH #1812)StitchedFactQuery.sort_by() changed sibling queries' row order. In a controlled input from GBDC's March 2025 10-Q (0001476765-25-000040), sorting one query made an unrelated limit(1) select $71,205,000 instead of $133,282,000 of interest paid. Sorting now returns a separate list and preserves the owner's order. (GH #1813)edgar.xbrl.facts.FactQuery could disagree with its cached DataFrame after a period-list edit. Appending GBDC's comparative period added its $71,205,000 interest-paid fact to execute() but left it out of the warmed frame. by_period_keys() now captures its keys at call time, including on stitched queries; use a new query for an edited selection. (GBDC March 2025 10-Q, accession 0001476765-25-000040) (GH #1814)NetChangeInCash row; it is now ForeignExchangeEffectOnCash. The six ASC 842 lessee operating-lease maturity elements now feed the OperatingLeaseCommitment* concepts their ASC 840 predecessors use. (#1446)dir/../file entry extract on every supported Python. Python before 3.11.17 and 3.12.15 failed with FileExistsError on such an entry; the entry name is now normalised before extraction. (GH #1598 follow-up)get_shares_outstanding_basic() and get_shares_outstanding_diluted() read the combined basic-and-diluted weighted-average count and no longer return a period-end share count. Snowflake's FY2021 10-K now gives 141,613,196 for both instead of None; a statement with only CommonStockSharesOutstanding returns None rather than that ending count. (GH #1672)WeightedAverageNumberOfShareOutstandingBasicAndDiluted. The share lookups used a misspelled element name, so Snowflake's Q4 FY2021 basic and diluted EPS came back empty; they are now -0.70, matching the reported $(0.70). (GH #1686)USD per share unit were skipped unless named EarningsPerShare; NVDA's 0.04 quarterly dividend (10-Q 0001045810-23-000093) now adjusts to 0.004 for the 10-for-1 split (GH #1687).calculate_ttm_trend rows carry a has_gaps flag. 52/53-week years (Costco's 16-week Q4) count as consecutive. (GH #1694)get_ttm("EarningsPerShare..."). AMC's 2017 TTM continuing-operations basic EPS was -3.75, the sum of its quarters, against the filed -3.80; a fiscal-year window now returns the filed annual amount (Citigroup FY2024: 6.03, with no Q4 quarter), and other windows raise ValidationError. Dividends per share still sum. (GH #1695)get_ttm() summed four quarterly weighted-average share counts, so a TTM share count was about four times too large. Snowflake's basic count for the window ending 2023-10-31 came back 1,300,937,000 instead of the 325,234,250 mean; calculate_ttm now averages share counts and rates the way calculate_ttm_trend does. (GH #1698)get_ttm() period labels now use each quarter's own fiscal year. Apple's TTM revenue to 2023-07-01 was labelled [(2022, 'Q4'), (2024, 'Q1'), (2024, 'Q2'), (2024, 'Q3')] because re-filed comparatives carry the later filing's fiscal year; it now reads 2022, 2023, 2023, 2023. A 52/53-week year closing early next month keeps its year (Costco's quarter to 2023-09-03 was FY2024, now FY2023). (GH #1699)income_statement() picks the parent's statement over a co-registrant's. In Williams' Q2 2026 10-Q it returned Transco's role, so get_revenue() read $804 million instead of Williams' $3,053 million; a co-registrant's own face statement now ranks below the parent's. (GH #1648)Statement.to_dataframe() signs each row by its own presentation arc. A concept presented twice took its first appearance's sign everywhere, so Chesapeake Utilities' second capital-expenditure line read -141.9 million where render() shows 141.9 million; the two frames now agree. (GH #1657)PaymentsOfDividends fact, so JPMorgan's FY2025 dividends ($16,625 million) appeared twice; a node no longer borrows a fact whose own concept has a row. (GH #1674)us-gaap_ ID while the instance files them under us-gaap-supplement, so Gladstone Investment's FY2023 "Total expenses, net of credits to fees" ($44.5 million) showed no value; the fact is now found under the extending prefix. (GH #1680)to_dataframe() view summed every member of the concept, so Deutsche Bank's 2024 operating cash flow read EUR -58,501 million against a filed -28,584 million; plain rows now add only the members the statement itself presents, in every view. (GH #1703)CurrencyConverter took exchange rates from other currencies and hedge contracts. It read every rate fact without its dimensions and kept the last one per year, so GSK's FY2025 20-F (0001131399-26-000004) gave a 2025 average of 1.19 (a EUR swap) instead of the filed US$/£ 1.31. Only target-currency facts off hedge axes are used now, and conflicting rates for a year are reported in extraction_warnings instead of guessed. (GH #1650)CurrencyConverter converted in the wrong direction, or 100x off, depending on how a filer quotes its rate. to_usd() always divided by the rate, so GSK's filed US$/£ 1.31 (20-F 0001131399-26-000004) turned GBP 100 into USD 76.34 instead of 131. Each rate is now read the only way that puts it in a plausible band for its currency. A rate that could read either way, with nothing in the filing to decide, gives None and a warning; rate_basis says how each rate was read. (GH #1651)CurrencyConverter returned no closing exchange rates. An instant fact's empty period_end is NaN, which skipped the period_instant fallback, so every closing rate was dropped with no warning. Novo Nordisk's FY2025 20-F (0000353278-26-000012) now gives a 2025 closing rate of 635 DKK per 100 USD instead of None, and a rate fact with no date is reported in extraction_warnings. (GH #1652)CurrencyConverter found no exchange rates in filings whose rate unit has an opaque ID. Units were matched by substrings of their ID, so SAP's Unit_Divide_USD_EUR_... facts were all skipped and exchange_rates was empty with no warning. Units are now read from their numerator and denominator, either way round: SAP's FY2025 20-F (0001104659-26-020058) gives a 2025 average of 1.1293 USD per EUR, so EUR 1,000 converts to USD 1,129.30. (GH #1653)Prospectus424B.shelf_registration no longer returns None depending on header file-number order. File numbers keep header order (was an unordered set), and the lifecycle loads the full filing history when the recent window holds no S-3/F-3/S-1. BofA Finance 424B2 0001918704-25-005439 now finds shelf 0001193125-22-301119 under either number. (GH #1707)extract_nonaccrual(bdc) no longer picks a data-less 10-K/A. It resolves the latest unamended filing (a later amendment with investment data wins), so ARCC fiscal 2025 reads 10-K 0001287750-26-000006 (20 investment records), and a result with no investment data now carries a warning. BDCEntity.get_filings gains amendments. (GH #1713)Part I, Item 1 no longer returns the contents page for Donnelley filings whose contents rows carry the item number in a separate cell. The Donnelley contents parser now reads the item from the row text. Tesla's 10-Q 0000950170-23-033872 returns Part I, Item 1 as 65,181 characters (was the 4,395-character contents page) and Items 3 and 4 are found. (GH #1749)EntityFacts.time_series("Assets") returned rows from other concepts. A bare name was matched by substring, so 18 of the 20 rows for Snowflake's Assets came from concepts such as AssetsCurrent and AmortizationOfIntangibleAssets. A bare name or label now selects exactly that concept, as get_fact() does; text that names no concept still falls back to a substring search, with a warning naming the concepts it mixed. (GH #1658)to_llm_context() and the MCP edgar_compare tool paired unrelated facts. Both took the first two time_series('Revenue') rows whatever their concept, unit or period, so Snowflake's Q1 FY2026 read -17.96% against a filed +25.75%. Growth now compares one concept, unit and period length a year apart, and reports "not available" when no such pair exists. (GH #1671)to_dataframe(), Statement.periods, period_filter='2024', equity_statement(), get_filings(count=4)). The pages now use the real API, and a test runs every example on Apple's FY2023 10-K: net sales 383,285 million. (GH #1706)xbrl.fiscal_year, statements.names, fact['unit'], unqualified exact=True and fact['period_end'] on instants now use the real API (entity_info, statement_by_type, unit_ref, us-gaap:Assets, period_instant), and a test runs them: Apple FY2023 Assets 352,583 million. (GH #1714).holdings frames. It now concatenates .infotable (110 + 4 = 114 entries) and groups by CUSIP and put/call: 39 securities, Lennar CL B once at 152,774 shares. (GH #1715)net_change > 10000 test is False for sales; Tim Cook's 191,753-share, $63.8M sale (Form 4 0001140361-26-038674) printed nothing. The pattern now tests abs() and says it covers purchases and sales. (GH #1716)ProxyStatement.sections docstring raised on proxies without a CD&A or governance section. Microsoft's 2025 proxy (0001193125-25-245150) has 4 sections and neither key, so .text() hit None and sections[...] raised KeyError. Both now test before calling .text(). (GH #1717)docs/common-pitfalls.md called Filings.filter(company=...), which raises TypeError. The 3 examples now filter by cik=1067983, which selects Berkshire Hathaway's 13F-HR 0000950123-24-008740, and a test runs them. (GH #1718)edgar_read returned full_text: null with success: true for 40-F, 424B5, S-4, S-1/A and 6-K filings. The handler indexed the typed object, which has no subscript support for those forms; it now reads the filing's text (HSBC's 6-K gives 2,490 characters), capped at 6,000, and says so when a filing has no text. (GH #1719)edgar_read earnings section returned only "Earnings data available". It called FinancialTable properties as methods and hid the TypeError. It now returns the income statement, balance sheet and cash flow from the release (Apple's 8-K 0000320193-25-000007: net sales 124,300), and names a missing statement as missing. (GH #1720)edgar_company and edgar_trends reported Python's 'NoneType' object has no attribute text when a company has no facts. For Circuit City (CIK 104599) both returned success: true with that text in 4 fields; they now return the documented NO_FACTS_DATA code. (GH #1721)install_skill() and package_skill() left the funds skill out. The three copy lists stopped at financials, holdings, ownership, reports and xbrl, so the symlink install, the copy install and the Desktop ZIP all shipped 0 of the 1 funds/skill.yaml that other installed skill files point to. All three now include it. (GH #1722)test_install_skill_uses_symlinks no longer fails on Windows accounts without the symlink privilege. The test now accepts the documented copy fallback when links cannot be made, and a regression test forces the refusal on every platform and checks that install_skill() copies SKILL.md and financials/. (GH #1723)get_financial_metrics()['debt_to_assets'] had no definition. It is total liabilities over total assets (Apple FY2023: 290,437 / 352,583 = 0.82), not borrowings over assets (0.32), which FormC.debt_to_asset_ratio uses under a similar name. The docstring and the extraction guide now say so; the value is unchanged. (GH #1729)Note truncated.
Release candidate for edgartools 6.0.0, planned for 2026-11-09.
Release candidate for edgartools 6.0.0, planned for 2026-11-09.
pip install --pre edgartools # or: pip install edgartools==6.0.0rc1
pip install edgartools stays on 5.62.1 until 6.0.0 is released.
6.0 is a major release: Python 3.11 or newer is required, pandas 2.2.2 or newer is required, edgar.files is removed, and a number of fixes return different values. Read the 6.0 upgrade guide before upgrading.
Please report anything that breaks at https://github.com/dgunning/edgartools/issues.
XML parsing is hardened. The minimum lxml is now 6.1.3, which includes an upstream security fix, and edgartools parses filing XML with stricter settin…
Statement.get_raw_data(), the rendered statement and every to_dataframe() export. The decrease concept and its -1 preferred_sign were lost. Oracle's FY2026 10-K (0001193125-26-277521) showed its FY2025 and FY2024 "Cumulative translation adjustments and other, net" decreases ($11M, $8M) as increases with preferred_sign 1; they now stay separate rows at -1. (GH #1471)Document.to_markdown() wrote headerless tables without a delimiter row, so Markdown parsers did not read them as tables. The first non-empty row now becomes the header, followed by the --- row; a table with only one non-empty row, such as a page footer, keeps its plain row. The AbbVie 10-K fixture gains 24 valid tables and no header-only ones. (GH #1484)Document.to_markdown() broke a pipe-table row when a body cell had a line break. A cell with <br> or several paragraphs was written across lines, so Markdown parsers ended the table early; First<br>second now renders as | First second | 10 |. (GH #1486)from_xbrl typed any identifier ending in a bracketed percent as a category roll-up, so PhenixFIN's $221,625 delayed-draw term loan ... (14.00%) left its 10-Q portfolio with no warning. A row that states a principal amount is now parsed without the suffix and kept; Hercules' roll-ups state none and stay excluded. (GH #1504)SecurityLending read every "Yes" securities-lending answer as None. A Yes is stated as attributes of cashCollateralCondition, nonCashCollateralCondition or loanByFundCondition, which from_xml never read. On SGOV's NPORT-P (0002071691-26-016719) the $441,347,500 cash-collateral Yes and two loans by the fund read None. They now read "Y", with the amounts in cash_collateral_value, non_cash_collateral_value and loan_value. (GH #1585)EightK.items missed 8-K items headed "Item - 5.02", and eightk["5.02"] raised SectionNotFoundError. Liaoning Shuiyun Qinghe Rice's 8-K 0000710782-24-000005 puts a dash after "Item", so it reported no items and content_type was other. The item label now accepts a dash, en dash, em dash or colon, and that filing reports ['Item 5.02'] and director_change. (GH #1589)search_filings(start_date=...) without end_date no longer searches every year. EFTS applies a date range only when it gets both bounds, so a lone start or end date was silently ignored: search_filings("tariff", forms="10-K", start_date="2025-01-01") returned 10,000+ hits led by 2010 filings, not 3,841 from 2025-26. A missing bound is now filled in.render_statement() and Statement.to_dataframe() gave an unclassified note role annual duration columns, so a point-in-time note such as IBM's long-term debt components rendered every cell blank; it now shows the note's instants ($824M and $838M for the unamortized discount). render_statement(period_filter=...) is also honoured now. (#1469)Document.to_markdown() let a literal | or \ in a table cell break the table. Only grouped headers were escaped, so a pipe in a flat header or a body cell opened an extra column; AbbVie's FY2024 10-K page footers ("1 | 2024 Form 10-K") rendered three cells for two. Every pipe-table cell is now escaped. (GH #1483)Financials.get_revenue() reads a bank's filed total when its income statement has no total row. U.S. Bancorp and Capital One returned None; they now return the filing's non-dimensional us-gaap:Revenues fact for the statement's period, U.S. Bancorp $28,656M for FY2025.get_revenue() and the income statement's Total Revenue tried contract revenue before Revenues, so GM's FY2025 revenue read $167,971M ("Automotive") instead of $185,019M. They now read Revenues first, as Financials.get_revenue() does; 10 of 106 large filers change. (bead edgartools-jrqvs)configure_http(proxy=...) no longer bypasses the rate limiter and the HTTP cache. The proxy was passed as a client option, and httpx routed every request through its own proxy transport in front of both, so the documented corporate-proxy setup sent unthrottled requests to SEC. The proxy now sits under the cache and rate limiter, and get_http_config()["proxy"] also reports one set by HTTPS_PROXY. (edgartools-kd2ap)Financials.get_revenue() returns a bank's total net revenue, not one part of it. Banks tag their top line as RevenuesNetOfInterestExpense, which was never searched, so JPMorgan's FY2025 revenue came back as $87.0B noninterest revenue instead of $182.4B, Wells Fargo's as $16.1B fee income and American Express's as $37.4B. A bank with no total row now returns None rather than a component.EDGAR_USE_SYSTEM_CERTS is set, so upgrading edgartools now also updates the trusted roots. (edgartools-a9bak)The last 5.x release. 6.0.0 follows on 2026-11-09. Until then, and for about three months afterwards, the 5.62.x branch gets only security and data-co
The last 5.x release. 6.0.0 follows on 2026-11-09. Until then, and for about three months afterwards, the 5.62.x branch gets only security and data-correctness hotfixes. The 6.0 schedule and what it changes are pinned on #936.
Upgrading. These change values you may already depend on:
Document.to_markdown() output changes for grouped tables. The Markdown for 76 of 79 measured filings changes, so rebuild any RAG index built from it (#1474).PortfolioInvestments.from_xbrl() can gain rows. ARCC's 2026-06-30 10-Q gains 3 loans and now sums to its reported $29,349.3M (#1505).XBRL.entity_info on a combined filing names the filer, not a co-registrant. name and ticker can change, as on Duke Energy's 10-K (#1420).SynonymGroups.identify_concepts() no longer matches an extension concept such as onto:NetIncome by its local name (#1418).XBRL.element_catalog reports period_type=None for a concept the filing gives no facts for, instead of 'duration' (#1425).Statement.render() shows share counts on one scale and says so in the caption (#1470). Raw facts and DataFrames are unchanged.Effect.effective_time returns the time of day a registration became effective. Effect.from_xml() dropped the notice's finalEffectivenessDispTime: Lyntris Inc.'s S-1 notice (9999999995-26-002694) gave only 2026-08-18, and now also gives 15:30:00. The property is None when the notice has no time, as on Fitness Fanatics' POS AM notice. effective_date now documents that SEC publishes the notice at about 00:15 ET on the next business day. (GH #1499)get_place_country_code() returns the ISO 3166-1 alpha-2 country for an EDGAR place code. place_codes.csv gains an ISO column for all 309 known codes. Accelerant Holdings (CIK 1997350) files its business address as E9, which now reads as KY (Cayman Islands). The unknown code XX returns None. (GH #1495)acceptance_datetime from get_filings() is now documented as unreliable for precise times. SEC's submissions JSON labels it UTC, but Apple's 10-Q 0000320193-26-000020 read 14:01:02 on 2026-10-08 for a 10:01:02 UTC acceptance. The docstring and filing guide point to filing.header.acceptance_datetime, which matches SEC's index page. (GH #1501)0001628280-26-050307 dropped a $50.5M Proofpoint second-lien loan and two $100k loans, leaving the portfolio $50.7M below its reported total; it now reconciles exactly. (GH #1505)edgar_read dropped the 13F-HR summary section when requested by name. "summary" was treated as the metadata-only inspection mode across all forms, filtering it out from 13F requests. It is now extracted whenever requested on forms that define a summary section. (GH #1415)tests._offline_harness blocked Windows asyncio startup and the audit mislabeled a no-data test as SEC-dependent. On Windows 11/Python 3.14.4, the probe passed normally but errored before its body with the harness. Internal socketpair connections now work while external TCP, DNS and ordinary loopback connections stay blocked. (GH #1490)scripts/check_offline_audit.py truncated failed parameter IDs containing spaces. A selector ending in [Net sales] became [Net; audit runs now retain native pytest IDs, keep distinct parameters separate, and quote complete rerun selectors. Ambiguous legacy reports fail before changing the baseline. (GH #1476)Document.to_markdown() loses duration/date headers on grouped tables. In Apple's 10-Q (0000320193-25-000057), $90,753 million of comparative net sales appeared under "Mar. 29, 2025"; it now appears under "3 Months Ended Mar. 30, 2024". Eaton's blank header rows retain Shares/Dollars placement. Column placement in some equity-table to_dataframe() and text() outputs also changes. Markdown bytes change in 76 of 79 measured filings, so rebuild RAG indexes built from it. (GH #1474)Statement.render() misstated share units in monetary captions. Intuit's 2026 10-K (0000896878-26-000037) displays 276 for 276,000,000 shares; its note now says "In millions, except per share data" instead of excluding shares. Recognized share cells use one consistent scale, keeping the finest existing whole share display. Raw facts, DataFrames, getters and existing per-share formatting are unchanged. (GH #1470)earnings_analysis, industry_overview, and filing_comparison prompts requested rejected inputs. Peer comparisons now request margins instead of net_margin; across-period financial trends request total_assets instead of assets. These templates guide clients and do not execute tools. (GH #1488)edgar_notes(detail="full") returned bare NaN for missing table metadata, which strict JSON parsers rejected. For Apple's 10-Q (0000320193-26-000020), Note 6 Debt, the output used NaN instead of null in balance and weight. Missing cells now become null across every exported record field, preserving the other recorded values. (GH #1489)XBRL.element_catalog called every us-gaap, dei and srt concept a duration. Entries for concepts declared outside the filing carried period_type='duration', data_type='' and balance=None whatever the concept, so on Oracle's 10-Q (0001193125-26-389274) 108 instant concepts with facts, us-gaap_Assets among them, read "duration". The period type now comes from the concept's facts; what the filing does not say is None. (GH #1425)FactQuery.by_statement() and Statement.get_concept_value() do not exist. The docs now use by_statement_type(), chained by_dimension(), to_dataframe() lookups, xbrl.facts.query() and SynonymGroups.export_to_json(), and menucat holds Statements, not S. They also say to_dataframe() defaults to the detailed view: 45 rows on NVIDIA's FY2026 10-K (0001045810-26-000021), where printing shows the 29-row standard view. (GH #1424)XBRL.entity_info named a co-registrant as the filer of a combined filing. The last of the cover's registrant names won, so Duke Energy's FY2025 10-K (0001326160-26-000014) reported PIEDMONT NATURAL GAS COMPANY, INC. beside Duke Energy's own CIK, 1326160; it now reports DUKE ENERGY CORPORATION. The filer's facts win over those tagged with a dei:LegalEntityAxis member, so ticker is also the filer's own symbol whenever it lists one. (GH #1420)SynonymGroups.identify_concepts() gave one XBRL element different answers in its prefix:Name and prefix_Name spellings. Statement frames use the underscore spelling, so on TSMC's FY2025 20-F (0001628280-26-025362) it returned [] for all 22 income-statement rows with a builtin synonym, such as ifrs-full_CostOfSales. Both spellings now drop only a standard prefix (us-gaap, ifrs-full, dei, srt); an extension such as onto:NetIncome no longer matches by local name. (GH #1418)Filing.html() returns an inline XBRL document as filed, without fetching the filing's index page. A primary document that opens with an XML declaration was treated as XML and re-downloaded through the index page, which costs a request and returns SEC's served copy: on JPM's 8-K that copy ends in a 109-character injected bot-detection <script>. html() now reads the document from the submission, so it also no longer depends on the index page being reachable. (GH #1421)A security release. It raises dependency minimums past known vulnerabilities and makes the MCP server's HTTP mode listen only on this machine by defau…
A security release. It raises dependency minimums past known vulnerabilities and makes the MCP server's HTTP mode listen only on this machine by default.
Upgrading.
--transport streamable-http and serve other machines (for example from a container) without passing --host, add --host 0.0.0.0. The default is now 127.0.0.1, so the server only accepts connections from the same machine until you do. With --host 0.0.0.0 the endpoint stays unauthenticated, and the server logs a warning saying so: keep it on a trusted network or behind an authenticating proxy.[ai]. A fresh install already resolved to these versions or newer, so most environments see no change.--transport streamable-http now binds 127.0.0.1 and rejects foreign Host (421) and Origin (403) headers there, which blocks DNS rebinding from a web page. Pass --host 0.0.0.0 to serve other machines, such as from a container; it logs that the endpoint is unauthenticated. stdio is unchanged.[ai] starlette 0.36.0 and mcp 1.12.3, which pip-audit flags; a fresh install already resolved past them. The floors are now each package's first patched release, so 0 advisories at the minimums.A fix release, mainly for the first item below. SEC has started refusing the filing index URL EdgarTools built, so in 5.60.0 Filing.html() returns Non
A fix release, mainly for the first item below. SEC has started refusing the filing index URL EdgarTools built, so in 5.60.0 Filing.html() returns None for many inline-XBRL filings, and anything else that reads a filing's index page fails too. Upgrade if you use 5.60.0 or earlier. This release also fixes installing from the Official MCP Registry, and adds Schedule14D9 for tender-offer recommendations.
Upgrading. No public API changes. Filing.homepage_url now returns the index page inside the accession folder (.../data/<cik>/<accession-no-dashes>/<accession>-index.html), so code that compares the URL as a string will see the new form. The edgar-test console script is gone. It never worked from an installed package; run the harness with python -m scripts.harness.cli from a checkout.
MCP. You can now install the server straight from the Official MCP Registry (io.github.dgunning/edgartools); clients launch it with uvx --with 'edgartools[ai]==5.61.0' edgartools@5.61.0 mcp. The existing uvx --from "edgartools[ai]" edgartools-mcp setup is unchanged.
Schedule14D9 for SC 14D-9 tender-offer recommendations. filing.obj() on a target board's solicitation/recommendation statement returns its position as accept, reject, neutral, or None when the wording is hedged, with the Item 4 text it was read from; an SC 14D9/A that does not restate Item 4 builds with item4_text=None. Classified 52 of the 55 SC 14D-9 originals filed in 2025, each checked against the filing.edgar.httprequests. Behaviour is unchanged: the fast suite passes the same 7,927 tests before and after, and 26 new tests check each branch for each of the four callers. (bead edgartools-fx90)edgar-test console script was removed. It pointed at tests.harness.cli, which ships in no wheel and left the repository when the harness moved, so it failed on every install. Run the harness with python -m scripts.harness.cli from a checkout. (bead edgartools-deq9)Filing.html() and anything else that reads a filing's index page returned None or raised, because SEC now refuses the index URL EdgarTools built. The short /data/<cik>/<accession>-index.html form answers 403 Access Denied for every filing tried; Filing.homepage_url now uses the index page inside the accession folder, which answers 200. (GH #1421)uvx edgartools@<version>, but the package shipped no edgartools executable. A new edgartools mcp command and a committed server.json make that launch start the server with all 13 tools. (GH #1413, bead edgartools-deq9)Section.markdown() dropped a section's own heading when the TOC anchor is nested inside it. The slicer started collecting after the anchor, by which point the heading's block had already begun. On Google's FY2004 10-K (0001193125-05-065298) 0 of 18 item sections began with their "ITEM N." heading; all 18 do now. (GH #1369)A gate that fails a pull request which stages a 6.0 change without updating the upgrade guide. Deprecations, will-raise warnings and "removed in 6.0"…
The last 5.x release before 6.0. It brings the ownership family to the top level (from edgar import Form4, Schedule13D), adds a timeout to the current-filings feed, and fixes 19 defects: most of them in BDC portfolio figures and Schedule 13D/G ownership, plus section text and grep.
6.0 is next (planned 2026-11-09). It drops Python 3.10 and turns several warnings you may already see in 5.x into errors. Running this release with warnings visible shows what will change for your code, and the upgrade guide says what to do about each one. Fixes made from now until 6.0 ship in 6.0.
Upgrading. Nothing is renamed or removed, but some values change because they were wrong:
portfolio_investments() no longer counts holdings twice, and joint-venture holdings, range facts and subtotals no longer inflate totals. Against the filed balance sheet, BCSF went from +90.5% to 0.0% and NMFC from +24.8% to +0.9%.NonAccrualResult.num_nonaccrual is None, not 0, when a filing gives only a portfolio-level non-accrual figure or none at all.total_shares does, instead of summing every reporting person. Some shares_change values change as a result.signatures section.Filing.grep(document="EX-10.1") searches only EX-10.1, not EX-10.10 through EX-10.19 as well. Counts that included those exhibits drop.edgar_fund returns AMBIGUOUS_BDC with up to five candidates when a name matches several BDCs, instead of silently picking one.get_current_filings(), iter_current_filings_pages() and get_all_current_filings() take a timeout. A slow feed page can get 90s without changing the 30s default for every other request, and a timed-out page still gets all 5 attempts where a process-wide 60s timeout allowed one. (GH #1391, bead edgartools-7xtb)from edgar import Form4, Schedule13D now works for the whole ownership family. Form3, Form4, Form5, Form144, Schedule13D and Schedule13G are top-level names, identical to their subpackage classes and loaded on first use, so import edgar time is unchanged (596 ms vs 602 ms). The filing.obj() import table in the API docs had four rows that raised ImportError; all thirteen now import, and a test keeps them that way.get_with_retry (sync and async); with vcr's httpcore patch removed, all three fail.docs/upgrade/6.0.md in the same PR; replayed over the 265 commits since the guide existed, it flags the four PRs that shipped staging with no guide entry, including #1201. Adds a PR template with the Definition of Done checklist.ReportingPerson.reported() and OwnershipComparison.reported_shares_change tell an unreported ownership figure from a reported 0. The 5.x fields still hold a placeholder 0 for a figure the filing leaves out and name it in ReportingPerson.unreported_fields; reported(), reported_shares_change and reported_percent_change return None for it, as the fields will in 6.0. (GH #1379)OwnershipComparison summed joint filers, while total_shares took the largest one. A fund and its adviser reporting one 10,000,000-share position compared as 20,000,000. The comparison now uses the same max, and the new aggregation_basis (single, identical, parent_sums_children, ambiguous) says when it may fall short: GAMCO's Nevro 13D is 1,322,950 by max and 2,123,900 by its own Item 5(a). (bead edgartools-qsk4)unfunded_commitments_fair_value (-1,448.1M) is added back, giving +0.3%. DERA summary_by_company() gains total_source (filed or summed). (bead edgartools-3vad)edgar_fund silently picked one BDC when a name matched several. "Golub" returned GOLUB CAPITAL BDC though four Golub BDCs score 99.4-100. A name now resolves only on one match equal to it (ignoring a suffix like Corp or Inc., so "Ares Capital" is still ARES CAPITAL CORP) or a clear top hit; otherwise the tool returns AMBIGUOUS_BDC with up to five candidates and their CIKs. (bead edgartools-1c6z)Filing.grep(document="EX-10.1") also searched EX-10.10 through EX-10.19. In SeeQC's S-1 it returned 378 matches for "agreement" where EX-10.1 holds 133. An exact filename or document type now wins, and a substring match applies only when nothing matches exactly, so document="EX-10" still selects every EX-10 exhibit. (bead edgartools-5qqy)grep returned shifted matches after a character whose lowercase is longer. Positions were found in the lowercased text and applied to the original, so four "İ" before "going concern" returned "g concern dou". Literal matches are now found in the original text, case-insensitively. (bead edgartools-6th2)BDCEntity.is_active would have turned False for BDCs the newest SEC report dropped, though they still file. Their rows come from the 2025 report, so Ares Capital's activity date stopped at 2025-05-29 and would leave the 18-month window around 2026-11-29. For these 22 BDCs, is_active now checks the company's own filings once the report date expires; the new in_latest_report flag marks them. (bead edgartools-huul)bdc_portfolio investment records named no borrower. Every record had only type, fair value, cost and rate because the builder read a name field PortfolioInvestment does not have. Records now carry company_name, identifier, principal, PIK rate, spread and percent of net assets, with None for a missing figure; MAIN's first record is "MSC Adviser I, LLC", $255.0M. (bead edgartools-rd23)NonAccrualResult.num_nonaccrual reported 0 when a filing gave only a portfolio-level non-accrual figure, or none. WhiteHorse Finance files $10.6M of non-accrual investments for 2025 with no per-investment detail and showed 0; the count is now None unless the filing itemizes, and the new evidence_level says which (investment, aggregate, none). (bead edgartools-vblr)HTMLParser now collapses whitespace runs inside a text node to one space, as a browser does, while <br> and <pre> still break lines. Item 7 of Google's FY2004 10-K had about 500 mid-sentence newlines. (#1370)portfolio_investments() no longer counts holdings twice. Company totals, tranche parents and a second schedule's copies of a row were read as holdings, so rows summed 2-48% above the filer's own total on every BDC checked; MAIN's FY2025 10-K summed to $6.30B against a reported $5.52B and now reconciles within 0.2%. Foreign-currency values no longer replace USD ones, and new reported_total_fair_value and reconciliation_gap show any remaining difference.TenK keeps both items of a combined heading such as "Items 1 and 2. Business and Properties". A TOC link reading "Items 1 and 2" (or "Items 1. and 2.", "Items 7. and 7A.") was dropped, so tenk.business was None on Viper, Devon, Freeport and Cheniere. The section is now keyed as the first item, lists both in Section.covered_items, and answers for either. Talos no longer shows a 1,311-character "Properties" fragment as Item 2. (GH #1382, #1383)OwnershipComparison.get_summary() warnings now point at the caller. It emitted two FutureWarnings attributed to amendments.py, so Python showed them once per process wherever it was called from; it now emits one, at the caller's line. A filing with an empty <reportingPersons> now displays its ownership as "not reported" instead of 0, matching the comparison. (GH #1379)OwnershipComparison.is_liquidating reported a sale against a Schedule 13D/G filing that gives no share figures. A 13D/A or 13G/A may omit them and a pre-2025 header-only filing has none, but they were read as 0: an amendment omitting them on the Aadi Bioscience 13D in the test data sold all 4,535,000 shares. The direction flags are now False when the change is unknown, and shares_change and percent_change warn that they return None in 6.0. (GH #1379)edgar.ownership.core.is_numeric() raised TypeError for pandas string and nullable numeric columns. The reported Arnaboldi Form 4 has UnderlyingShares of 0 [F1]; its value calculation now returns None instead of raising. Ownership share totals also return None for unreadable text rather than raising. (GH #1368, thanks @manantlerio)EntityFacts.get_total_liabilities() returned total assets for filers with no standalone liabilities total. Its last fallback was LiabilitiesAndStockholdersEquity, which always equals total assets: NIKE's FY2026 figure came back as $38.41 billion against $23.545 billion of liabilities, and 16 of 51 large US filers were affected. It now returns None there, matching Financials.get_total_liabilities(). (GH #1279)PortfolioInvestment.company_name returned the instrument (First Lien, Ordinary Shares) when the filer tagged the company as a member. On PSEC's FY2026 10-K (0001287032-26-000269) 142 of 248 identifiers named an instrument. A member candidate that fills the whole first pipe segment is now the company, not a prefix to strip. (GH #1373)PortfolioInvestment.company_name named a schedule heading or sub-sector instead of the company on category-led identifiers. On CGBD's Q2 2026 10-Q (0001544206-26-000055) 348 of 352 positions read Investment, Credit Fund or a sub-sector such as Durable; SLRC and PFX were hit too. Only the axis prefix is stripped now, and leading heading segments are skipped. (GH #1372)Form4.shares_traded raised AttributeError on a derivative-only filing. market_trades is None when the non-derivative table has no transactions, and the accessor read .Shares off it without the None/empty guard its neighbours already had: Vertex's 2020-10-19 Form 4 parsed and returned its derivative transaction but crashed on shares_traded, which now reports 0 market trades. (GH #1359)A patch release with two fixes. One is a regression from 5.59.0: ExxonMobil's 10-K Item 7 opened on a "Table of Contents" breadcrumb again. The other:
A patch release with two fixes. One is a regression from 5.59.0: ExxonMobil's 10-K Item 7 opened on a "Table of Contents" breadcrumb again. The other: edgar_read returned no 13F holdings section for any filing.
Upgrading. No action needed:
edgar_read returns a 13F holdings section. It lists the top 30 securities by value, with issuer, shares and value. The header counts securities ("29 of 29" for Berkshire), while the summary's Total Holdings counts information-table lines (89), so the two numbers can differ.edgar_read returned no holdings section for any 13F-HR. A filing whose summary reported Total Holdings: 211 came back with holdings set to None, because the section tested a DataFrame for truth and then looped over its column names. It now lists the top 30 rows by value (issuer, shares, value), and an extraction failure logs at warning level. (GH #1337)`EDGAR_ACCESS_MODE` and `NORMAL`/`CAUTION`/`CRAWL` emit `DeprecationWarning`. They never changed how edgartools talks to the SEC; use EDGAR_RATE_LIMIT…
The headline is filing sections. Seven fixes change how 10-K and 10-Q items are found through the table of contents. Items no longer return a neighbouring section, a summary, or each other's text: FirstEnergy's Item 7 no longer returns Item 8, Southern Co.'s Item 7A no longer duplicates Item 7, and Citi's FY2022 Item 7A no longer opens on Item 1. Tables in those sections render cell by cell instead of fusing numbers ("20217,294,800"). Alongside that: a 13F with a couple of holdings could be valued 1,000× too high without a warning, SGML downloads no longer cache an empty filing header when the SEC errors or refuses a request, and four XBRL fixes, including balance-sheet debt reported under combined concepts.
Upgrading. Several of these change values or behaviour you may rely on:
doc.text() keeps content it used to drop. Short figures such as "$85" and a cell's own label ahead of its <div>s are now included.Filing.sgml() and Filing.header raise on HTTP errors. A 5xx, 403 or 429 now propagates, where it used to return a header with every field None. The 429 keeps its retry_after. A later call on the same Filing retries.Ambiguous13FValueUnitWarning, so code that promotes warnings to errors may see it.sections['risk_factors'] works when the section is keyed part_i_item_1a, and 'mda' in sections is true when Item 7 exists.EDGAR_ACCESS_MODE and NORMAL/CAUTION/CRAWL emit DeprecationWarning. They never changed how edgartools talks to the SEC; use EDGAR_RATE_LIMIT_PER_SEC and EDGAR_HTTP_TIMEOUT. They are removed in 6.0.balance_sheet() gains debt lines for filers reporting under combined or short-term debt concepts.Notes.from_xbrl() without a FilingSummary now returns disclosure-category stems with their Tables, Policies and Details. The fallback builder only accepted note-category roles, and a family that hangs from a *DisclosureAbstract is classified disclosure, so XBRL.from_directory() callers got an empty concept index. gahc goes from 0 notes to 7, aapl from 0 to 16. (GH #1218)EDGAR_ACCESS_MODE and the NORMAL/CAUTION/CRAWL modes are deprecated and have no effect. They advertised a timeout, connection-limit and retry policy that was never wired into the HTTP client; nothing in the package read edgar_mode, so all 3 modes behaved identically. Use EDGAR_RATE_LIMIT_PER_SEC and EDGAR_HTTP_TIMEOUT. The names still import and now warn; removed in 6.0. (GH #1326)20217,294,800 for the year 2021 and a 7,294,800 share count; 144 such tokens across 70 fixtures are now 0. Tables render with the cells doc.text() shows, and doc.text() itself now keeps short figures such as Regions' "$85" and cell labels it used to drop. (bead edgartools-wzgu)Filing.sgml() and Filing.header now propagate every status-bearing HTTP error in both error modes, preserving the original 429 retry_after value. After an injected refusal, a later successful download on the same filing recovers Apple's 2024-09-28 report period; content-error homepage fallback is unchanged.obj["Item 7"] returned Item 8's financial statements (306,341 chars) instead of the MD&A. Such rows now keep their item label. Patch by @sf1tzp. (GH #1347)Item 7 and Item 7A sliced to one span; they are now separated by their headings, so Southern Co's obj['Item 7A'] returns its 347-char cross-reference instead of Item 7's 282,239 chars. Prospectus markdown() now also stops at the financial statements, as text() did. (GH #1345, bead edgartools-rc46)Item 7A opened three pages early on Item 1's human-capital section. The offset is now calibrated from the printed footers and is a no-op when they are aligned. (GH #1346)Filing.sgml() and Filing.header propagate HTTP 5xx errors in both default and strict error modes, leaving the same filing retryable. An injected 503 followed by Apple's checked-in 2024 10-K now recovers its 2024-09-28 report period on the second call.dei:LegalEntityAxis and a classification axis; the member hierarchy keyed rows by member alone, so the LegalEntity row — carrying the more precise $82.723B against the other's $82.7B — was dropped. A member on two axes is no longer reordered.Disclosure marker. UNP names most of one family DisclosureDebtDetails1 but two members DebtDetails6, so those surfaced as spurious top-level notes while the real Debt note lost two Details. The marker is now ignored when matching a stem, taking UNP from 20 notes to 18 and 318 reachable concepts to 336.Filing.html() raised AttributeError instead of returning None when a filing's homepage listed no primary document. The homepage property is optional and comes back None for some filings, but two call sites used it unguarded, so the scheduled build went red on a PDF-primary APP NTC filing. Both are guarded now.Tables and Details roles were returned by xbrl.notes() while their parent was returned by xbrl.disclosures(). In a filing that falls back to role names only the family stem hangs from a *DisclosureAbstract, so gahc's ConvertiblePromissoryNotesPayable moved alone and its 4 children stayed notes. A child now follows its nearest stem when that stem declares a disclosure; no other role in the 7 committed fixtures changes. (GH #1218)Statement.get_raw_data(view="detailed") dropped NVIDIA's reportable-segment revenue. On the FY2026 10-K (0001045810-26-000021) the default view kept Compute & Networking $193.479B and Graphics $22.459B; DETAILED returned neither and showed ProductOrService Compute/Networking instead. Member hierarchy keyed rows by the first axis, so two-axis facts collided once Data Center's children entered the set. Hierarchy now nests only single-axis members. (GH #1331)edgar could fail. Importing edgar once on the main thread before starting threads avoids it; this is now documented under Common Pitfalls. A structural fix through lazy submodule loading is planned for 6.0. (GH #1325)set_rate_limit() and enable_local_storage() appeared in the performance, SEC-compliance and Form 4 guides but are not in the package; the working names are EDGAR_RATE_LIMIT_PER_SEC (set before import; the default is 9, not 10) and use_local_storage(). The configuration page said submissions are cached up to 10 minutes; MAX_SUBMISSIONS_AGE_SECONDS has been 30 seconds since #471.PortfolioInvestments.from_xbrl() cut the first word off a borrower's name when the filer labelled a range member with it. BXSL's FY2025 10-K (0001736035-26-000004) returned Street Buyer, Inc. 2 with industry='High' for its four High Street Buyer, Inc. positions: BXSL labels srt:MaximumMember "High", and every member label was an industry candidate. The five srt:RangeAxis members no longer are; across 12 filings and 8,212 positions only those rows change. (GH #1324)balance_sheet() dropped debt reported under combined or short-term concepts. Filers using LongTermDebtAndCapitalLeaseObligations, DebtCurrent or ShortTermBorrowings lost the line entirely: CSX's balance sheet showed no debt, and now shows $18.165B long-term plus $708M current. (PR #1330, thanks @wittling)parse_investment_identifier() recompiled up to 933 regexes on every call. Six sites built patterns from INVESTMENT_TYPES per call, past the 512 re caches, so a piped Company | Type identifier never hit the cache. They are now compiled once and cached: parsing BXSL's 703 identifiers on its Q2 2026 10-Q (0001736035-26-000016) drops from 40.8 s to 1.1 s, every field identical. (GH #1342)to_dataframe(*columns) no longer builds the full fact width before projecting. Selecting columns now narrows the frame at construction instead of at the end, so a two-column call on the JPM 10-K fixture drops from 7.59 MiB to 0.71 MiB peak and from 78 ms to 8 ms, with a byte-identical frame. (GH #1181)The guide now requires Python 3.10, matching package metadata, and identifies cashflow_statement() as deprecated in favor of cash_flow_statement().
The headline is the SEC's BDC data sets. When the SEC relocated them, the fair value column changed its heading and BDCDataset.summary_by_industry() quietly reported $0 for every industry. It now resolves fair value and industry from every header DERA has used, and portfolio investments carry an industry, where it came from, and a normalized sector. Alongside that: seven XBRL statement fixes, a warning when a 13F's value units are ambiguous, and a BDC register that stops losing companies each January.
Upgrading. Several of these change values you already read:
summary_by_industry() sums fair value from the current period only, counting a filer's industry subtotals rather than adding its line items on top, and gains a num_bdcs column. ARCC's 2024Q4 Software And Services exposure reads the filed $6.572B. summary_by_industry(by='sector') merges spellings such as "Software Sector" and "Software & Services".summary_by_company() reports fair value and counts investments, not rows. It gains total_fair_value, taken from the filer's own total where reported, and is sorted by it. num_investments is now the number of distinct positions at the period end.to_dataframe(clean=True) coalesces BDC columns. fair_value, cost, shares, principal and industry are single canonical columns, followed by industry_source and sector.get_bdc_list() combines the last three yearly reports, so is_bdc_cik() recognises Ares Capital again.[Axis], [Domain] and member nodes no longer appear as line items; rendered output is unchanged.decimals can be the string "INF" on assembled statement rows, as filed, where it used to be 0. Code that assumes an integer should allow for it.annual_report=True, and 20-F and 40-F annual reports are recognised.Credit Agency. Fannie Mae, Freddie Mac and Farmer Mac move off Operating Company, and is_financial_institution() counts them.XBRLS.from_filings() logs a warning naming any filing that fails to parse, instead of dropping it silently.edgar.dates. edgar.core re-exports them until 6.0.PortfolioInvestment.industry, .industry_source (axis, enumeration, identifier, peer) and .sector, with PortfolioInvestments.filter(industry=) and .by_industry(); the DERA path reads the same fallbacks and gains summary_by_industry(by='sector') and a summary_by_company() with fair value. Filings with a resolvable industry in 2025Q2 rise from 123 to 158 of 160. (bead edgartools-t1wr)get_period_views("CashFlowStatement") returned an empty list because the statement type had no entry in the period-view configuration, so to_dataframe(period_view=...) had no name to accept. Cash flow now offers the same named views as the income statement, which selects from the same periods. (GH #1253)edgar.dates. filing_date_to_year_quarters, current_year_and_quarter, is_start_of_quarter and parse_acceptance_datetime now live beside extract_dates instead of in edgar.core; edgar.core re-exports the same objects until 6.0 removes the shim. Continues the bounded extraction that produced edgar/settings.py, taking core.py from 579 to 514 lines.Path(__file__).parent, which breaks the moment a file moves directory; they now import TESTS_DIR, REPO_ROOT, FIXTURES_DIR, CASSETTES_DIR and DATA_DIR from a new tests/paths.py. Behaviour-preserving — test-fast reports the same 7,261 passed as before — and it is the prerequisite for the test-tree reorganisation.edgar/documents/utils/html_utils.py as text_content, text_stripped, text_joined and text_skipping_tables, with an html_to_text wrapper. Behaviour-preserving: the five characterization suites pass unchanged against their recorded bs4 baselines, 297 tests.docs/installation.md listed Python 3.8+ in two places (System Requirements and the venv setup section) while pyproject.toml requires >=3.10. Both now read 3.10, matching package metadata.cashflow_statement() as deprecated in favor of cash_flow_statement().ThirteenF could silently inflate ambiguous filing values by 1,000x. It now warns on ambiguous thousands conversions and exposes raw values and unit diagnostics. A verified value_unit='dollars' override reads Kahn Brothers' Q1 2022 holdings as $787,553,692; the SEC summary remains $1 higher. Automatic unit choices are unchanged.BDCDataset.summary_by_industry() reported $0 for every industry. The SEC's relocated BDC data sets head the fair value column Initial fair value of Investment and leave Investment Owned, Fair Value present but empty, so it summed an empty column. Fair value, cost, shares and industry now resolve from every header DERA has used, the summary counts only current-period rows and each filing's industry subtotals, and ARCC's Software And Services exposure comes back as the filed $6.572B.find_mutual_fund_cik() could answer from a stale ticker frame. The ticker-to-CIK dict had its own cache stacked on the cached frame, so clearing or replacing the frame left the lookup frozen for the rest of the process. The dict now follows the frame it was built from, and the fast test serves a 3-row slice of the SEC file instead of downloading it.entity_info compared the whole dei:DocumentType against 10-K, so Shopify's 10-K/A reported amendment=True alongside annual_report=False, contradicting the filing's own dei:DocumentAnnualReport. The amendment suffix is now stripped before classifying, and 20-F/40-F annual reports are recognised too. (GH #1226)Credit Agency category that is_financial_institution() counts; 6120 joins the bank codes as a depository. (GH #1120)get_bdc_list() read one year's CSV, and the 2026 report dropped 15 registrants the 2025 one carried — Ares Capital among them — so is_bdc_cik(1287750) answered False. The last three reports are now combined, restoring 22 registrants (212 to 234) with the most recent row per CIK. (GH #1146)point_in_time. Both lookups walked only the displayed columns, which are durations on a cash flow statement, so Apple's beginning and ending cash rows returned point_in_time=None and unit=NaN despite their metadata saying instant and usd. Period type and unit belong to the concept and are now read that way. (GH #1228)decimals="INF" was rewritten to integer 0 in assembled statements. Apple's exactly-filed $0.00001 par value reached get_statement() claiming to be rounded to the dollar, while the same fact still read INF through FactQuery. The sentinel is now preserved; scaling and formatting ask decimals_for_scaling for a number, so rendered statements are unchanged. (GH #1229)show_date_range=True was ignored by a stitched statement's normal rendering. The accessor stored the option, but StitchedStatement.render() defaulted the parameter to False and never read it, so repr() showed Dec 31, 2024 where an explicit render(show_date_range=True) showed Jan 1, 2024 - Dec 31, 2024. An explicit argument still wins. (GH #1176)get_statement() returned the presentation linkbase's [Axis], [Domain] and member nodes as line items — four empty automotive headings in Tesla's operations statement. Across the 8 committed fixtures this drops 2,390 such rows from 268 statements, with no value-bearing row lost and no change to rendered output. (GH #1224)Statement.to_markdown() emitted labels verbatim, so Global Arena Holding's "conversion of debt | shares" row carried four unescaped pipes where a two-column row has three, and Markdown parsers read it as a different shape than the header. Labels, headers and values are now escaped. (GH #1227)XBRLS with no trace. XBRLS.from_filings() swallowed every exception, so a batch that lost a filing looked identical to a complete one. The failure is now logged as a warning naming the accession number, form and exception; a filing with no XBRL at all is still skipped quietly. (GH #1174)Thirty-eight entries, all in the XBRL parser, the fact queries, stitching, and the Financials getters. The recurring shape is a lookup keyed too narro
Thirty-eight entries, all in the XBRL parser, the fact queries, stitching, and the Financials getters. The recurring shape is a lookup keyed too narrowly to tell filed facts apart — a context ID, a currency, a concept name, a label pattern — that then returned the first match or silently dropped the second. Every one of these now identifies the fact the filing actually reported, or returns None rather than a substitute.
Upgrading. Several of these change values you already read:
xbrl.us labels, statements selected by role URI, render(standard=False), and Statements.to_dataframe() all agree with Statement.to_dataframe(). Apple's FY2023 capex is −10,959,000,000 everywhere; it was positive on three of those paths.Financials scalar getters identify concepts, not labels. get_total_liabilities() no longer returns liabilities and equity (NIKE's debt_to_assets was exactly 1.0); where a filing reports no consolidated us-gaap:Liabilities — Coca-Cola, Amazon — it now returns None. get_revenue() returns the filed total rather than a component; get_shares_outstanding_basic() no longer stops at an abstract heading.FactQuery.to_dataframe(), EntityFacts, and stitched queries return the same columns and dtypes regardless of which rows match; stitched frames keep columns the matched rows leave empty. pivot_by_period() and _deduplicate_facts keep facts that collide on unit or currency — expect more rows.calculate_ratios() computes every ratio from one period; analyze_trends() works on a two-period 10-K; RenderedStatement.to_dict() is JSON-safe (Infinity → None).xbrl.axes / domains_for_role() keep an axis's second domain root and childless roots.FactQuery.sort_by() no longer reorders the shared cache, so independent queries stop changing each other's results.get_financial_metrics() renders each statement once instead of once per metric.aggregate() summed the filed strings and raised TypeError on ordinary numeric facts. It now aggregates numeric_value and skips facts that have none; Apple's FY2023 product revenue by srt:ProductOrServiceAxis returns totals. (GH #1277, bead edgartools-n93w.3)by_concept() could not match a concept whose name contains an underscore. Every _ was rewritten to :, so YUM's yum:YUM_LesseeOperatingLease… matched nothing in either spelling. The pattern is now matched against both spellings unaltered; 20,094 exact queries across 15 filings lose no row. (GH #1276, bead edgartools-n93w.4)facts_history(include_dimensions=True) returned the undimensioned series. The flag never reached the query. It does now, and the date column falls back to period_instant for instant facts — previously every balance-sheet concept returned an empty frame. (GH #1286, bead edgartools-n93w.5)sort_by() sorted the shared fact list in place, so an unrelated .limit(5) changed on 15 of 15 filings measured. The sort now builds a new list. (GH #1287, bead edgartools-n93w.1)sort_by() raised TypeError on most filings, and silently did nothing on a sparse column. None values broke the comparator, and the column was looked for in the first row only. The key is now a total order over mixed types and the column is found anywhere in the result. (GH #1288, bead edgartools-n93w.2)analyze_trends() returned {} for a 10-K reporting exactly two annual periods. It required a named three-period view; it now falls back to the periods the statement displays, so Auburn National's $603,000 and $598,000 of contract revenue are reported. (GH #1292, bead edgartools-hsgs.3)calculate_ratios() let each operand take the first value in its own dictionary, so Netflix's Q3 2024 net margin was 0.2767 against a correct 0.2406. Every operand now comes from one period, and a ratio missing an operand is omitted rather than borrowed. (GH #1280, bead edgartools-hsgs.1)RenderedStatement.to_dict() produced output that was not JSON. A prior-period value of zero yielded Infinity in the comparison, which 17 corpus filings hit. Non-finite comparisons now serialize as null. (GH #1290, bead edgartools-hsgs.4)efc_Year2038Member was absent from domains_for_role() though five facts use it. Every root is now registered; Axis.domain_id still names the first. (GH #1278, bead edgartools-xn1u.2)19-Sep-2024 failed all thirteen word-date formats. The tokenizer now accepts the registry's separators; 40,219 facts across 16 filings show no already-transformed value changing. (GH #1284, bead edgartools-xn1u.1)render(standard=False) displayed cash-flow outflows as positive, disagreeing with its own DataFrame. The statement type was stamped on rows only along the standardization path; it is now stamped regardless of the flag. Apple's FY2023 capex reads −10,959 in both. (GH #1289, bead edgartools-vaau.2)get_total_liabilities() returned liabilities and equity, so debt_to_assets came back as exactly 1.0. A Liabilities$ label pattern matched the combined total; the getter now identifies the concept and returns None where no consolidated total is filed (Coca-Cola, Amazon). NIKE's FY2026 ratio goes from 1.0 to the correct 0.613. (GH #1279, bead edgartools-y4yp.1)get_revenue() returned a component of revenue instead of the filed total. Contract revenue was tried before Revenues, understating Cato's FY2023 top line by $7.7M of other income. The order now matches Statement.REVENUE_CONCEPTS. (GH #1294, bead edgartools-y4yp.2)get_shares_outstanding_basic() returned None for shares present in the statement. The lookup stopped at the abstract heading above the populated row; it now skips abstract rows and iterates its matches, returning Auburn National's 3,498,030. (GH #1291, bead edgartools-y4yp.3)_1 collided with an internal duplicate-occurrence index, and the parser silently dropped one of the two facts. ClearOne's Q2 2025 10-Q lost 6 of 590 facts, including the quarter's own revenue. Key parts are now joined with |, which no NCName admits. (GH #1295, bead edgartools-jj77.1)unit was discarded as a structural element. ctso:NumberOfSharesInAunit was skipped by a bare-suffix test; structural elements are now matched by fully qualified name. (GH #1293, bead edgartools-jj77.2)by_statement_type('BalanceSheet') lost period_start on 5 of 5 stitched pairs. Columns and dtypes now follow the query's configuration per engineering/decisions/facts-dataframe-schema.md. (bead edgartools-trhe)pivot_by_period() on company facts discarded facts that collided on a key too narrow to tell them apart, and for an IFRS filer discarded nearly all of them. IFRS concepts carry no label upstream, so 765 of LPA's 768 facts shared one row. Unit is now part of fact identity and the pivot widens its key only for rows that collide; the filing pivots to 188 rows. (GH #1196, bead edgartools-6byc)source_filing_index was always None and by_filing_index() matched nothing. Rows now carry the filing the stitcher selected each period from; a two-filing ORCL stitch splits 115/100. (GH #1205, bead edgartools-qbm7)value flipped float64→int64 on 16 of 36 narrowing queries, and an empty result had no columns. FactQuery.to_dataframe() and EntityFacts.to_dataframe() now declare their columns and dtypes; fiscal_year stays int64 until 6.0. (bead edgartools-7wtj)Total* concepts, so a standardized statement returned several rows all claiming to be the same total. Citizens Financial's balance sheet had three "Total Assets" rows. A tag not marked is_total no longer resolves to a total (74 spellings), and IFRS tags answer only from IFRS entries. (GH #914, bead edgartools-gnx5)render(standard=True) returned a DataFrame with no standardization in it. The standard_concept column was never emitted. It is now present for every standardized statement and absent otherwise; filer labels are unchanged. (GH #1083, bead edgartools-t3zh)_merge_complementary_rows() matched dimensional rows on label alone, merging 213 different-concept pairs across the corpus. A dimensional row must now also match on concept; 2 merges remain. (GH #1206, bead edgartools-j7iz)xbrli:divide was unreachable, so usdPerShare was recorded as USD. Divided units are now parsed, and the unit-to-currency rule lives once in edgar.xbrl.core.unit_currency. (GH #1170, bead edgartools-uetp)PresentationTree now carries root_element_ids; 9 of 2,052 corpus roles are multi-root and gain 106 rows. (GH #1245, bead edgartools-0q0d)xbrl.us roles matched neither of two independent matchers, so Apple's FY2010 securities purchases showed +57,793M. Both paths now use is_negated_label_role; Apple FY2023 goes from 0 to 130 negated facts. (GH #1249, bead edgartools-hxnw)Statements.to_dataframe() and Statement.to_dataframe() returned different signs for the same line. The sign now lives in one function called by both; 13 rows of Apple's FY2023 cash flow change. to_dataframe(presentation=False) returns the raw filed values. (GH #1248, bead edgartools-5ztr)get_structured_statement(fiscal_period='FY') could return the prior-year comparative. A 10-K tags every instant fp=FY, and the filing-date tiebreak never fired. Period end now breaks the tie; Apple's FY2023 balance sheet reports AssetsNoncurrent 209,017,000,000 instead of the FY2022 217,350,000,000. (bead edgartools-i24a, GH discussion #946)select_periods(max_periods=4) returned 10 periods; balance-sheet views admitted empty incidental instants; parenthetical=True returned the face statement. The cap is applied, instants are fact-checked, and a filing with no parenthetical role raises StatementNotFoundError. (GH #1221, #1222, #1246, beads edgartools-rj4b, edgartools-zt8u, edgartools-z0y2)Statement helpers had never returned a correct result for any filing. analyze_trends() read a key that was never emitted, calculate_ratios() looked up reversed concept names, and to_dataframe(matrix=True) read every cell from standard_concept. All three now return filed values; Apple's FY2023 equity matrix goes from 33 null cells to its filed balances. (GH #1240, #1241, #1244, bead edgartools-ysr8)get_financial_metrics() rendered and converted the same three statements once per metric. The three getter helpers share one _render_statement_frame() with a per-call memo, so a call renders 3 times instead of 14–18. JPMorgan's warm median drops from 95.2 to 24.8 ms with every metric byte-identical. (GH #1283, bead edgartools-y4yp.4)It backs Filing.search(), and its old backend is deleted in 6.0, so it needed a replacement rather than a deprecation. Chunks are cut at headings with…
Twenty-seven entries, led by the XBRL dimensional and calculation models. Both were keyed too coarsely — axes and domains by concept when they belong to a role, calculation edges by concept when a concept can roll up into two totals — so each returned one answer where the filing gives several.
Upgrading. Several of these change values you already read, all in the direction of returning more or refusing to guess:
xbrl.axes and xbrl.domains keep their shape but are now a union across roles rather than whichever role was parsed last. Use the new xbrl.axes_for_role(role_uri) / xbrl.domains_for_role(role_uri) wherever the role is known — the flat views cannot distinguish a domain's two-way breakdown on the income statement from its five-way one in a revenue note.calculation_linkbase() returns more rows: one per filed relationship rather than one per concept. Coca-Cola's FY2023 10-K goes from 278 rows to its filed 306. Code that assumed one row per concept will now see a concept twice, with a different parent and weight — which is what the filing says.get_revenue(unit=...) and get_gross_profit(unit=...) return None for a unit the company does not report in, where they previously returned a USD figure. If you passed unit='EUR' and got a number, that number was dollars.<div> is no longer dropped and nested text is no longer duplicated. related_fact_ids now hold fact ids rather than locator labels, so get_footnotes_for_fact() resolves on filings where it previously returned nothing.Report.get_dataframe() returns data instead of an empty DataFrame, and Document.chunks() no longer raises ModuleNotFoundError. Both were broken for every filing.CurrentReport.doc now warns; it was the last un-warned public route into edgar.files, which 6.0 deletes. Use .document, or .items / report[item].pip install --upgrade edgartools
XBRL.axes_for_role(role_uri) and XBRL.domains_for_role(role_uri) return the axes and domains a single extended link role declares, keyed on element ID. Prefer them over xbrl.axes / xbrl.domains wherever the role is known: the same domain routinely carries different members in different roles, so the flat views can only offer the union. Axis and Domain gain role_uri, and XBRL.axes_by_role / XBRL.domains_by_role expose the underlying store.CalculationTree.all_arcs, one entry per filed calculation relationship. A concept may roll up into two totals with a different weight, and often a different sign, under each, which the one-node-per-concept all_nodes map cannot represent. all_nodes is unchanged and still the right thing for lookup and membership; all_arcs is the graph.ElementCatalog.typed_domain_ref and ElementCatalog.substitution_group, read from the element declaration, which is what lets Axis.is_typed_dimension report a dimension the filer declared as typed.get_revenue(unit=...) answered a request for one unit with a figure in another. When the unit filter correctly rejected every candidate fact, _get_standardized_concept_value fell through to its component calculation, and that calculation checked the two components against each other — which establishes only that they can be added — before normalising with no target unit at all. Apple's GrossProfit and CostOfGoodsAndServicesSold are both USD, so their sum was returned as the answer to unit='EUR' and to unit='shares' alike: 364,357,000,000 in both cases, and both now None. The default USD path is unchanged and still answers from the revenue fact itself. Strictness now reaches the fallbacks, and a derived figure has to answer the unit that was asked for — an exact match always does, a merely compatible one only where the caller did not pin the unit, which is the rule the direct path already used. get_gross_profit() was leaking the same way for a second reason — its own direct branch never asked for strict matching at all, so a compatible-but-different currency succeeded before the guard was even reached — and both fallbacks now answer by the same rule as the direct path. This was latent until bare concept names were indexed (GH #1202, same release): before that, get_fact('GrossProfit') returned None, the components were never found, and the fallback returned None for the wrong reason. (bead edgartools-885h)
Report.get_dataframe() returned an empty DataFrame for every filing. FinancialTableExtractor read table_node._processed, an attribute that exists only on the legacy TableNode, while the module parses with the modern one — so every call raised AttributeError into a bare except that returned an empty frame. Measured before the fix: 642 of 642 tables across 21 filings raised, and 0 of 42 Apple R-files produced data; after, 41 of 42, with Apple's Q2 FY2025 net sales reading the filed $95,359M and $219,659M. Two further defects surfaced once the code could run: a short period-header row was padded on the left, attributing every figure to the wrong period, and AUD without a word boundary matched inside "Unaudited", so an unaudited USD statement reported Australian dollars. The money scale now prefers the currency-qualified phrase, since "shares in Thousands, $ in Millions" was read as thousands. (bead edgartools-yq1l)
Document.chunks() raised ModuleNotFoundError on every call. It imported edgar.documents.extractors.chunk_extractor, a module that was never written, so this public retrieval API — token-budgeted chunks with overlap, the shape a RAG pipeline asks for — has been unusable since the parser rewrite, along with the DocumentChunk class that was only reachable through it. The module now exists; overlap >= chunk_size raises ValueError rather than advancing the window by zero and looping forever. (bead edgartools-vwtb)
Two taxonomies whose namespace URIs end in the same path segment became one concept. A concept is identified by its expanded name — namespace URI plus local name — and the prefix is only a display choice, but when a namespace was declared on the fact element rather than the instance root the parser fell back to deriving the prefix from the URI's final path segment. {http://example.com/l01/alpha/2024}Collision and {http://example.com/l01/beta/2024}Collision both became 2024:Collision: both facts survived, their technical identity did not, and grouping by concept silently merged concepts from different taxonomies. The prefix a document declares is now read wherever it declares it, and a derived segment already claimed by a different namespace is suffixed rather than shared. Real filings declare their namespaces on the instance root, so nothing changes for them — Apple FY2023, Microsoft FY2024 and Coca-Cola FY2023 return the same 1,640 concept strings over 6,138 facts as before, byte for byte. Separately, Axis.is_typed_dimension and Axis.typed_domain_ref were declared on the model and written by no code path at all, so every dimension read as explicit; xbrldt:typedDomainRef is now carried from the element declaration through the element catalog onto the axis. An axis declared only in an imported taxonomy still reads as explicit, because following xs:import into a remote taxonomy is a separate capability that remains open. (GH #1188, #1234, bead edgartools-0c1q.16)
Footnote-to-fact links resolved to nothing, and footnote text was truncated and duplicated at once. A footnoteArc's xlink:from names a link:loc whose href fragment is the fact id, and edgartools stored the locator label instead — so Coca-Cola's FY2011 10-K put six ids on its footnote that name no fact, no fact carried a footnote back, and get_footnotes_for_fact() returned nothing for a fact that has one. This resolved almost everywhere by coincidence, because 18,686 of the 18,699 footnote locators in the fixture corpus label a locator with the fact id verbatim; the 13 that do not are in Coca-Cola FY2011 and JPMorgan FY2012, and both are now fully linked. Footnote text was built from descendant <div>s whenever any existed, so direct-child and sibling text was dropped, and since findall('.//div') returns nested divs alongside their ancestors while itertext() already descends, nested text was emitted twice; a footnote in Golub Capital's Q1 FY2025 10-Q lost its opening sentence and its entire table while repeating one sentence, reading 187 words against 352. Text is now assembled in one ordered walk with a break after each block, so words no longer glue across boundaries. Footnote resources are also matched within their own footnoteLink rather than a document-wide dict, since two extended links may legitimately reuse a label — a collision returned a real footnote belonging to a different fact. Of the 123 footnotes across the three fixtures that exercise this, only the 3 with nested markup change; the rest are byte-identical. (GH #1169, #1190, #1230, beads edgartools-0c1q.15, edgartools-05gk)
calculation_linkbase() dropped a concept's second calculation parent, and the weight that came with it. Calculation relationships were stored one node per concept, so a concept rolling up into two totals in the same role kept only whichever edge the traversal reached last — and calculation_linkbase() documents its result as one row per parent-to-child relationship. Multiple parents are ordinary rather than exotic: Coca-Cola's FY2023 10-K returned 278 rows against 306 filed arcs, Apple's 194 against 215 and JPMorgan's 438 against 456, and each returns exactly its filed count now. The sign was the sharper half — Coca-Cola files OtherComprehensiveIncomeAvailableforsaleSecuritiesAdjustmentNetOfTaxPortionAttributableToNoncontrollingInterest into one total at +1 and another at -1, so the single row that survived reported a weight whose sign depended on traversal order. CalculationTree now carries all_arcs, one entry per filed edge, beside the all_nodes map that callers use for lookup and membership. The same error at the tree level is fixed with it: a role split across two calculationLink elements, which is legal XBRL, had its second element replace the whole tree built from the first, so relationships are now accumulated across every element and file before the trees are built. (GH #1184, #1238, bead edgartools-0c1q.14)
The dimensional model was not role-scoped, so a domain reused across roles kept only one role's members. Hypercubes were stored per extended link role, but axes and domains sat in dicts keyed on element ID alone with no role on either model, and the members of a domain were assigned unconditionally once per role. Apple's FY2023 10-K splits srt:ProductsAndServicesDomain two ways on the income statement and five ways in the revenue note; a single global object held whichever was parsed last, so one of those two roles was always answered with the other's breakdown. Axes and domains are now kept per role, reachable through xbrl.axes_for_role() and xbrl.domains_for_role(), and the existing xbrl.axes / xbrl.domains keep their shape as a union across roles rather than a truncation. Five further gaps in the same loop close with it: xbrldt:targetRole is followed across extended links, so a hypercube whose dimensions are declared in another role is no longer reported as having none; dimension-default arcs set Axis.default_member_id, without which a fact at the axis default cannot be told from an undimensioned one; Table.closed and Table.context_element are read from the all arc instead of being hardcoded; Domain.parent is written for a member that is itself a parent; and axes and members follow the filed order attribute, which was already being extracted and then discarded. Measured on Exxon's FY2022 10-K: 49 of 49 hypercubes now report the closed="true" they were filed with against 0 before, all 39 axes carry their default member against 0, and 37 domains carry a parent against 0. (GH #1171, #1194, #1219, #1225, #1235, #1236, bead edgartools-0c1q.13)
get_statement_facts() returned nothing for concepts that are demonstrably in the statement. A fact's statement membership was read from whichever presentation role the enrichment loop reached first and stored as a single scalar, but membership is a set — us-gaap:NetIncomeLoss is presented in the income statement, the cash flow statement, the statement of equity and more. Apple's FY2024 10-K returned 0 rows for NetIncomeLoss in CashFlowStatement, ComprehensiveIncome, StatementOfEquity, Notes and Disclosures, and 6 in each after the fix; statement row counts rise correspondingly, for instance StatementOfEquity from 25 to 63. The full membership is now accumulated once into an index — strictly less work than the per-fact tree scan it replaces — and by_statement_type() tests membership rather than equality. The scalar statement_type keeps its meaning as the primary statement, and the membership lists stay off the declared DataFrame schema. (GH #1242)
get_facts_with_dimensions() removed the dimension columns it selected rows on. The predicate matches a fact because it carries dim_* keys, then to_dataframe()'s projection dropped every one of them because include_dimensions was left falsy, so the caller got the right rows with the information that made them the right rows deleted. Intel's FY2024 10-K returned 1,135 correctly-selected rows and 0 dimension columns; it now returns the same rows with all 56. (GH #1243)
pivot_by_dimension() silently returned a table that was not a pivot. The lookup column was built by interpolating the caller's spelling, while the projected column is always normalised, so the natural QName form us-gaap:AwardTypeAxis missed and the method fell through to returning the plain unpivoted frame — a DataFrame of plausible shape that a caller who does not already know the pivoted layout cannot distinguish. All three spellings (dei_LegalEntityAxis, dei:LegalEntityAxis, LegalEntityAxis) now resolve to the same pivot, and an axis no fact carries raises ValidationError naming the available axes instead of returning an empty frame that reads as "no data". Both pivot methods also now keep colliding rows apart by unit where that is what distinguishes them, and report any collision they cannot resolve rather than quietly keeping one row of each. (GH #1223)
Unitless facts were classified as numeric, so numeric queries returned document metadata. float(value) ran on every fact unconditionally, so dei:DocumentFiscalYearFocus — a gYear with no unit — carried a numeric_value of 2024.0 and a query for facts valued between 2020 and 2030 returned it alongside real monetary facts. XBRL requires a unitRef on numeric items and forbids one on non-numeric items, which makes the unit an exact test where the element catalog is not reachable. Five to seven facts per filing are reclassified — CIK, fiscal-year focus, area code, auditor firm ID, postal code — with every genuine numeric fact untouched. numeric_value is what consumers read as "this is a numeric fact", so this reached well beyond by_value(). (GH #1220)
Inline XBRL facts from parse_html() carried no resolved context or unit. _extract_contexts and _extract_units ran namespace-aware XPath (//xbrli:context) against a tree built by lxml.html.fromstring(), whose elements carry the literal tag "xbrli:context" and no namespace URI, so the lookups matched nothing and every fact's .context and .unit came back None while the raw refs sat on the fact untouched. Matching on the local name instead resolves 180,747 of 180,747 context references and 169,255 of 169,255 unit references across the 60 iXBRL filings in the fixture corpus, where the previous count was zero of each. A <unit> built from xbrli:divide now resolves to USD/shares rather than to plain USD, and dimensional members are read from xbrli:scenario as well as xbrli:segment. (GH #1232)
A fact's value was the text before its first child, and nothing else. _get_fact_value read element.text, so descendant markup, tail text and the entire continuedAt chain were dropped and the truncated string looked like a complete value — the mechanism behind four separate reports. Relevant content is now assembled properly: descendants and their tails contribute, an ix:exclude subtree does not while the text following it does, escape="true" serializes child markup instead of flattening it, and continuation chains are followed in order with a cycle guard. ix:footnote and ix:continuation are resources rather than facts and are no longer routed through the fact extractor, which removes 451 blank fake facts from the same corpus. Since continuedAt is how issuers split long narrative disclosures, the truncation fell on exactly the text an LLM reads. (GH #1189, #1237, #1239)
Unsupported inline transformations passed human display text through as the fact value. The registry held five entries and matched them on an exact QName, so it missed even ixt:num-dot-decimal — the TR4 spelling that accounts for 68,880 of the 90,000 format attributes in the fixture corpus — and returned "Delaware" where the value is "DE" and "ten years" where it is "P10Y". It now implements Transformation Registry 1–4 and the SEC's ixt-sec extensions keyed on a version-independent name, covering every format those 60 filings use, and a format that cannot be applied is recorded on the fact and logged rather than silently yielding display text. Scale is computed with Decimal: applying it in binary float wrote the error into the lexical value, so a filed 0.7 at scale="-2" became 0.006999999999999999 and one filing reported 8199999999.999999 for $8.2 billion. (GH #1250, #1251)
Document.has_xbrl reported False on documents with thousands of extracted facts. The inline pre-process pass stores its facts on metadata.xbrl_data, while Document.xbrl_facts called a separate extractor that scanned the node tree for ix_tag metadata the inline pass never sets — two unconnected pipelines, so the public surface answered from one that was never populated. This is worse than missing data, because a working extraction looked like an absent feature and a caller correctly branched away from it. The property now reads the pass that is populated, and has_xbrl is True for all 60 iXBRL filings in the fixture corpus against zero before. (GH #1233)
Every XBRL duration was reported one day short. XBRL 2.1 reads a date-only endDate as the end of that day, so the context runs to 24:00 on it and the final day belongs to the period; reporting_periods[*]['days'] subtracted the two dates and dropped it. A calendar year of 2023-01-01 to 2023-12-31 measured 364, Apple's 53-week FY2023 measured 370 instead of 371, and a duration context covering a single day measured 0. The count now lives in one place, edgar.xbrl.core.duration_days, used for the stored field and for the two sites in period_selector that recompute a duration and compare it against that field. periods.py keeps its own exclusive count, because its bucket bounds were calibrated against it and a 53-week fiscal year sits exactly on the <= 370 bound - it no longer writes that local value back over the public field, which is what made the field mean different things depending on whether a statement had been rendered first. Measured over the 8 filings under data/xbrl/datafiles: 191 duration periods change their days value and nothing else changes - no period_type, fiscal_period, label or key, and no change to the periods selected for any statement. (GH #1247)
Table rows were dropped at random. <thead> rows were deduplicated by id(), but nothing kept the lxml element proxies alive — lxml frees a proxy once unreferenced and CPython reuses the address, so a tbody row could inherit a stale thead id, match it, and be skipped as "already processed". The row vanished from the table with no error. Measured on the ABRAMS Form 4 0000001923-04-000001: 1 to 22 divergences per 25 renders before, 0 in 100 after. Membership is now tested by identity against a list held for the whole function, so the collision is impossible by construction rather than dependent on GC timing. This is what made the regression lane fail at random. (bead edgartools-gf6v)
EntityFacts lookups reported that present data did not exist. The fact index was keyed by the qualified name a fact is tagged with (us-gaap:StockholdersEquity) and by the lowercased label, never by the bare concept name — so get_annual_fact('StockholdersEquity') returned None and warned "No fact found", while the value was present and correct. get_annual_fact('Assets') worked only because us-gaap:Assets happens to be labelled "Assets" and matched the label key by coincidence, which is what made this look intermittent rather than systematic: in Snowflake's company facts that coincidence holds for 4 of 339 concepts. The local name is now indexed too, in its filed case — a lowercased form would merge with existing label keys. Across AAPL, JPM and KO, 1,954 of 4,264 concept and label lookups that previously answered None now return a fact, with no existing answer changed or lost. get_fact() and available_periods() shared the defect and share the fix. (GH #1202)
A filtered or paged Reports collection returned reports whose content could not be read. Report.content fetches its R-file through reports._filing_summary._filing_sgml, and three of the four methods that return a new collection rebuilt it without that back-reference, so filter(), next() and previous() handed back reports with the right filenames and raised AttributeError: 'NoneType' object has no attribute '_filing_sgml' on .content. On Apple's FY2024 10-K a multi-row filter now reads R3.htm as the 79,891 characters get_by_category() always returned. Only the multi-row branch of filter() was affected — filtering to a single report returns a Report still attached to the original collection, which is why this stayed hidden. Deriving a collection now goes through one place, so the next method to return a subset carries the context by default. (GH #1191)
FactQuery.scale() was a silent no-op on every parsed fact. Transforms were applied to fact['value'], which holds the filed string, so scale()'s own isinstance(value, (int, float, Decimal)) guard rejected it and handed it straight back — while numeric_value, the float every consumer actually reads, was never transformed on any path. scale(1000), the example in its own docstring, changed nothing. A numeric fact is now transformed on its numeric value and both fields are written together so they cannot disagree; a non-numeric fact keeps being transformed on value, so text transforms still work and scale() leaves TextBlocks alone through the same guard. StitchedFactQuery duplicated the defect and shares the fix. (GH #1187)
FactQuery.to_dataframe() served a stale table after the query was narrowed. The cache was keyed on the column projection alone, so the first to_dataframe() answered every later one: after .by_concept('us-gaap:Assets'), execute() returned 2 rows while to_dataframe() still returned the original 1,075, with nothing to say the two disagreed. FactQuery is a fluent mutable builder whose methods return self, so one object describes a different population after each call, and the key now covers the query configuration as well as the projection. Identical queries are still served from the cache. StitchedFactQuery duplicated this one too. (GH #1186)
A disclosure whose subject is notes payable was routed to notes(). get_all_statements() falls back to keyword matching when a role carries no FilingSummary menu category, and that fallback tested the bare substring "note" before it tested "disclosure". "Note" names a financial-statement section and a debt instrument, so Oracle's four Disclosure - NOTES PAYABLE AND OTHER BORROWINGS roles (10-Q 0000950170-23-047713) reported type="Notes" / category="note", came back from xbrl.notes(), and were absent from xbrl.disclosures(). The ambiguous word no longer outranks a role that states what it is - by its Disclosure category marker, or by the concept it hangs from (us-gaap_DebtDisclosureAbstract) - unless the definition names the notes section itself, so Notes to Consolidated Financial Statements and Note 1 - Organization still classify as notes. Roles whose definition does not contain "note" are classified exactly as before. (GH #1207)
Filing.sections() is chunked by edgar.documents instead of the legacy parser. It backs Filing.search(), and its old backend is deleted in 6.0, so it needed a replacement rather than a deprecation. Chunks are cut at headings with each table its own chunk, capped so no chunk swallows a whole Item. Measured across 41 era-stratified fixtures the two chunkers index the same words to within 0.10%, and across 375 phrase queries mean recall@5 is unchanged at 99.2%. (bead edgartools-07lk.3)CurrentReport.doc warns; it was the last un-warned public route into edgar.files. It returns the legacy ChunkedDocument, while CompanyReport.doc on every other report class returns the modern edgar.documents document — so the same attribute name yields two unrelated types depending on the form, and a survey starting at the base class reads .doc as already migrated. Use .document for the parsed document, or .items / report[item] for item access. Measured across 133 era-stratified fixtures and 2,128 item lookups, no internal code path reaches the legacy parser through the report classes any more. (bead edgartools-07lk.3)6.0 deletes the edgar.files package, and these three were the user-reachable modules with no deprecation, so code importing them would have got no not…
Thirteen fixes, led by two clusters: period identity now comes from the duration a fact covers rather than the SEC's filing-focus label, and three typed fields on Form D / Form 144 that were silently wrong.
Upgrading: Form144.nothing_to_report is Optional[bool] now, so callers comparing it against the string "N" need updating; period_length is honoured rather than ignored, so callers who passed it were reading annual figures under a quarterly label and will see different numbers. edgar.effect and edgar.form144 still work but warn — they now live at edgar.offerings.effect and edgar.ownership.form144, and edgar.files.markdown / .tables / .text warn ahead of 6.0 removing them.
pip install --upgrade edgartools
edgar.muniadvisors is a package rather than a single module. The MA-I parser moved to edgar/muniadvisors/core.py, matching edgar/ownership/ and edgar/entity/. Nothing changes for callers: a module converting to a package of the same name keeps the import path identical, and all 19 of its public names still resolve to the same objects — __all__ named only MunicipalAdvisorForm, so a naive re-export would have dropped the other 18. (bead edgartools-07lk.12.1)[tool.hatch.build].include opens with an unrestricted edgar/**/*.py, so every install also downloaded the evaluation harnesses, entity training scripts and demos under edgar/ — about 8,900 lines across ai/evaluation/, ai/examples/, entity/training/ and thirteenf/demo_comparison.py. They are excluded now; ai/exporters/ stays, since export_skill is public API. A regression test asserts that no shipped module imports any excluded path, which is the property that makes the exclusion safe rather than the exclusion itself.CurrencyConverter reported "no rates found" when extraction had crashed, and an MA-I summary whose disclosure block failed to parse read as a clean compliance record. All three now log a warning and carry the failure — XBRL.period_validation_unavailable, CurrencyConverter.extraction_warnings, and an explicit line in the advisor summary. No exception propagates that did not before._require_element() helper raising ValidationError with the element name and the namespace hint that explains most of these failures. ValidationError IS-A ValueError, so anything catching the old raise still catches this one.edgar.files.markdown, edgar.files.tables and edgar.files.text now warn. 6.0 deletes the edgar.files package, and these three were the user-reachable modules with no deprecation, so code importing them would have got no notice at all — just an ImportError on upgrade. They warn for external callers and stay silent for edgartools' own internal use. edgar.files.styles is deliberately left unwarned: parse_style runs thousands of times per document and the frame check is not free, and its entry points (Document, SECHTMLParser) already warn.edgar.effect and edgar.form144 moved, with the old paths kept until 6.0. Effect now lives in edgar.offerings.effect, beside the S-1/S-3/S-4, 424B and Form C/D parsers — a notice of effectiveness is a registration-lifecycle document. Form144 now lives in edgar.ownership.form144, beside the Forms 3/4/5 parsers — Form 144 is an affiliate's notice of a proposed sale. Neither name was in edgar.__all__ nor re-exported at the top level, so the module path was their whole public surface; both old paths still work and emit a DeprecationWarning naming the new one. The objects are the same objects, not copies, so isinstance checks against either path agree. filing.obj() routes to the new paths, so ordinary use never triggers the warning. (bead edgartools-07lk.12.1)period_length on EntityFacts.income_statement() and .cash_flow_statement() warns before 6.0 removes it. period= is the supported spelling and the only one that can also ask for 'ttm'. The parameter is honoured now rather than ignored — see Fixed below. (GH #1177)edgar/files/html_documents_id_parser.py. 687 lines and three classes (AssembleText, ParsedHtml10K, ParsedHtml10Q) with no callers anywhere in the library or the tests. The id_parse_document path it once served survives only in comments.edgar/tools/ (an empty package), edgar/analysis/ and edgar/xbrl/analysis/ (both already emptied of their modules, leaving only stale caches), and edgar/reference/financials.py (a seven-line scratch script with a __main__ block and no importers). Nothing in the library or the tests referenced any of them.cash_flow_statement(period='quarterly') reported Snowflake's PaymentsToAcquireInvestments for Q3 FY2022 as 3,042,396,000, the nine-month cumulative total, against a discrete quarter of 1,053,763,000 — a 2.89x overstatement. Two causes: comparative re-filings were discarded as inconsistent, which removed the six-month operand the quarter is derived from, and the quarterly path had no duration filter to mirror the annual one. (GH #1180)TTMCalculator.quarterize() returned periods that were not quarters. Deriving a quarter pairs a cumulative fact with an earlier one, and when the two came from different fiscal cycles the subtraction produced a real number over a nonsense interval: Snowflake's goodwill yielded -45,711,000 across 457 days labelled Q2 FY2025. Everything quarterize() returns is now checked to span a quarter. (GH #1197)latest_periods(1) returned three years of facts. Periods were keyed on (fiscal_year, fiscal_period), which is the SEC's filing focus rather than the period itself, so the 2022, 2023 and 2024 comparatives in Logistic Properties' 20-F — all tagged fy=2024, fp=FY — counted as one period. Periods are now keyed on the dates the facts cover, and annual-ness comes from the duration. (GH #1185)discrete_quarters=True, a concept present on the 12-month period but not the nine-month one had nothing to subtract and kept its cumulative value, while the period was relabelled a quarter regardless. Meta's FY2024 "Deferred income taxes" changes concept between the 10-K and the 10-Q, so $(4.738)B was shown as Q4. A quarter that cannot be derived is now dropped, so the cell reads empty rather than wrong. (GH #1179)recipient.address.zipcode was always None: the parser looked for a child element named 30361 — the sample ZIP from the code's own docstring, pasted in where the tag name belongs. The four recipients in the 1685 38th REIT notice now read 30361 as filed. Related-person addresses were never affected. (GH #1193)FormD.is_new was inverted, so every base notice rendered as FORMD/A and every amendment as FORMD. The field was assigned the filing's isAmendment answer, which asks the opposite question — AP Fund IV, a base D, reported is_new=False. to_context() now takes its heading from submission_type, so the heading cannot contradict the field beside it. (GH #1192)Form144.nothing_to_report returned a truthy "N". It carried the raw SEC flag against a bool annotation, and bool("N") is True, so if form.nothing_to_report: read a notice reporting a sale as reporting none — accession 0001958244-23-000454 has one prior-sale row. Y/N now parse to True/False, and an unanswered flag is None rather than a default. Callers comparing against the string need updating. (GH #1195)Document.parse(...).to_markdown() emitted four warnings, two naming modules the caller never touched. It now skips only the warning's own module.period_length was accepted, documented, and never read. EntityFacts.income_statement(period_length=3) and .cash_flow_statement(period_length=3) returned the annual statement with no warning — the parameter appeared nowhere in either method body and had not since the Facts API landed. It now selects the period it names, and a value contradicting period= or annual= raises ValidationError instead of being silently resolved. Numbers can change for callers who were passing it: they were reading annual figures under a quarterly label. (GH #1177)XBRLS.get_statement(use_optimal_periods=False) crashed instead of returning a statement. The non-optimal path appended the (statements, role, statement_type) tuple from XBRL.find_statement() to the list handed to StatementStitcher, which then called .get() on it and raised AttributeError: 'tuple' object has no attribute 'get'. It now calls get_statement_by_type(), the same accessor the optimal path uses, and skips filings that lack the statement type. (GH #1173)XBRLS.query(standardize=False) still standardized. The option was stored under the keyword standardize and read back as standard, so it never reached get_statement() and every query came back with standard-concept mapping applied. Both spellings are now accepted. (GH #1172)FactQuery.transform() and .scale() mutated the shared fact cache. Transformed values were written back into the row dictionaries returned by get_facts(), which come from the facts view's cache, so .scale(1000) scaled the cache itself and a second identical query returned values divided by a million. Rows are now copied before transformation. StitchedFactQuery had the same defect. (GH #1175)find() did not recognize tickers containing X. The ordinary-ticker pattern was ^[A-WYZ]{1,5}([.-][A-Z])?$, a character class excluding X in every position rather than only the trailing position that marks a mutual fund, so find("XOM") returned CompanySearchResults instead of the Company that Company("XOM") resolves. The ^[A-Z]{4}X$ fund pattern is now tested first and the ticker pattern admits the full alphabet. (GH #1178)$3. Many filers render a fee-waiver footnote as its own single-row table, and its prose contains "waive its management fees", so _classify_table labelled it operating_expenses on the keyword alone. That inflated the fee-table count, which routed the expense example to the single-column extractor, which parsed the digit out of the header string 3 Years — Ocean Park High Income ETF reported expense_1yr=3 against the filed $109/$381. Prose-only tables no longer fall through into the data classifiers, fee values are shape-validated without rejecting SEC footnote formats, and column alignment is preserved for indented operating-expense and shareholder-fee tables. Contributed by @sambai-dev. (GH #912)Six fixes, two of them contributed by @hashd1ve (#1144) and @biautomator (#1151). Thanks to both.
Six fixes, two of them contributed by @hashd1ve (#1144) and @biautomator (#1151). Thanks to both.
Upgrade note: FundSeries.get_filings() now returns that series' own filings rather than the whole trust's, so counts drop and filings[0] changes — the previous result could belong to a sibling fund. Fund.get_filings() without series_only is unchanged.
detect_page_breaks() and mark_page_breaks() warn before 6.0 removes them. No replacement: the edgar.documents parser treats page-break <hr>s and page-number containers as print chrome and discards them, so page-break positions are not part of the supported document model. Internal callers stay silent, so a suite running -W error is unaffected unless it calls these itself.get_financials() no longer goes quiet on a filing with no XBRL. It returned a truthy Financials whose income_statement(), balance_sheet() and cash_flow_statement() all answered None — Company(104599), Circuit City's 2008 10-K, is the case. It now emits a FutureWarning and will raise XBRLFilingWithNoXbrlData in 6.0; set EDGARTOOLS_STRICT_ERRORS=1 for that behaviour today. filing.xbrl() still returns None quietly, which is a true absence rather than a failure.get_revenue() returned a years-stale value after a company migrated its XBRL tag. The first variant with any matching fact won, regardless of period, so NVIDIA answered FY2022 (26,914,000,000) instead of FY2026 (215,938,000,000). Candidates are now gathered across all variants and ranked by recency. Every standardized getter built on the same helper was exposed. (GH #1149)get_revenue() returned the ASC-606 slice for insurers and banks. MetLife tags both the contract-revenue slice (2,436,000,000) and consolidated Revenues (77,084,000,000) for FY2025, and the slice ranked first — a 32x understatement, with an income statement showing less revenue than operating income. A same-period candidate that dwarfs the ranked pick is now taken as the consolidated total. Revenue only: net income's variants are not slices of one another. The statement builder held a second copy of the priority list and now shares one threshold with the getter.edgar_trends returned whatever concept contained the word. Concepts were matched as substrings, so "Revenue" also matched CostOfRevenue and DeferredRevenue; rows were truncated before the period-type filter, so the correct annual fact could be cut from the window; and quarterly facts labelled FY entered annual series under the same year label as the real annual figure. Apple's revenue read as falling from $383B to $7.7B. Concepts are now matched exactly, filtered before truncation, and period type comes from the reporting window rather than fiscal_period. Revenue trends take the same consolidated-total cross-check as get_revenue(). (GH #1138)FundSeries.get_filings() delegated to the fund company, and a trust files one report per series under one CIK, so every series returned the same list and filings[0] belonged to whichever sibling filed last: Vanguard's Extended Market Index fund, which excludes the S&P 500 by construction, was handed the 500 Index fund's portfolio. It now resolves its own series through SEC browse-edgar and returns empty rather than falling back to the trust. Fund.get_portfolio() and get_latest_report() took the same trust-wide path and are fixed with it. Fund.get_filings() without series_only still returns the whole trust. (GH #1143)year and quarter accept a list on the fund-series path. A list was untranslatable, silently dropped, and the caller received the series' entire unfiltered history. An unusable pair now raises rather than answering with everything. (GH #1143)edgar_ownership called a fund's unparsed positions "nan", and edgar_screen wrote a bare NaN into its JSON. A missing DataFrame cell is truthy, so if issuer: admitted it and str() rendered it; where nothing guarded it, the NaN reached to_json as a literal no strict parser reads back. One cell reader now lives in tools/base.py. (GH #1137)find_bdc("Ares") — the example in its own docstring — returned nothing. Names were scored with whole-string fuzz.ratio against a threshold of 50, so a short query against a longer registered name missed: ARES against ares core infrastructure scores 28.6. Candidates already come from an exact word index, so they are now scored with token_set_ratio. (GH #1145)edgar_fund(action="portfolio")'s total_value summed share counts, not dollars. None of the candidate columns existed on FundReport.investment_data() — the dollar column is value_usd — so it fell through to balance and summed raw units. VFINX reported a total_value of $14.4B against a top-5-holdings sum of ~$366B. Per-holding records were unaffected. (GH #1148)One column per quarter.