Every export

The complete public surface of truecopy, by entry point. Each module is also its own import, so a project that wants one pays for one.

Every module is its own entry point (truecopy/signature loads signature and nothing else), and the package is side-effect free, so a bundler drops what you do not import. The barrel (truecopy) re-exports all of it.

truecopy/table


`readTable(file, options?)`	rows, warnings, boundaries, document (the two-line path)
`Table`	the shape it returns

truecopy/open


`openDocument(file, options?)`	the door: caps, deadline, engine released
`UnreadableDocument`	carries `reason` so your application writes the sentence
`Unreadable`	`'empty' \| 'too-big' \| 'no-text' \| 'too-slow' \| 'binary' \| 'no-engine' \| 'not-opened'`
`DEFAULT_LIMITS`, `Limits`	20 MB, 40 pages, 30 s
`OpenOptions`	the limits, plus `workerSrc` and `pdfjs`
`PdfEngine`	what this library needs of an engine, structurally
`withDeadline(promise, ms?)`	bound any read in time
`positionedItems(items)`	the pure step between an engine and this library’s data

truecopy/layout


`cellsOf(page, boundaries?)`	a page as a table of cells
`placesOf(page, boundaries?)`	where each cell is, laid out identically
`placeOf(items, page)`	where any group of items sits
`boundariesFromRecurrence(rows, support?)`	keep only the x that come back row after row
`boundariesFromAnchors(anchors)`	cut halfway between anchors you trust
`columnBoundaries(items)`	the raw spread of x
`columnAt(x, boundaries)`	which column an x falls into
`rowToCells(row, boundaries)`	one row, cut
`rowsFrom(items, boundaries)`	items into rows, by proximity of baseline
`pageFrom(n, w, h, items)`	items into a page
`documentFrom(pages, origin, name)`	pages into a document
`documentFromText(text, name)`	a paste, a CSV, an export

truecopy/notation


`readNumber(raw)`	how a page writes a figure, `null` rather than a guess
`findNumbers(text, decimals?)`	every figure on a line, in order
`isOnlyNumber(text, decimals?)`	a cell of figures, not prose quoting one
`numberToken(decimals?)`	the pattern, to compose with
`readDate(text, notation)`	a delimited date, in the stated order
`readLeadingDate(text, notation, yearWhenMissing?)`	the date a line starts with, and its length
`decimalMarkOf(text)`	which mark the document uses, `null` when nothing settled it
`wellGrouped(written, mark)`	is this number grouped the way its document groups thousands
`accentFree(text)`	lower-cased and stripped of accents, to compare a label
`Notation`, `FoundNumber`, `LeadingDate`, `DecimalMark`

truecopy/records


`recordsFrom(rows, options?)`	which rows belong to the same record
`spineWidthOf(rows)`	how many filled cells a complete row carries, measured
`Records`	`records`, `loose` and `findings` - no row is ever grouped away
`RecordBlock`	the `spine` row, and every row index it carries
`RecordFinding`, `RecordDoubt`	`'spine-not-sharp' \| 'no-spine'`, with its numbers
`RecordsOptions`	`spineWidth`, `reach`

truecopy/labels


`labelledValues(rows, isLabel, isValue, options?)`	the cells that could be a label’s value
`columnOfHeader(rows, isHeader)`	the column a header sits in, `null` when several match
`Labelled`, `Candidate`, `Cell`	the label, its candidates closest first, and `raw`
`LabelOptions`, `Look`	`reach`, and `'row' \| 'column' \| 'both'`

truecopy/signature


`findRowAnomalies(rows, options, judged?)`	the rows that break the table’s shape
`thresholdsFor(kinds, share)`	one share for every kind named
`SignatureOptions`, `KindThreshold`, `RowAnomaly`

truecopy/columns


`profileColumns(rows, options)`	what each column holds, counted once
`dominantKind(profile, thresholds)`	the kind a column is made of
`columnCount(rows)`, `cellAt(row, column)`
`ColumnProfile`, `ProfileOptions`

truecopy/roles


`assignRoles(profiles, rules)`	what each column IS
`RoleRule`	`role`, `kind`, `minimum`, `take`, `among`

truecopy/schema


`schemaOf(schema)`	one declaration, two outputs
`RecordOf<typeof schema>`	the record type, read back off it
`validate(schema, records)`	the first rule that broke, or `null`
`validateWith(standardSchema, records, rule)`	bring your own Zod, Valibot or ArkType
`conforms(schema, record)`	one record
`Schema`, `Field`, `FieldFormat`, `SchemaViolation`, `StandardSchema`

truecopy/classify


`classifyDocument(text, kinds)`	the kind, with precedence
`DocumentKind`, `Requirement`, `PatternSet`

truecopy/pattern


`compilePattern(raw, options?)`	domain knowledge as data, behind a guard
`toRawPattern(pattern)`	back to JSON
`countMatches(text, pattern)`
`NEVER_MATCHES`	what a rejected pattern compiles to
`RawPattern`, `CompileOptions`

truecopy/explain


`explainDocument(document, options?)`	what the reading decided, as text
`explainRows(rows, options?)`	the same, for rows that never came from a page
`describeAnomaly(anomaly)`	one anomaly, in words
`ExplainOptions`

truecopy/contract


`readDocument(document, reader)`	drive a reading, and enforce the one rule
`Reader`	`read`, `selfCheck`, `rowsToReview`, and two optional
`ReadResult`, `Reading`, `SelfCheck`, `Discrepancy`, `Refusal`, `ReviewableRow`, `Verdict`

truecopy/kit


`checkContract(reader, corpus, options)`	six rules, in your own suite
`contractReport(results)`	the run as a file you commit
`failures(results)`	what went red, in sentences
`pdfWithText(words)`, `pdfWithPages(pages)`	a real PDF, xref table included
`docxWithText(blocks)`, `docxWithBody(bodyXml)`	a real `.docx`, real CRC, both methods
`documentWithoutSubstance()`	for rule 5
`CorpusCase`, `CheckResult`, `ContractOptions`, `PlacedWord`

truecopy (types)

Document, TextPage, Row, PositionedItem, Place.