Package: exporter

LinkState is one trace link as the views present it. Title is the target'shuman-readable title; Pin/Direction describe the authored version pin andits health against the target's current Version.

NodeState is the presentation state of one requirement, adapted from graphanalysis. Views render these states instead of inferring coverage orverification from link presence.

Analysis is the semantic layer shared by every HTML view: coverage,verification, and trace link health adapted from the graph so the exportcannot disagree with reqmd check.

NewAnalysis adapts graph results into presentation state. titleMap supplieslink titles; missing entries fall back to the bare ID.

realRequirementIDs returns the IDs of authored requirements. Synthetic nodes(RESULT: pseudo-requirements and synthesized test cases) are evidence, notrequirements: they have no page and must not appear as trace links.

nodeState assembles one requirement's presentation state.

evidenceItems converts graph evidence into the exporter's render type.

upLinks keeps the authored trace refs that point at a real requirement (orare unresolved, so broken links stay visible).

CoverageSummary counts the coverage states of the given requirements.

VerdictSummary counts verification states over the eligible measures of thegiven requirements. Unverified eligible measures are counted as NoVerdict,so percentages always divide by the eligible total.

LinkSummary counts trace link health across the given requirements.

isSyntheticID reports whether an ID belongs to a synthesized node (resultpseudo-requirement or test case) rather than an authored requirement.

idsOf returns the given document's requirement IDs in document order.

DocBoundary describes a document directory's position in the V-model chain.IsRoot is true when the document has no upstream sources (or isdeclared external). IsLeaf is true when no other document referencesit as an upstream source.

ComputeDocBoundaries returns the boundary for each document directory path.A document is a root boundary when it has no upstream sources, or isdeclared external. A document is a leaf boundary when it is notreferenced as an upstream source by any other document.

browserFixture builds a document whose analysis exercises every state:covered, missing coverage, a failing verdict, and a draft requirement thatthe default status filter hides.

exportBrowserFixture writes the fixture page and returns its path.

browserHarness drives the exported page and reports results through thedocument title, which --dump-dom prints.

findChrome returns the first usable headless-capable browser.

runHeadless loads one exported page through file:// with the given harnessappended and returns the decoded RESULT payload.

stringSet converts a decoded JSON array of strings into a set.

TestHTMLExportBrowserSmoke is the end-to-end guard for workflows that onlyexist in the client: offline load, filtering, sorting, the on-demand views,the bounded matrix, and keyboard operation.

TestHTMLExportBrowserDeepLink verifies a #REQ-ID deep link opens Documentmode and reveals the target.

TestHTMLExportBrowserRestoresURLState verifies a shared URL reproduces theexact filtered view, including values with commas and colons.

TestHTMLExportBrowserExplicitEmptyFilters verifies ?f= keeps the documentdefault status filter cleared after reload.

TestHTMLExportBrowserMobileOutline verifies the outline becomes a drawer atmobile width and the toolbar stops sticking when vertical space is tight.

TestHTMLExportBrowserMatrixBound verifies the full matrix refuses to drawbeyond its cell limit and points the reviewer at the trace neighbourhood.

TestHTMLExportBrowserLinkSemantics covers the navigation fixes: externaltrace nodes are links, requirement references inside analytical views openthe Document view, and the shortcuts dialog receives focus.

TestHTMLExportBrowserContextJumps verifies that jumping from a coveragequeue keeps the reviewer's filters and that the result count names thereason requirements are hidden.

CSV exports document content as comma-separated values. When verdictsare set via SetVerdicts, extra columns (Verdict, Verdict Source, VerdictCases) are appended per row.

BuildTitleMap builds a reqID → title map across all documents, for use astrace-link labels. Requirements with an empty title are omitted.

RenderContext holds precomputed document chain and link resolution datafor cross-document HTML traceability. It is shared by the export and serve commands.

NewRenderContext precomputes the cross-document link map and trace chain data.

ChainCard represents a single document in the chain graph visualization.

ChainTier is a horizontal row of docs at the same trace distance.

DocChainGraph is the full tiered document graph for a single current doc.It is the source of truth for the Confluence-Flow visualization.

tierLabel formats a chain tier header: the bare noun at one level ofdistance, otherwise " · N levels ".

eachChainCard calls fn for every card in the graph's upstream anddownstream tiers (never the current card). Shared by the three cardpath-rewriting functions below.

ResolveChainOutputs replaces every non-current card's Path — which therender context seeds with the flat output file name (e.g."01-stakeholder-requirements.html") — with the real output path for thatdocument, as reported by outPath(dirName). Callers know where each pageis actually written (flat into an output dir, or into each document's owndir), so they supply the mapping. Relative card paths are then producedby RelativizeChainGraph.

RelativizeChainGraph rewrites every non-current card's Path to berelative to baseDir. This is a free function (not a method) so callersin other packages can use it without exposing the unexported fieldsof DocChainGraph.

BasenameChainGraph collapses every non-current card's Path to itsbasename. Useful when all HTML files are served at a single URL root(the dev server).

makeDoc constructs a minimal model.Document with the given path, title andoptional x-reqmd block. Requirements are empty (the chain graph doesn't needany) and schema is the bare map containing the title.

makeXReqmd is a tiny constructor for XReqmd pointer values.

ExportGraph writes all parsed requirements as a ladybugdb database at outDir.Each requirement becomes a node with id and file properties. Every tracelink becomes a TracesTo edge between Requirement nodes.

The resulting database is at /reqmd-graph.lbug/ and can be queriedwith the ladybug CLI or SQL ODBC driver.

escapeCypherString escapes single quotes by doubling them for safe inlineCypher string literals.

go:embed static

EvidenceItem is one verification result contributing to a measure'srolled-up evidence set, ready for rendering in exports.

VerdictInfo carries the rolled-up verification verdict for a singlemeasure: the outcome, a representative source file, the case keys thatcontributed (for CSV), and the full evidence list (for HTML).

HTML exports requirements as standalone HTML: a server-rendered Documentview plus on-demand analytical views driven by one shared view model.

schemaTitle safely extracts the title field from a document's parsedschema. Returns "" when Schema is nil, not a map, or has no string title.

docTitle returns the title used for rendering a document page or card:the parsed file h1 (document title), falling back to the schema title.

bodyRenderer renders requirement body markdown to HTML.

renderMarkdown renders markdown content to safe HTML.

pageBufferSize is the buffered-writer size for one exported page.

renderModelJSON writes the shared view model the client renders from.

formatAttrLabel converts a snake_case or kebab-case attribute name toa human-readable title-case label (e.g. "disposition-reason" → "Disposition Reason").

formatAttrHTML renders an attribute value as rich HTML for the card grid.Arrays become inline pill tags. Nested maps become a compact JSON block.All output is already HTML-escaped where needed by the caller.

renderChainGraph writes the Document Stack chain as HTML to b.The output is pure HTML/CSS — no JavaScript.

renderChainZone writes one horizontal zone (upstream or downstream)with its header and stacked tiers.

renderChainSeparator writes the chevron rule between chain zones.

renderChainCard writes one (or for the current doc) card.

analysisFor builds the presentation analysis from a document set.

exportPage renders one document and returns the HTML.

pageModelOf extracts and decodes the embedded page model.

TestHTMLExportOfflineContent is the Phase 0/1 regression guard: primarycontent is visible by default, and nothing loads from a CDN.

TestHTMLExportModelIsHTMLSafe pins that author-controlled values in theembedded JSON cannot break out of the script element.

TestHTMLExportModelOmitsBodies verifies the duplicate-data fix: the readingview is the only representation of requirement prose.

TestHTMLExport_UndeclaredAttributeRendered verifies undeclared attributesstill render (flagged) and feed the filter menu.

TestHTMLExportTraceLinkState checks pins, versions, and link health in theshared model and the server-rendered links.

TestHTMLExportDownstreamLinkDirection pins that a requirement's inboundlinks appear as Down (who traces to it), never as Up.

TestMatrixTooltipDirection pins the client's direction-specific matrixtooltip, the regression from the old downstream matrix.

TestHTMLExportCoverageStates checks that coverage comes from graphsemantics, not link presence: a draft downstream does not cover.

TestHTMLExportNoVerdictIsFirstClass verifies unverified eligible measuresshow a no-verdict chip and are marked eligible in the model, so the clientdenominator includes them.

TestHTMLExportEvidenceList keeps the rolled-up evidence rendering.

TestHTMLExportLargeDocumentBound guards against quadratic hidden markup:a large document's output grows linearly and emits no matrix/table DOM.

TestProjectIndex checks the project home page content and queues.

JSON exports the parsed tree as NDJSON tables. Verdicts set viaSetVerdicts are emitted by ExportResults; the graph passed to theExport* methods is the single source of truth for trace edges,verdicts, and findings.

promotedAttrs maps a built-in requirement attribute to the top-levelNDJSON key it is promoted to, so warehouse SQL can read it directlyinstead of digging into the attrs envelope.

jsonRequirement is one row of requirements.ndjson. Built-in attributesare promoted to top-level keys; the remaining declared attributes stayin Attrs so nothing is lost.

Every fixed key is present in every row (null when unset) so warehouseschema inference — DuckDB, Spark, Snowflake's infer_schema — sees thefull column set regardless of which row it samples. Attrs is the oneintentional exception: it is the per-document attribute envelope and isnull on nodes that declare no extra attributes.

jsonTrace is one row of traces.ndjson: one declared trace reference.To is the resolved bare requirement ID when resolved, otherwise the rawdeclared target. Ref always holds the raw form as written, and DocIDholds the doc-id qualifier when the ref uses one. Resolved is falsefor a dangling reference; such a ref also appears in findings.ndjson, soconsumers can count or filter it themselves.

jsonCase is one contributing verification result in a measure'sevidence set.

jsonResult is one row of results.ndjson: one rolled-up measure verdict.

jsonFinding is one row of findings.ndjson, mirroring the reporter'sCheckResult fields. ERROR and WARNING both appear; consumers filter.

walkJSONNodes visits every node in document order, depth-first.

remainingAttrs returns a copy of attrs without the promoted built-inkeys, so a value never appears both top-level and in the envelope.

measureVerify returns the verify attribute of an authored measure, or"" when the ID is unknown or the attribute is absent.

jsonDocs builds a small two-document tree: one upstream with a version,one downstream that traces to it with a pin and a dangling ref. Enough toexercise every table.

filterSet is the per-attribute sorted unique values collected acrossall requirements, used to populate the search toolbar filter menus.

pageLink is one trace link in the client model. Href is resolved for thepage that embeds the model; Unresolved links carry the authored Ref only.

pageNode is one content node in the shared view model. Body and rationaleare deliberately absent: the server-rendered Document view is their singlerepresentation and analytical views read them from the DOM.

pageDocMeta carries the document facts the client toolbar and views show.

pageModel is the single view model for one exported document page. Everyanalytical view, count, aggregate, and outline is derived from it, and thefilters are the only filtering input.

renderData is the single O(n) intermediate produced by buildRenderData: itpowers the server-rendered TOC, the filter menus, and the view model.

buildRenderData runs a single O(n) pass over the document content tree toproduce the per-document data the rendering helpers need: the filter set,the per-status breakdown, and the stable DOM anchors.

Every node gets a stable anchor: requirements anchor on their ID,containers and info items on a generated "info-N" anchor (N = documentorder sequence). The same anchors are used by the TOC and by the renderedDOM so scroll-spy stays consistent.

buildPageModel turns the document tree and graph analysis into the sharedview model. resolveLink maps a requirement ID to a cross-document href; anil resolver yields same-page fragments. propOrder is the schema propertyorder used to order the table's attribute columns.

pageMeta fills the document-level model fields.

modelHref resolves cross-document links for the model, falling back to anempty href (the client uses the same-page anchor).

pageNodeFor converts one content node, attaching graph-derived state forrequirement nodes.

attributeColumns returns every attribute present on at least onerequirement, in schema property order followed by undeclared extras A→Z.trace and status are excluded: both have dedicated presentation(trace link columns and the status chip/column).

attrsForModel copies a requirement's attributes for the client model,dropping trace (represented by the Up links).

pageLinks converts analysis links for JSON. Same-page links resolve to"#id", which needs no href in the model; links to another document keeptheir page-relative target so the client can navigate to them.

sortedKeys returns the keys of a string-keyed map in sorted order.

ProjectDoc is one document row on the project home page.

ProjectIndex is the project-level home page: document discovery,project-wide health, actionable queues, and cross-document search.

NewProjectIndex builds the project home model. hrefFor maps a documentdirectory path to the exported page href.

queueItem is one actionable finding on the project page.

filepathBase returns the last path element, tolerating URL separators.

queueSpec is one actionable queue: a title, an explanation, and the matcherthat decides which requirement states belong to it.

projectQueueSpecs is the fixed set of exception queues on the project home.

missingCoverageNote matches requirements whose coverage expectation is notmet by an approved downstream requirement.

unresolvedCoverageNote matches requirements expecting coverage from anunknown document-id or level.

missingVerdictNote matches eligible measures with no result.

failingVerdictNote matches measures whose rolled-up verdict is fail.

brokenTraceNote matches requirements with unresolved, outdated, or predatedoutbound trace links.

attentionClass returns a modifier class when a count should stand out.

projectIndexCapacity pre-sizes the cross-document search index.

renderProjectJSON writes the cross-document requirement index used by theproject search.

The HTML export has one server-rendered reading view and three analyticalviews rendered on demand from the embedded page model:document — quiet card-per-requirement reading view (works without JS) table — dense sortable grid with a column chooser and row expansion trace — selected-requirement trace explorer with a bounded matrix coverage — actionable coverage, verification, and link-health lists

Every view, count, aggregate, and outline reads the same filteredrequirement set, so the toolbar means one thing.normalizeAttrValue collapses []any / []string / scalar attributevalues into a uniform []string, dropping nil entries and nested maps.

buildFilters collects per-attribute unique values across all requirementsinto a filterSet. Attributes with fewer than two distinct values are omitted(single-value filters are not useful).

collectFilterValues gathers the distinct values of every filterableattribute. The trace attribute never becomes a filter; status values areskipped when the document opts out of the status lifecycle.

orderFilterValues returns one attribute's values in display order: status iscanonical (draft, approved, declared additions, then A→Z); everything elseis sorted A→Z.

sortedStatusValues returns the given status values in canonical order:built-ins [draft, approved] first, then declared additions A→Z, then anyunrecognized values A→Z.

orderStatusFilterValues pulls the declared additions from doc.XReqmd(nil-safe) and returns the values in canonical order.

statusBreakdownItem describes one status value and its count.

countStatusBreakdown tallies status values across all requirements in adocument, in display order: built-in (draft, approved) first, thenadditions A→Z. Values not declared are treated as StatusDefault.

statusColor returns the CSS color expression for a status value. Built-insreturn a CSS variable so the active theme selects the right value.

effectiveStatus returns the requirement's status, defaulting to approved.

countItems tallies non-requirement nodes (containers + info items).

Content headings start at h2 (h1 is the document title) and stop at h6.

Content headings start at h2 (h1 is the document title) and stop at h6.

itemHeadingTag returns the HTML heading tag for an item's native markdownlevel. Top-level items start at h2 (h1 is the document title); nested itemskeep their native depth, floored one level below their parent.

linkHref resolves a requirement ID to an href for this page: thecross-document resolver when available, else a same-page fragment.

renderDocHead writes the + + block. Thestylesheet and client script are inlined: there is no runtime CDN request,so an exported page is fully offline. The theme is applied by a small headscript before first paint to avoid a light/dark flash.

renderBodyOpen writes the , the app bar, the shell open, and theserver-rendered outline.

renderTOC writes the server-rendered outline. It works without JavaScript;the client hides items that fall outside the filtered requirement set.

renderDocHeader writes the document title, count line, status breakdown,boundary badges, and the frontmatter description. The trace chain lives atthe end of the Document view so it can never push a reader past the toolbar.

renderDocChain writes the optional document chain navigation block.

renderToolbar writes the view switcher, search field, active-filter chipsplaceholder, and the per-attribute filter menus.

renderDocumentView writes the server-rendered reading view: one card perrequirement, containers as collapsible sections, and info items as plainblocks. Contents are visible by default (no x-cloak), so the reading viewsurvives a JavaScript failure or an offline file:// open.

renderEmptyState writes a consistent empty-state block.

renderContainerBlock writes a collapsible section for a container node.

renderInfoBlock writes a plain block for a leaf info item.

renderCard writes one requirement card. Semantic state is carried on dataattributes (used by CSS and the client) plus at most a few state chips, sothe reading view stays quiet.

renderCardHeader writes the ID/title heading and the anchor link.

renderCardFlags writes the state chips. Only exceptional or decision-relevant states are shown: draft/custom status, coverage problems, and anyverdict. A covered requirement with an approved status adds no chrome.

statusChip renders a chip only for non-approved statuses: an approvedrequirement adds no chrome.

coverageChip renders coverage problems only (missing or unresolved); acovered or not-required requirement adds no chrome.

verdictChip renders the rolled-up verdict, or an explicit no-verdict chipfor an eligible measure when results were loaded.

attrRow is one rendered attribute cell in a requirement card.

attrRows returns the attribute cells to render, in display order: theschema property order first, then any remaining attributes actually presenton the requirement sorted A→Z. The trace attribute is skipped.

renderAttrGrid writes the attribute definition list.

renderRationale writes the rationale block when present.

renderEvidenceList writes the rolled-up verification evidence.

renderTraceLinks writes the upstream/downstream trace columns with theauthored reference, title, pin, and resolution state.

renderTraceColumn writes one direction column.

renderTraceLink writes one link with its pin/version state.

renderPalette writes the command-palette overlay. Overlays are the onlycontent hidden until the client boots.

renderShortcuts writes the keyboard-shortcut dialog.

renderBodyClose writes the end of the shell and document.

testDoc builds a document with a container, requirement children, and aninfo item so the render data covers every node kind.

VerdictInfoFromGraph converts a graph-side rolled-up NodeVerdict into theexporter's render type, carrying the outcome, representative source file,the contributing case keys (for CSV), and the evidence items (for HTML).