@graphty/graph-io 0.2.4 → 0.3.0

package/dist/chunks/{records-CGpxszm1.js.map → records-BKSMowhR.js.map}

@@ -1 +1 @@
-{"version":3,"file":"records-CGpxszm1.js","sources":["../../src/formats/csv/records.ts"],"sourcesContent":["/**\n * The one streaming CSV record reader of the package (design sections 8.2 and 8.4: hand-written\n * tokenisers per format; CSV and Neo4j CSV are line-oriented over a byte stream): RFC 4180 records\n * with a configurable single-character delimiter and quote, a doubled quote inside a quoted field,\n * quoted fields spanning lines, LF / CRLF / lone-CR terminators, read from the common text reader\n * one chunk at a time as a character state machine. Only the open field is ever held, so a field\n * spanning many chunks costs its length once and a multi-gigabyte file is never buffered.\n *\n * `RecordReader` (used by the Neo4j importer) keeps one reusable cell array and one reusable\n * \"was quoted\" array: a quoted empty field (`\"\"`) is an empty string while an unquoted empty field\n * means \"not set\", and only the tokeniser can tell the two apart. `CsvRecordReader` (the CSV\n * importer) wraps it with delimiter sniffing over a bounded preview and yields a fresh cell array\n * per row. A malformed quoted field is fatal for both: an unterminated quote swallows the rest of\n * the file, and text after a closing quote makes every later cell boundary unreliable.\n */\n\nimport { GraphFormatError } from \"@graphty/graph-format\";\n\nimport { type ReadOptions, textChunks } from \"../../common/input.js\";\nimport { type ImportReportBuilder } from \"../../common/report.js\";\nimport { type ImportInput } from \"../../types.js\";\n\n/** Issue code: a quoted field is never closed; the import aborts (everything after it would be one cell). */\nexport const UNCLOSED_QUOTE_CODE = \"E_CSV_UNCLOSED_QUOTE\";\n\n/** Issue code: a closing quote is followed by text other than a delimiter or a line break; the import aborts. */\nexport const BAD_QUOTE_CODE = \"E_CSV_QUOTE\";\n\n/** The delimiters tried, in priority order, when none is given. */\nconst DELIMITER_CANDIDATES: readonly string[] = Object.freeze([\",\", \"\\t\", \";\", \"|\", \" \"]);\n\n/** Rows the delimiter sniff looks at. */\nconst PREVIEW_ROWS = 10;\n\n/** Characters buffered for the sniff when the input has fewer than PREVIEW_ROWS line breaks. */\nconst PREVIEW_CHARS = 64 * 1024;\n\nconst LF = 10;\nconst CR = 13;\n\n/** Tokeniser state: at the start of a field. */\nconst START = 0;\n/** Tokeniser state: inside an unquoted field. */\nconst UNQUOTED = 1;\n/** Tokeniser state: inside a quoted field. */\nconst QUOTED = 2;\n/** Tokeniser state: just after a quote inside a quoted field (a second quote is a literal, anything else ends the field). */\nconst CLOSING = 3;\n/** Tokeniser state: after a closed quoted field, before the delimiter (text here is malformed). */\nconst AFTER_QUOTED = 4;\n/** Inside a leading comment line (skipped up to its line break). */\nconst COMMENT = 5;\n\ntype State = typeof START | typeof UNQUOTED | typeof QUOTED | typeof CLOSING | typeof AFTER_QUOTED | typeof COMMENT;\n\n/** The delimiter and quote of a reader. */\nexport interface RecordSyntax {\n    /** The field delimiter, one character; null sniffs it from the first rows. */\n    readonly delimiter: string | null;\n    /** The quote character, one character. */\n    readonly quote: string;\n    /** The delimiters tried when `delimiter` is null; DELIMITER_CANDIDATES by default. */\n    readonly candidates?: readonly string[] | undefined;\n    /**\n     * Characters that open a comment line while no record has been read yet (SNAP `#`, KONECT\n     * `%`): such leading lines are skipped, kept in `leadingComments` and left out of the delimiter\n     * sniff. A later line starting with one of them is an ordinary record. Default: none.\n     */\n    readonly comments?: readonly string[] | undefined;\n}\n\n/**\n * Check a delimiter or quote option: exactly one character that is not a line break, and the two\n * must differ.\n * @param syntax - the delimiter (or null for sniffing) and quote\n * @returns the syntax unchanged; E_UNSUPPORTED when invalid\n */\nexport function checkRecordSyntax(syntax: RecordSyntax): RecordSyntax {\n    for (const [option, value] of [\n        [\"delimiter\", syntax.delimiter],\n        [\"quote\", syntax.quote],\n    ] as const) {\n        if (value === null) {\n            continue;\n        }\n        if (value.length !== 1 || value === \"\\n\" || value === \"\\r\") {\n            throw new GraphFormatError(\n                \"E_UNSUPPORTED\",\n                `option ${option}: ${JSON.stringify(value)} is not a single non-line-break character`,\n                { option, found: value },\n            );\n        }\n    }\n    if (syntax.delimiter !== null && syntax.delimiter === syntax.quote) {\n        throw new GraphFormatError(\"E_UNSUPPORTED\", \"options delimiter and quote must differ\", {\n            option: \"delimiter\",\n            found: syntax.delimiter,\n        });\n    }\n    return syntax;\n}\n\n/**\n * Sniff the line terminator of a text: a lone `\\r` only when the text has `\\r` and no `\\n` at all\n * (classic Mac files); `\\n` otherwise.\n * @param text - the preview text\n * @returns \"\\n\" or \"\\r\"\n */\nexport function sniffNewline(text: string): \"\\n\" | \"\\r\" {\n    return !text.includes(\"\\n\") && text.includes(\"\\r\") ? \"\\r\" : \"\\n\";\n}\n\n/**\n * Split a text into records synchronously (the first `maxRows` of them), honouring quotes; for\n * the delimiter sniff and the registry's head sniff, where the input is a bounded preview.\n * @param text - the text\n * @param delimiter - the delimiter\n * @param quote - the quote character\n * @param maxRows - the most rows to return\n * @returns the rows as cell arrays (blank lines skipped)\n */\nfunction splitRecords(text: string, delimiter: string, quote: string, maxRows: number): string[][] {\n    const rows: string[][] = [];\n    const delimiterCode = delimiter.charCodeAt(0);\n    const quoteCode = quote.charCodeAt(0);\n    let cells: string[] = [];\n    let state: State = START;\n    let segment = 0;\n    let field = \"\";\n    const n = text.length;\n    for (let i = 0; i < n && rows.length < maxRows; i++) {\n        const c = text.charCodeAt(i);\n        if (state === QUOTED) {\n            if (c === quoteCode) {\n                field += text.slice(segment, i);\n                state = CLOSING;\n            }\n            continue;\n        }\n        if (state === CLOSING) {\n            if (c === quoteCode) {\n                field += quote;\n                segment = i + 1;\n                state = QUOTED;\n                continue;\n            }\n            state = AFTER_QUOTED;\n            segment = i;\n        }\n        if (c === delimiterCode || c === LF || c === CR) {\n            if (state === UNQUOTED || state === AFTER_QUOTED) {\n                field += text.slice(segment, i);\n            }\n            if (c === delimiterCode) {\n                cells.push(field);\n                field = \"\";\n                state = START;\n            } else {\n                if (state !== START || cells.length > 0) {\n                    cells.push(field);\n                    rows.push(cells);\n                    cells = [];\n                    field = \"\";\n                }\n                state = START;\n                if (c === CR && text.charCodeAt(i + 1) === LF) {\n                    i++;\n                }\n            }\n            segment = i + 1;\n            continue;\n        }\n        if (state === START) {\n            if (c === quoteCode) {\n                state = QUOTED;\n                segment = i + 1;\n            } else {\n                state = UNQUOTED;\n                segment = i;\n            }\n        }\n    }\n    if (rows.length < maxRows && (state !== START || cells.length > 0)) {\n        if (state === UNQUOTED || state === AFTER_QUOTED || state === QUOTED) {\n            field += text.slice(segment, n);\n        }\n        cells.push(field);\n        rows.push(cells);\n    }\n    return rows;\n}\n\n/**\n * The preview text without its leading comment lines (those starting with one of the comment\n * characters), so the sniff sees the records only.\n * @param text - the preview text\n * @param comments - the comment characters\n * @returns the text from the first non-comment line on\n */\nfunction stripLeadingComments(text: string, comments: readonly string[]): string {\n    if (comments.length === 0) {\n        return text;\n    }\n    let at = 0;\n    while (at < text.length && comments.includes(text[at])) {\n        const lf = text.indexOf(\"\\n\", at);\n        const cr = text.indexOf(\"\\r\", at);\n        const end = Math.min(lf < 0 ? Infinity : lf, cr < 0 ? Infinity : cr);\n        if (end === Infinity) {\n            return \"\";\n        }\n        at = end + 1;\n        if (text[at - 1] === \"\\r\" && text[at] === \"\\n\") {\n            at++;\n        }\n    }\n    return at === 0 ? text : text.slice(at);\n}\n\n/**\n * Sniff the delimiter of a text: every candidate is tried over the first rows and the one whose\n * field count is most consistent across rows wins (ties broken by the higher field count); a\n * candidate that yields fewer than two fields per row on average is never chosen. `null` when no\n * candidate qualifies (a single-column file).\n * @param text - the preview text\n * @param newline - the sniffed line terminator (unused by the splitter, which reads every kind;\n * kept for callers that sniffed it)\n * @param candidates - the delimiters to try, in priority order\n * @param quote - the quote character\n * @returns the delimiter, or null\n */\nexport function sniffDelimiter(\n    text: string,\n    newline: \"\\n\" | \"\\r\" = \"\\n\",\n    candidates: readonly string[] = DELIMITER_CANDIDATES,\n    quote = '\"',\n): string | null {\n    let best: string | null = null;\n    let bestDelta = Infinity;\n    let bestAverage = 0;\n    const terminated = text.endsWith(\"\\n\") || text.endsWith(\"\\r\");\n    for (const delimiter of candidates) {\n        if (delimiter === quote || delimiter === newline) {\n            continue;\n        }\n        let rows = splitRecords(text, delimiter, quote, PREVIEW_ROWS).filter((row) => !isBlankRow(row));\n        if (rows.length > 1 && !terminated) {\n            // the preview is a prefix of the file: its last row may be cut short\n            rows = rows.slice(0, -1);\n        }\n        if (rows.length === 0) {\n            continue;\n        }\n        let total = 0;\n        let delta = 0;\n        for (let i = 0; i < rows.length; i++) {\n            const count = rows[i].length;\n            total += count;\n            if (i > 0) {\n                delta += Math.abs(count - rows[i - 1].length);\n            }\n        }\n        const average = total / rows.length;\n        if (average > 1.99 && (delta < bestDelta || (delta === bestDelta && average > bestAverage))) {\n            best = delimiter;\n            bestDelta = delta;\n            bestAverage = average;\n        }\n    }\n    return best;\n}\n\n/**\n * Whether a parsed row is a blank line: one cell holding only whitespace.\n * @param row - the row\n * @returns true for a blank line\n */\nfunction isBlankRow(row: readonly string[]): boolean {\n    return row.length === 1 && row[0].trim().length === 0;\n}\n\n/**\n * Reads the records of one input. Iterate with `for await (const count of reader)`: each\n * iteration fills `reader.cells[0..count)` and `reader.quoted[0..count)` and sets `reader.line`\n * to the 1-based line the record started on. Blank lines (an empty unquoted record of one field)\n * are skipped. The arrays are reused between records. When the syntax gives no delimiter, the\n * first rows (PREVIEW_ROWS, or PREVIEW_CHARS characters) are buffered and the delimiter sniffed\n * from them before the first record is yielded.\n */\nexport class RecordReader implements AsyncIterable<number> {\n    /** The cells of the record most recently yielded; only the first `count` entries are valid. */\n    readonly cells: string[] = [];\n\n    /** Whether each cell of the record most recently yielded was quoted. */\n    readonly quoted: boolean[] = [];\n\n    /** The leading comment lines (without their line breaks), in order; empty without `syntax.comments`. */\n    readonly leadingComments: string[] = [];\n\n    private readonly input: ImportInput;\n\n    private readonly report: ImportReportBuilder;\n\n    private readonly readOptions: ReadOptions;\n\n    /** The delimiter (null while unsniffed), quote and comment characters. */\n    readonly syntax: RecordSyntax;\n\n    private delimiterText: string | null;\n\n    private recordLine = 0;\n\n    /**\n     * Create a reader; nothing is read until iteration starts.\n     * @param input - the input (or an async iterable of already decoded text chunks)\n     * @param report - the report decode and quoting errors are recorded in\n     * @param syntax - the delimiter (null to sniff) and quote (already checked)\n     * @param readOptions - cancellation and progress\n     */\n    constructor(input: ImportInput, report: ImportReportBuilder, syntax: RecordSyntax, readOptions: ReadOptions) {\n        this.input = input;\n        this.report = report;\n        this.readOptions = readOptions;\n        this.syntax = syntax;\n        this.delimiterText = syntax.delimiter;\n    }\n\n    /**\n     * The 1-based line the most recently yielded record started on.\n     * @returns the line number; 0 before the first record\n     */\n    get line(): number {\n        return this.recordLine;\n    }\n\n    /**\n     * The delimiter in use.\n     * @returns the given or sniffed delimiter; null before the sniff ran\n     */\n    get delimiter(): string | null {\n        return this.delimiterText;\n    }\n\n    /**\n     * Iterate the records.\n     * @yields the number of cells of each record\n     * @returns nothing\n     */\n    async *[Symbol.asyncIterator](): AsyncGenerator<number, void, undefined> {\n        const scanner = new RecordScanner(this);\n        for await (const chunk of this.chunks()) {\n            // the first chunk arrives after the sniff (or with the given delimiter)\n            scanner.setDelimiter((this.delimiterText ?? \",\").charCodeAt(0));\n            let i = 0;\n            while (i < chunk.length) {\n                i = scanner.scan(chunk, i);\n                if (scanner.ready) {\n                    scanner.ready = false;\n                    this.recordLine = scanner.recordLine;\n                    yield scanner.count;\n                    scanner.count = 0;\n                }\n            }\n            scanner.endChunk(chunk);\n        }\n        if (scanner.finish()) {\n            this.recordLine = scanner.recordLine;\n            yield scanner.count;\n        }\n    }\n\n    /**\n     * Record a text-after-quote error (fatal).\n     * @param line - the line the field started on\n     * @returns never\n     */\n    badQuote(line: number): never {\n        return this.report.fail(BAD_QUOTE_CODE, `line ${line}: text after the closing quote of a quoted field`, {\n            line,\n        });\n    }\n\n    /**\n     * Record an unclosed-quote error (fatal).\n     * @param line - the line the field started on\n     * @returns never\n     */\n    unclosedQuote(line: number): never {\n        return this.report.fail(UNCLOSED_QUOTE_CODE, `a quoted field starting on line ${line} is never closed`, {\n            line,\n        });\n    }\n\n    /**\n     * The decoded text chunks of the input; when the delimiter is to be sniffed, the first rows\n     * are buffered (bounded by PREVIEW_ROWS line breaks or PREVIEW_CHARS characters), the sniff\n     * runs, and the buffered text is yielded as one chunk before the rest streams through.\n     * @yields the text chunks\n     * @returns nothing\n     */\n    private async *chunks(): AsyncGenerator<string, void, undefined> {\n        const source = textChunks(this.input, this.report, this.readOptions);\n        if (this.delimiterText !== null) {\n            yield* source;\n            return;\n        }\n        const pieces: string[] = [];\n        let length = 0;\n        let breaks = 0;\n        let lastWasCr = false;\n        let sniffed = false;\n        for await (const chunk of source) {\n            if (sniffed) {\n                yield chunk;\n                continue;\n            }\n            pieces.push(chunk);\n            length += chunk.length;\n            for (let i = 0; i < chunk.length && breaks < PREVIEW_ROWS; i++) {\n                const c = chunk.charCodeAt(i);\n                if (c === LF) {\n                    if (!lastWasCr) {\n                        breaks++;\n                    }\n                } else if (c === CR) {\n                    breaks++;\n                }\n                lastWasCr = c === CR;\n            }\n            if (breaks < PREVIEW_ROWS && length < PREVIEW_CHARS) {\n                continue;\n            }\n            const text = pieces.join(\"\");\n            pieces.length = 0;\n            this.sniff(text);\n            sniffed = true;\n            yield text;\n        }\n        if (!sniffed) {\n            const text = pieces.join(\"\");\n            if (text.length > 0) {\n                this.sniff(text);\n                yield text;\n            }\n        }\n    }\n\n    /**\n     * Sniff the delimiter from the preview text and record it.\n     * @param text - the preview\n     */\n    private sniff(text: string): void {\n        const candidates = this.syntax.candidates ?? DELIMITER_CANDIDATES;\n        // the sniff is a heuristic over the first rows: a single chunk holding a huge quoted cell\n        // is capped so the candidate scans stay bounded\n        const body = stripLeadingComments(text.slice(0, PREVIEW_CHARS), this.syntax.comments ?? []);\n        this.delimiterText = sniffDelimiter(body, sniffNewline(body), candidates, this.syntax.quote) ?? candidates[0];\n    }\n}\n\n/**\n * The record state machine of RecordReader, resumable at every record end so the reader can\n * yield a record before the next one overwrites the shared cell arrays. Every state lives on the\n * instance; `scan()` runs the per-character loop over one chunk from a position and returns where\n * it stopped (after the character that completed a record, with `ready` set, or the chunk end).\n */\nclass RecordScanner {\n    /** Whether a record just completed. */\n    ready = false;\n\n    /** The cell count of the record in progress (the completed one while `ready`). */\n    count = 0;\n\n    /** The line the completed record started on (valid while `ready`, and after finish()). */\n    recordLine = 1;\n\n    /** The line the record in progress started on. */\n    private startLine = 1;\n\n    private readonly reader: RecordReader;\n\n    private readonly quoteCode: number;\n\n    private readonly quoteText: string;\n\n    private readonly commentCodes: readonly number[];\n\n    private delimiterCode = -1;\n\n    private state: State = START;\n\n    private field = \"\";\n\n    private fieldQuoted = false;\n\n    private line = 1;\n\n    private lastWasCr = false;\n\n    /** Whether no record has started yet (comment lines are recognised until one does). */\n    private leading: boolean;\n\n    private comment = \"\";\n\n    /** The start of the unconsumed run of the current chunk that belongs to the field or comment. */\n    private segment = 0;\n\n    /**\n     * Create a scanner over a reader's cell arrays.\n     * @param reader - the reader\n     */\n    constructor(reader: RecordReader) {\n        this.reader = reader;\n        this.quoteCode = reader.syntax.quote.charCodeAt(0);\n        this.quoteText = reader.syntax.quote;\n        this.commentCodes = (reader.syntax.comments ?? []).map((c) => c.charCodeAt(0));\n        this.leading = this.commentCodes.length > 0;\n    }\n\n    /**\n     * Set the delimiter once it is known (before the first chunk is scanned).\n     * @param code - the delimiter's character code\n     */\n    setDelimiter(code: number): void {\n        if (this.delimiterCode < 0) {\n            this.delimiterCode = code;\n        }\n    }\n\n    /**\n     * Scan one chunk from a position until a record completes or the chunk ends.\n     * @param chunk - the chunk\n     * @param from - where to start\n     * @returns the position after the last character consumed\n     */\n    scan(chunk: string, from: number): number {\n        const { quoteCode, delimiterCode } = this;\n        const n = chunk.length;\n        let i = from;\n        this.segment = from;\n        while (i < n) {\n            const c = chunk.charCodeAt(i);\n            if (this.state === COMMENT) {\n                if (c === LF || c === CR) {\n                    this.comment += chunk.slice(this.segment, i);\n                    this.reader.leadingComments.push(this.comment);\n                    this.comment = \"\";\n                    this.state = START;\n                    this.endLine(c);\n                    this.segment = i + 1;\n                }\n                i++;\n                continue;\n            }\n            if (this.state === QUOTED) {\n                // jump to the next quote; the run in between is field text whose line breaks are counted\n                const q = chunk.indexOf(this.quoteText, i);\n                const end = q < 0 ? n : q;\n                if (end > i) {\n                    this.countLineBreaks(chunk, i, end);\n                }\n                if (q < 0) {\n                    i = n;\n                    break;\n                }\n                this.field += chunk.slice(this.segment, q);\n                this.state = CLOSING;\n                this.lastWasCr = false;\n                i = q + 1;\n                continue;\n            }\n            if (this.state === CLOSING) {\n                if (c === quoteCode) {\n                    this.field += this.quoteText;\n                    this.segment = i + 1;\n                    this.state = QUOTED;\n                    this.lastWasCr = false;\n                    i++;\n                    continue;\n                }\n                if (c !== delimiterCode && c !== LF && c !== CR) {\n                    this.reader.badQuote(this.startLine);\n                }\n                this.state = AFTER_QUOTED;\n                this.segment = i;\n                // fall through to the delimiter / line-end handling below without consuming c\n            }\n            if (c === delimiterCode || c === LF || c === CR) {\n                if (c === LF && this.lastWasCr) {\n                    // the second half of a CRLF already ended the record\n                    this.lastWasCr = false;\n                    this.segment = i + 1;\n                    i++;\n                    continue;\n                }\n                if (this.state === UNQUOTED) {\n                    this.field += chunk.slice(this.segment, i);\n                }\n                if (c === delimiterCode) {\n                    this.push();\n                    this.state = START;\n                    this.lastWasCr = false;\n                    this.segment = i + 1;\n                    i++;\n                    continue;\n                }\n                const blank = this.state === START && this.count === 0;\n                if (!blank) {\n                    this.push();\n                    this.ready = true;\n                    this.recordLine = this.startLine;\n                }\n                this.state = START;\n                this.endLine(c);\n                this.segment = i + 1;\n                i++;\n                if (this.ready) {\n                    return i;\n                }\n                continue;\n            }\n            this.lastWasCr = false;\n            if (this.state === START) {\n                if (this.leading && this.count === 0 && this.commentCodes.includes(c)) {\n                    this.state = COMMENT;\n                    this.segment = i;\n                    i++;\n                    continue;\n                }\n                this.leading = false;\n                if (c === quoteCode) {\n                    this.state = QUOTED;\n                    this.fieldQuoted = true;\n                    this.segment = i + 1;\n                } else {\n                    this.state = UNQUOTED;\n                    this.segment = i;\n                }\n            }\n            i++;\n        }\n        return i;\n    }\n\n    /**\n     * Keep the tail of a chunk that belongs to an open field or comment.\n     * @param chunk - the chunk just scanned to its end\n     */\n    endChunk(chunk: string): void {\n        const n = chunk.length;\n        if (this.segment >= n) {\n            return;\n        }\n        if (this.state === COMMENT) {\n            this.comment += chunk.slice(this.segment, n);\n        } else if (this.state === QUOTED || this.state === UNQUOTED) {\n            this.field += chunk.slice(this.segment, n);\n        }\n        this.segment = n;\n    }\n\n    /**\n     * The end of the input: an open comment is kept, an open quote is fatal, an open record\n     * (no trailing line break) is completed.\n     * @returns true when a last record is ready\n     */\n    finish(): boolean {\n        const { state } = this;\n        if (state === COMMENT) {\n            this.reader.leadingComments.push(this.comment);\n            return false;\n        }\n        if (state === QUOTED) {\n            this.reader.unclosedQuote(this.startLine);\n        }\n        if (state !== START || this.count > 0) {\n            this.push();\n            this.recordLine = this.startLine;\n            return true;\n        }\n        return false;\n    }\n\n    /**\n     * Count the line breaks of a run of quoted text (a CRLF counts once, like the record loop).\n     * @param chunk - the chunk\n     * @param from - the first index of the run\n     * @param to - the index after the run\n     */\n    private countLineBreaks(chunk: string, from: number, to: number): void {\n        let { line, lastWasCr } = this;\n        for (let i = from; i < to; i++) {\n            const c = chunk.charCodeAt(i);\n            if (c === LF) {\n                if (!lastWasCr) {\n                    line++;\n                }\n                lastWasCr = false;\n            } else if (c === CR) {\n                line++;\n                lastWasCr = true;\n            } else {\n                lastWasCr = false;\n            }\n        }\n        this.line = line;\n        this.lastWasCr = lastWasCr;\n    }\n\n    /** Store the field in progress as the next cell. */\n    private push(): void {\n        const { reader, count } = this;\n        reader.cells[count] = this.field;\n        reader.quoted[count] = this.fieldQuoted;\n        this.count = count + 1;\n        this.field = \"\";\n        this.fieldQuoted = false;\n    }\n\n    /**\n     * Count a line break that ends a record or a comment line.\n     * @param c - the LF or CR\n     */\n    private endLine(c: number): void {\n        if (!(c === LF && this.lastWasCr)) {\n            this.line++;\n        }\n        this.startLine = this.line;\n        this.lastWasCr = c === CR;\n    }\n}\n\n/** What the CSV importer's reader needs besides the input. */\nexport interface CsvReaderOptions extends ReadOptions {\n    /** The delimiter; null or undefined sniffs it from the preview. */\n    readonly delimiter?: string | null | undefined;\n    /** The characters opening a leading comment line (RecordSyntax.comments); none by default. */\n    readonly comments?: readonly string[] | undefined;\n}\n\n/**\n * Records of a CSV input, one string array per row, streamed: iterate with\n * `for await (const row of reader)` and read `reader.line` for the 1-based line the row starts\n * on. The delimiter is known after the first row (`reader.delimiter`). Rows come out as fresh\n * arrays of cell texts exactly as written (no trimming, no typing); `reader.quoted` tells, for\n * the row just yielded, which cells were quoted (a quoted empty cell is the empty string, an\n * unquoted one is \"not set\"). Blank lines and whitespace-only single-cell lines are skipped.\n */\nexport class CsvRecordReader implements AsyncIterable<string[]> {\n    private readonly inner: RecordReader;\n\n    /**\n     * Create a reader; nothing is read until iteration starts.\n     * @param input - the input\n     * @param report - the report parse errors are recorded in\n     * @param options - the delimiter, cancellation and progress\n     */\n    constructor(input: ImportInput, report: ImportReportBuilder, options: CsvReaderOptions = {}) {\n        this.inner = new RecordReader(\n            input,\n            report,\n            { delimiter: options.delimiter ?? null, quote: '\"', comments: options.comments },\n            { signal: options.signal, onProgress: options.onProgress },\n        );\n    }\n\n    /**\n     * The leading comment lines the reader skipped (SNAP `#` lines, the KONECT `%` header), in order.\n     * @returns the lines without their line breaks\n     */\n    get leadingComments(): readonly string[] {\n        return this.inner.leadingComments;\n    }\n\n    /**\n     * The 1-based line the most recently yielded row starts on.\n     * @returns the line; 0 before the first row\n     */\n    get line(): number {\n        return this.inner.line;\n    }\n\n    /**\n     * The delimiter in use.\n     * @returns the given or sniffed delimiter; null before the sniff ran\n     */\n    get delimiter(): string | null {\n        return this.inner.delimiter;\n    }\n\n    /**\n     * Whether each cell of the row most recently yielded was quoted.\n     * @returns the flags (valid for the first `row.length` entries)\n     */\n    get quoted(): readonly boolean[] {\n        return this.inner.quoted;\n    }\n\n    /**\n     * Iterate the rows.\n     * @yields one row at a time as its cell texts\n     * @returns nothing\n     */\n    async *[Symbol.asyncIterator](): AsyncGenerator<string[], void, undefined> {\n        const { inner } = this;\n        for await (const count of inner) {\n            if (count === 1 && !inner.quoted[0] && inner.cells[0].trim().length === 0) {\n                continue;\n            }\n            yield inner.cells.slice(0, count);\n        }\n    }\n}\n"],"names":[],"mappings":";;AAuBO,MAAM,sBAAsB;AAG5B,MAAM,iBAAiB;AAG9B,MAAM,uBAA0C,OAAO,OAAO,CAAC,KAAK,KAAM,KAAK,KAAK,GAAG,CAAC;AAGxF,MAAM,eAAe;AAGrB,MAAM,gBAAgB,KAAK;AAE3B,MAAM,KAAK;AACX,MAAM,KAAK;AAGX,MAAM,QAAQ;AAEd,MAAM,WAAW;AAEjB,MAAM,SAAS;AAEf,MAAM,UAAU;AAEhB,MAAM,eAAe;AAErB,MAAM,UAAU;AA0BT,SAAS,kBAAkB,QAAoC;AAClE,aAAW,CAAC,QAAQ,KAAK,KAAK;AAAA,IAC1B,CAAC,aAAa,OAAO,SAAS;AAAA,IAC9B,CAAC,SAAS,OAAO,KAAK;AAAA,EAAA,GACd;AACR,QAAI,UAAU,MAAM;AAChB;AAAA,IACJ;AACA,QAAI,MAAM,WAAW,KAAK,UAAU,QAAQ,UAAU,MAAM;AACxD,YAAM,IAAI;AAAA,QACN;AAAA,QACA,UAAU,MAAM,KAAK,KAAK,UAAU,KAAK,CAAC;AAAA,QAC1C,EAAE,QAAQ,OAAO,MAAA;AAAA,MAAM;AAAA,IAE/B;AAAA,EACJ;AACA,MAAI,OAAO,cAAc,QAAQ,OAAO,cAAc,OAAO,OAAO;AAChE,UAAM,IAAI,iBAAiB,iBAAiB,2CAA2C;AAAA,MACnF,QAAQ;AAAA,MACR,OAAO,OAAO;AAAA,IAAA,CACjB;AAAA,EACL;AACA,SAAO;AACX;AAQO,SAAS,aAAa,MAA2B;AACpD,SAAO,CAAC,KAAK,SAAS,IAAI,KAAK,KAAK,SAAS,IAAI,IAAI,OAAO;AAChE;AAWA,SAAS,aAAa,MAAc,WAAmB,OAAe,SAA6B;AAC/F,QAAM,OAAmB,CAAA;AACzB,QAAM,gBAAgB,UAAU,WAAW,CAAC;AAC5C,QAAM,YAAY,MAAM,WAAW,CAAC;AACpC,MAAI,QAAkB,CAAA;AACtB,MAAI,QAAe;AACnB,MAAI,UAAU;AACd,MAAI,QAAQ;AACZ,QAAM,IAAI,KAAK;AACf,WAAS,IAAI,GAAG,IAAI,KAAK,KAAK,SAAS,SAAS,KAAK;AACjD,UAAM,IAAI,KAAK,WAAW,CAAC;AAC3B,QAAI,UAAU,QAAQ;AAClB,UAAI,MAAM,WAAW;AACjB,iBAAS,KAAK,MAAM,SAAS,CAAC;AAC9B,gBAAQ;AAAA,MACZ;AACA;AAAA,IACJ;AACA,QAAI,UAAU,SAAS;AACnB,UAAI,MAAM,WAAW;AACjB,iBAAS;AACT,kBAAU,IAAI;AACd,gBAAQ;AACR;AAAA,MACJ;AACA,cAAQ;AACR,gBAAU;AAAA,IACd;AACA,QAAI,MAAM,iBAAiB,MAAM,MAAM,MAAM,IAAI;AAC7C,UAAI,UAAU,YAAY,UAAU,cAAc;AAC9C,iBAAS,KAAK,MAAM,SAAS,CAAC;AAAA,MAClC;AACA,UAAI,MAAM,eAAe;AACrB,cAAM,KAAK,KAAK;AAChB,gBAAQ;AACR,gBAAQ;AAAA,MACZ,OAAO;AACH,YAAI,UAAU,SAAS,MAAM,SAAS,GAAG;AACrC,gBAAM,KAAK,KAAK;AAChB,eAAK,KAAK,KAAK;AACf,kBAAQ,CAAA;AACR,kBAAQ;AAAA,QACZ;AACA,gBAAQ;AACR,YAAI,MAAM,MAAM,KAAK,WAAW,IAAI,CAAC,MAAM,IAAI;AAC3C;AAAA,QACJ;AAAA,MACJ;AACA,gBAAU,IAAI;AACd;AAAA,IACJ;AACA,QAAI,UAAU,OAAO;AACjB,UAAI,MAAM,WAAW;AACjB,gBAAQ;AACR,kBAAU,IAAI;AAAA,MAClB,OAAO;AACH,gBAAQ;AACR,kBAAU;AAAA,MACd;AAAA,IACJ;AAAA,EACJ;AACA,MAAI,KAAK,SAAS,YAAY,UAAU,SAAS,MAAM,SAAS,IAAI;AAChE,QAAI,UAAU,YAAY,UAAU,gBAAgB,UAAU,QAAQ;AAClE,eAAS,KAAK,MAAM,SAAS,CAAC;AAAA,IAClC;AACA,UAAM,KAAK,KAAK;AAChB,SAAK,KAAK,KAAK;AAAA,EACnB;AACA,SAAO;AACX;AASA,SAAS,qBAAqB,MAAc,UAAqC;AAC7E,MAAI,SAAS,WAAW,GAAG;AACvB,WAAO;AAAA,EACX;AACA,MAAI,KAAK;AACT,SAAO,KAAK,KAAK,UAAU,SAAS,SAAS,KAAK,EAAE,CAAC,GAAG;AACpD,UAAM,KAAK,KAAK,QAAQ,MAAM,EAAE;AAChC,UAAM,KAAK,KAAK,QAAQ,MAAM,EAAE;AAChC,UAAM,MAAM,KAAK,IAAI,KAAK,IAAI,WAAW,IAAI,KAAK,IAAI,WAAW,EAAE;AACnE,QAAI,QAAQ,UAAU;AAClB,aAAO;AAAA,IACX;AACA,SAAK,MAAM;AACX,QAAI,KAAK,KAAK,CAAC,MAAM,QAAQ,KAAK,EAAE,MAAM,MAAM;AAC5C;AAAA,IACJ;AAAA,EACJ;AACA,SAAO,OAAO,IAAI,OAAO,KAAK,MAAM,EAAE;AAC1C;AAcO,SAAS,eACZ,MACA,UAAuB,MACvB,aAAgC,sBAChC,QAAQ,KACK;AACb,MAAI,OAAsB;AAC1B,MAAI,YAAY;AAChB,MAAI,cAAc;AAClB,QAAM,aAAa,KAAK,SAAS,IAAI,KAAK,KAAK,SAAS,IAAI;AAC5D,aAAW,aAAa,YAAY;AAChC,QAAI,cAAc,SAAS,cAAc,SAAS;AAC9C;AAAA,IACJ;AACA,QAAI,OAAO,aAAa,MAAM,WAAW,OAAO,YAAY,EAAE,OAAO,CAAC,QAAQ,CAAC,WAAW,GAAG,CAAC;AAC9F,QAAI,KAAK,SAAS,KAAK,CAAC,YAAY;AAEhC,aAAO,KAAK,MAAM,GAAG,EAAE;AAAA,IAC3B;AACA,QAAI,KAAK,WAAW,GAAG;AACnB;AAAA,IACJ;AACA,QAAI,QAAQ;AACZ,QAAI,QAAQ;AACZ,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AAClC,YAAM,QAAQ,KAAK,CAAC,EAAE;AACtB,eAAS;AACT,UAAI,IAAI,GAAG;AACP,iBAAS,KAAK,IAAI,QAAQ,KAAK,IAAI,CAAC,EAAE,MAAM;AAAA,MAChD;AAAA,IACJ;AACA,UAAM,UAAU,QAAQ,KAAK;AAC7B,QAAI,UAAU,SAAS,QAAQ,aAAc,UAAU,aAAa,UAAU,cAAe;AACzF,aAAO;AACP,kBAAY;AACZ,oBAAc;AAAA,IAClB;AAAA,EACJ;AACA,SAAO;AACX;AAOA,SAAS,WAAW,KAAiC;AACjD,SAAO,IAAI,WAAW,KAAK,IAAI,CAAC,EAAE,OAAO,WAAW;AACxD;AAUO,MAAM,aAA8C;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA8BvD,YAAY,OAAoB,QAA6B,QAAsB,aAA0B;AA5B7G,SAAS,QAAkB,CAAA;AAG3B,SAAS,SAAoB,CAAA;AAG7B,SAAS,kBAA4B,CAAA;AAarC,SAAQ,aAAa;AAUjB,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,cAAc;AACnB,SAAK,SAAS;AACd,SAAK,gBAAgB,OAAO;AAAA,EAChC;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,OAAe;AACf,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC3B,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,QAAQ,OAAO,aAAa,IAA6C;AACrE,UAAM,UAAU,IAAI,cAAc,IAAI;AACtC,qBAAiB,SAAS,KAAK,UAAU;AAErC,cAAQ,cAAc,KAAK,iBAAiB,KAAK,WAAW,CAAC,CAAC;AAC9D,UAAI,IAAI;AACR,aAAO,IAAI,MAAM,QAAQ;AACrB,YAAI,QAAQ,KAAK,OAAO,CAAC;AACzB,YAAI,QAAQ,OAAO;AACf,kBAAQ,QAAQ;AAChB,eAAK,aAAa,QAAQ;AAC1B,gBAAM,QAAQ;AACd,kBAAQ,QAAQ;AAAA,QACpB;AAAA,MACJ;AACA,cAAQ,SAAS,KAAK;AAAA,IAC1B;AACA,QAAI,QAAQ,UAAU;AAClB,WAAK,aAAa,QAAQ;AAC1B,YAAM,QAAQ;AAAA,IAClB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAS,MAAqB;AAC1B,WAAO,KAAK,OAAO,KAAK,gBAAgB,QAAQ,IAAI,oDAAoD;AAAA,MACpG;AAAA,IAAA,CACH;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAAqB;AAC/B,WAAO,KAAK,OAAO,KAAK,qBAAqB,mCAAmC,IAAI,oBAAoB;AAAA,MACpG;AAAA,IAAA,CACH;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAe,SAAkD;AAC7D,UAAM,SAAS,WAAW,KAAK,OAAO,KAAK,QAAQ,KAAK,WAAW;AACnE,QAAI,KAAK,kBAAkB,MAAM;AAC7B,aAAO;AACP;AAAA,IACJ;AACA,UAAM,SAAmB,CAAA;AACzB,QAAI,SAAS;AACb,QAAI,SAAS;AACb,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,qBAAiB,SAAS,QAAQ;AAC9B,UAAI,SAAS;AACT,cAAM;AACN;AAAA,MACJ;AACA,aAAO,KAAK,KAAK;AACjB,gBAAU,MAAM;AAChB,eAAS,IAAI,GAAG,IAAI,MAAM,UAAU,SAAS,cAAc,KAAK;AAC5D,cAAM,IAAI,MAAM,WAAW,CAAC;AAC5B,YAAI,MAAM,IAAI;AACV,cAAI,CAAC,WAAW;AACZ;AAAA,UACJ;AAAA,QACJ,WAAW,MAAM,IAAI;AACjB;AAAA,QACJ;AACA,oBAAY,MAAM;AAAA,MACtB;AACA,UAAI,SAAS,gBAAgB,SAAS,eAAe;AACjD;AAAA,MACJ;AACA,YAAM,OAAO,OAAO,KAAK,EAAE;AAC3B,aAAO,SAAS;AAChB,WAAK,MAAM,IAAI;AACf,gBAAU;AACV,YAAM;AAAA,IACV;AACA,QAAI,CAAC,SAAS;AACV,YAAM,OAAO,OAAO,KAAK,EAAE;AAC3B,UAAI,KAAK,SAAS,GAAG;AACjB,aAAK,MAAM,IAAI;AACf,cAAM;AAAA,MACV;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,MAAM,MAAoB;AAC9B,UAAM,aAAa,KAAK,OAAO,cAAc;AAG7C,UAAM,OAAO,qBAAqB,KAAK,MAAM,GAAG,aAAa,GAAG,KAAK,OAAO,YAAY,CAAA,CAAE;AAC1F,SAAK,gBAAgB,eAAe,MAAM,aAAa,IAAI,GAAG,YAAY,KAAK,OAAO,KAAK,KAAK,WAAW,CAAC;AAAA,EAChH;AACJ;AAQA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA;AAAA,EA6ChB,YAAY,QAAsB;AA3ClC,SAAA,QAAQ;AAGR,SAAA,QAAQ;AAGR,SAAA,aAAa;AAGb,SAAQ,YAAY;AAUpB,SAAQ,gBAAgB;AAExB,SAAQ,QAAe;AAEvB,SAAQ,QAAQ;AAEhB,SAAQ,cAAc;AAEtB,SAAQ,OAAO;AAEf,SAAQ,YAAY;AAKpB,SAAQ,UAAU;AAGlB,SAAQ,UAAU;AAOd,SAAK,SAAS;AACd,SAAK,YAAY,OAAO,OAAO,MAAM,WAAW,CAAC;AACjD,SAAK,YAAY,OAAO,OAAO;AAC/B,SAAK,gBAAgB,OAAO,OAAO,YAAY,CAAA,GAAI,IAAI,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC;AAC7E,SAAK,UAAU,KAAK,aAAa,SAAS;AAAA,EAC9C;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,aAAa,MAAoB;AAC7B,QAAI,KAAK,gBAAgB,GAAG;AACxB,WAAK,gBAAgB;AAAA,IACzB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,KAAK,OAAe,MAAsB;AACtC,UAAM,EAAE,WAAW,cAAA,IAAkB;AACrC,UAAM,IAAI,MAAM;AAChB,QAAI,IAAI;AACR,SAAK,UAAU;AACf,WAAO,IAAI,GAAG;AACV,YAAM,IAAI,MAAM,WAAW,CAAC;AAC5B,UAAI,KAAK,UAAU,SAAS;AACxB,YAAI,MAAM,MAAM,MAAM,IAAI;AACtB,eAAK,WAAW,MAAM,MAAM,KAAK,SAAS,CAAC;AAC3C,eAAK,OAAO,gBAAgB,KAAK,KAAK,OAAO;AAC7C,eAAK,UAAU;AACf,eAAK,QAAQ;AACb,eAAK,QAAQ,CAAC;AACd,eAAK,UAAU,IAAI;AAAA,QACvB;AACA;AACA;AAAA,MACJ;AACA,UAAI,KAAK,UAAU,QAAQ;AAEvB,cAAM,IAAI,MAAM,QAAQ,KAAK,WAAW,CAAC;AACzC,cAAM,MAAM,IAAI,IAAI,IAAI;AACxB,YAAI,MAAM,GAAG;AACT,eAAK,gBAAgB,OAAO,GAAG,GAAG;AAAA,QACtC;AACA,YAAI,IAAI,GAAG;AACP,cAAI;AACJ;AAAA,QACJ;AACA,aAAK,SAAS,MAAM,MAAM,KAAK,SAAS,CAAC;AACzC,aAAK,QAAQ;AACb,aAAK,YAAY;AACjB,YAAI,IAAI;AACR;AAAA,MACJ;AACA,UAAI,KAAK,UAAU,SAAS;AACxB,YAAI,MAAM,WAAW;AACjB,eAAK,SAAS,KAAK;AACnB,eAAK,UAAU,IAAI;AACnB,eAAK,QAAQ;AACb,eAAK,YAAY;AACjB;AACA;AAAA,QACJ;AACA,YAAI,MAAM,iBAAiB,MAAM,MAAM,MAAM,IAAI;AAC7C,eAAK,OAAO,SAAS,KAAK,SAAS;AAAA,QACvC;AACA,aAAK,QAAQ;AACb,aAAK,UAAU;AAAA,MAEnB;AACA,UAAI,MAAM,iBAAiB,MAAM,MAAM,MAAM,IAAI;AAC7C,YAAI,MAAM,MAAM,KAAK,WAAW;AAE5B,eAAK,YAAY;AACjB,eAAK,UAAU,IAAI;AACnB;AACA;AAAA,QACJ;AACA,YAAI,KAAK,UAAU,UAAU;AACzB,eAAK,SAAS,MAAM,MAAM,KAAK,SAAS,CAAC;AAAA,QAC7C;AACA,YAAI,MAAM,eAAe;AACrB,eAAK,KAAA;AACL,eAAK,QAAQ;AACb,eAAK,YAAY;AACjB,eAAK,UAAU,IAAI;AACnB;AACA;AAAA,QACJ;AACA,cAAM,QAAQ,KAAK,UAAU,SAAS,KAAK,UAAU;AACrD,YAAI,CAAC,OAAO;AACR,eAAK,KAAA;AACL,eAAK,QAAQ;AACb,eAAK,aAAa,KAAK;AAAA,QAC3B;AACA,aAAK,QAAQ;AACb,aAAK,QAAQ,CAAC;AACd,aAAK,UAAU,IAAI;AACnB;AACA,YAAI,KAAK,OAAO;AACZ,iBAAO;AAAA,QACX;AACA;AAAA,MACJ;AACA,WAAK,YAAY;AACjB,UAAI,KAAK,UAAU,OAAO;AACtB,YAAI,KAAK,WAAW,KAAK,UAAU,KAAK,KAAK,aAAa,SAAS,CAAC,GAAG;AACnE,eAAK,QAAQ;AACb,eAAK,UAAU;AACf;AACA;AAAA,QACJ;AACA,aAAK,UAAU;AACf,YAAI,MAAM,WAAW;AACjB,eAAK,QAAQ;AACb,eAAK,cAAc;AACnB,eAAK,UAAU,IAAI;AAAA,QACvB,OAAO;AACH,eAAK,QAAQ;AACb,eAAK,UAAU;AAAA,QACnB;AAAA,MACJ;AACA;AAAA,IACJ;AACA,WAAO;AAAA,EACX;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,SAAS,OAAqB;AAC1B,UAAM,IAAI,MAAM;AAChB,QAAI,KAAK,WAAW,GAAG;AACnB;AAAA,IACJ;AACA,QAAI,KAAK,UAAU,SAAS;AACxB,WAAK,WAAW,MAAM,MAAM,KAAK,SAAS,CAAC;AAAA,IAC/C,WAAW,KAAK,UAAU,UAAU,KAAK,UAAU,UAAU;AACzD,WAAK,SAAS,MAAM,MAAM,KAAK,SAAS,CAAC;AAAA,IAC7C;AACA,SAAK,UAAU;AAAA,EACnB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAkB;AACd,UAAM,EAAE,UAAU;AAClB,QAAI,UAAU,SAAS;AACnB,WAAK,OAAO,gBAAgB,KAAK,KAAK,OAAO;AAC7C,aAAO;AAAA,IACX;AACA,QAAI,UAAU,QAAQ;AAClB,WAAK,OAAO,cAAc,KAAK,SAAS;AAAA,IAC5C;AACA,QAAI,UAAU,SAAS,KAAK,QAAQ,GAAG;AACnC,WAAK,KAAA;AACL,WAAK,aAAa,KAAK;AACvB,aAAO;AAAA,IACX;AACA,WAAO;AAAA,EACX;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQQ,gBAAgB,OAAe,MAAc,IAAkB;AACnE,QAAI,EAAE,MAAM,UAAA,IAAc;AAC1B,aAAS,IAAI,MAAM,IAAI,IAAI,KAAK;AAC5B,YAAM,IAAI,MAAM,WAAW,CAAC;AAC5B,UAAI,MAAM,IAAI;AACV,YAAI,CAAC,WAAW;AACZ;AAAA,QACJ;AACA,oBAAY;AAAA,MAChB,WAAW,MAAM,IAAI;AACjB;AACA,oBAAY;AAAA,MAChB,OAAO;AACH,oBAAY;AAAA,MAChB;AAAA,IACJ;AACA,SAAK,OAAO;AACZ,SAAK,YAAY;AAAA,EACrB;AAAA;AAAA,EAGQ,OAAa;AACjB,UAAM,EAAE,QAAQ,MAAA,IAAU;AAC1B,WAAO,MAAM,KAAK,IAAI,KAAK;AAC3B,WAAO,OAAO,KAAK,IAAI,KAAK;AAC5B,SAAK,QAAQ,QAAQ;AACrB,SAAK,QAAQ;AACb,SAAK,cAAc;AAAA,EACvB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,QAAQ,GAAiB;AAC7B,QAAI,EAAE,MAAM,MAAM,KAAK,YAAY;AAC/B,WAAK;AAAA,IACT;AACA,SAAK,YAAY,KAAK;AACtB,SAAK,YAAY,MAAM;AAAA,EAC3B;AACJ;AAkBO,MAAM,gBAAmD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS5D,YAAY,OAAoB,QAA6B,UAA4B,CAAA,GAAI;AACzF,SAAK,QAAQ,IAAI;AAAA,MACb;AAAA,MACA;AAAA,MACA,EAAE,WAAW,QAAQ,aAAa,MAAM,OAAO,KAAK,UAAU,QAAQ,SAAA;AAAA,MACtE,EAAE,QAAQ,QAAQ,QAAQ,YAAY,QAAQ,WAAA;AAAA,IAAW;AAAA,EAEjE;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,kBAAqC;AACrC,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,OAAe;AACf,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC3B,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,SAA6B;AAC7B,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,QAAQ,OAAO,aAAa,IAA+C;AACvE,UAAM,EAAE,UAAU;AAClB,qBAAiB,SAAS,OAAO;AAC7B,UAAI,UAAU,KAAK,CAAC,MAAM,OAAO,CAAC,KAAK,MAAM,MAAM,CAAC,EAAE,KAAA,EAAO,WAAW,GAAG;AACvE;AAAA,MACJ;AACA,YAAM,MAAM,MAAM,MAAM,GAAG,KAAK;AAAA,IACpC;AAAA,EACJ;AACJ;"}
+{"version":3,"file":"records-BKSMowhR.js","sources":["../../src/formats/csv/records.ts"],"sourcesContent":["/**\n * The one streaming CSV record reader of the package (design sections 8.2 and 8.4: hand-written\n * tokenisers per format; CSV and Neo4j CSV are line-oriented over a byte stream): RFC 4180 records\n * with a configurable single-character delimiter and quote, a doubled quote inside a quoted field,\n * quoted fields spanning lines, LF / CRLF / lone-CR terminators, read from the common text reader\n * one chunk at a time as a character state machine. Only the open field is ever held, so a field\n * spanning many chunks costs its length once and a multi-gigabyte file is never buffered.\n *\n * `RecordReader` (used by the Neo4j importer) keeps one reusable cell array and one reusable\n * \"was quoted\" array: a quoted empty field (`\"\"`) is an empty string while an unquoted empty field\n * means \"not set\", and only the tokeniser can tell the two apart. `CsvRecordReader` (the CSV\n * importer) wraps it with delimiter sniffing over a bounded preview and yields a fresh cell array\n * per row. A malformed quoted field is fatal for both: an unterminated quote swallows the rest of\n * the file, and text after a closing quote makes every later cell boundary unreliable.\n */\n\nimport { GraphFormatError } from \"@graphty/graph-format\";\n\nimport { type ReadOptions, textChunks } from \"../../common/input.js\";\nimport { type ImportReportBuilder } from \"../../common/report.js\";\nimport { type ImportInput } from \"../../types.js\";\n\n/** Issue code: a quoted field is never closed; the import aborts (everything after it would be one cell). */\nexport const UNCLOSED_QUOTE_CODE = \"E_CSV_UNCLOSED_QUOTE\";\n\n/** Issue code: a closing quote is followed by text other than a delimiter or a line break; the import aborts. */\nexport const BAD_QUOTE_CODE = \"E_CSV_QUOTE\";\n\n/** The delimiters tried, in priority order, when none is given. */\nconst DELIMITER_CANDIDATES: readonly string[] = Object.freeze([\",\", \"\\t\", \";\", \"|\", \" \"]);\n\n/** Rows the delimiter sniff looks at. */\nconst PREVIEW_ROWS = 10;\n\n/** Characters buffered for the sniff when the input has fewer than PREVIEW_ROWS line breaks. */\nconst PREVIEW_CHARS = 64 * 1024;\n\nconst LF = 10;\nconst CR = 13;\n\n/** Tokeniser state: at the start of a field. */\nconst START = 0;\n/** Tokeniser state: inside an unquoted field. */\nconst UNQUOTED = 1;\n/** Tokeniser state: inside a quoted field. */\nconst QUOTED = 2;\n/** Tokeniser state: just after a quote inside a quoted field (a second quote is a literal, anything else ends the field). */\nconst CLOSING = 3;\n/** Tokeniser state: after a closed quoted field, before the delimiter (text here is malformed). */\nconst AFTER_QUOTED = 4;\n/** Inside a leading comment line (skipped up to its line break). */\nconst COMMENT = 5;\n\ntype State = typeof START | typeof UNQUOTED | typeof QUOTED | typeof CLOSING | typeof AFTER_QUOTED | typeof COMMENT;\n\n/** The delimiter and quote of a reader. */\nexport interface RecordSyntax {\n    /** The field delimiter, one character; null sniffs it from the first rows. */\n    readonly delimiter: string | null;\n    /** The quote character, one character. */\n    readonly quote: string;\n    /** The delimiters tried when `delimiter` is null; DELIMITER_CANDIDATES by default. */\n    readonly candidates?: readonly string[] | undefined;\n    /**\n     * Characters that open a comment line while no record has been read yet (SNAP `#`, KONECT\n     * `%`): such leading lines are skipped, kept in `leadingComments` and left out of the delimiter\n     * sniff. A later line starting with one of them is an ordinary record. Default: none.\n     */\n    readonly comments?: readonly string[] | undefined;\n}\n\n/**\n * Check a delimiter or quote option: exactly one character that is not a line break, and the two\n * must differ.\n * @param syntax - the delimiter (or null for sniffing) and quote\n * @returns the syntax unchanged; E_UNSUPPORTED when invalid\n */\nexport function checkRecordSyntax(syntax: RecordSyntax): RecordSyntax {\n    for (const [option, value] of [\n        [\"delimiter\", syntax.delimiter],\n        [\"quote\", syntax.quote],\n    ] as const) {\n        if (value === null) {\n            continue;\n        }\n        if (value.length !== 1 || value === \"\\n\" || value === \"\\r\") {\n            throw new GraphFormatError(\n                \"E_UNSUPPORTED\",\n                `option ${option}: ${JSON.stringify(value)} is not a single non-line-break character`,\n                { option, found: value },\n            );\n        }\n    }\n    if (syntax.delimiter !== null && syntax.delimiter === syntax.quote) {\n        throw new GraphFormatError(\"E_UNSUPPORTED\", \"options delimiter and quote must differ\", {\n            option: \"delimiter\",\n            found: syntax.delimiter,\n        });\n    }\n    return syntax;\n}\n\n/**\n * Sniff the line terminator of a text: a lone `\\r` only when the text has `\\r` and no `\\n` at all\n * (classic Mac files); `\\n` otherwise.\n * @param text - the preview text\n * @returns \"\\n\" or \"\\r\"\n */\nexport function sniffNewline(text: string): \"\\n\" | \"\\r\" {\n    return !text.includes(\"\\n\") && text.includes(\"\\r\") ? \"\\r\" : \"\\n\";\n}\n\n/**\n * Split a text into records synchronously (the first `maxRows` of them), honouring quotes; for\n * the delimiter sniff and the registry's head sniff, where the input is a bounded preview.\n * @param text - the text\n * @param delimiter - the delimiter\n * @param quote - the quote character\n * @param maxRows - the most rows to return\n * @returns the rows as cell arrays (blank lines skipped)\n */\nfunction splitRecords(text: string, delimiter: string, quote: string, maxRows: number): string[][] {\n    const rows: string[][] = [];\n    const delimiterCode = delimiter.charCodeAt(0);\n    const quoteCode = quote.charCodeAt(0);\n    let cells: string[] = [];\n    let state: State = START;\n    let segment = 0;\n    let field = \"\";\n    const n = text.length;\n    for (let i = 0; i < n && rows.length < maxRows; i++) {\n        const c = text.charCodeAt(i);\n        if (state === QUOTED) {\n            if (c === quoteCode) {\n                field += text.slice(segment, i);\n                state = CLOSING;\n            }\n            continue;\n        }\n        if (state === CLOSING) {\n            if (c === quoteCode) {\n                field += quote;\n                segment = i + 1;\n                state = QUOTED;\n                continue;\n            }\n            state = AFTER_QUOTED;\n            segment = i;\n        }\n        if (c === delimiterCode || c === LF || c === CR) {\n            if (state === UNQUOTED || state === AFTER_QUOTED) {\n                field += text.slice(segment, i);\n            }\n            if (c === delimiterCode) {\n                cells.push(field);\n                field = \"\";\n                state = START;\n            } else {\n                if (state !== START || cells.length > 0) {\n                    cells.push(field);\n                    rows.push(cells);\n                    cells = [];\n                    field = \"\";\n                }\n                state = START;\n                if (c === CR && text.charCodeAt(i + 1) === LF) {\n                    i++;\n                }\n            }\n            segment = i + 1;\n            continue;\n        }\n        if (state === START) {\n            if (c === quoteCode) {\n                state = QUOTED;\n                segment = i + 1;\n            } else {\n                state = UNQUOTED;\n                segment = i;\n            }\n        }\n    }\n    if (rows.length < maxRows && (state !== START || cells.length > 0)) {\n        if (state === UNQUOTED || state === AFTER_QUOTED || state === QUOTED) {\n            field += text.slice(segment, n);\n        }\n        cells.push(field);\n        rows.push(cells);\n    }\n    return rows;\n}\n\n/**\n * The preview text without its leading comment lines (those starting with one of the comment\n * characters), so the sniff sees the records only.\n * @param text - the preview text\n * @param comments - the comment characters\n * @returns the text from the first non-comment line on\n */\nfunction stripLeadingComments(text: string, comments: readonly string[]): string {\n    if (comments.length === 0) {\n        return text;\n    }\n    let at = 0;\n    while (at < text.length && comments.includes(text[at])) {\n        const lf = text.indexOf(\"\\n\", at);\n        const cr = text.indexOf(\"\\r\", at);\n        const end = Math.min(lf < 0 ? Infinity : lf, cr < 0 ? Infinity : cr);\n        if (end === Infinity) {\n            return \"\";\n        }\n        at = end + 1;\n        if (text[at - 1] === \"\\r\" && text[at] === \"\\n\") {\n            at++;\n        }\n    }\n    return at === 0 ? text : text.slice(at);\n}\n\n/**\n * Sniff the delimiter of a text: every candidate is tried over the first rows and the one whose\n * field count is most consistent across rows wins (ties broken by the higher field count); a\n * candidate that yields fewer than two fields per row on average is never chosen. `null` when no\n * candidate qualifies (a single-column file).\n * @param text - the preview text\n * @param newline - the sniffed line terminator (unused by the splitter, which reads every kind;\n * kept for callers that sniffed it)\n * @param candidates - the delimiters to try, in priority order\n * @param quote - the quote character\n * @returns the delimiter, or null\n */\nexport function sniffDelimiter(\n    text: string,\n    newline: \"\\n\" | \"\\r\" = \"\\n\",\n    candidates: readonly string[] = DELIMITER_CANDIDATES,\n    quote = '\"',\n): string | null {\n    let best: string | null = null;\n    let bestDelta = Infinity;\n    let bestAverage = 0;\n    const terminated = text.endsWith(\"\\n\") || text.endsWith(\"\\r\");\n    for (const delimiter of candidates) {\n        if (delimiter === quote || delimiter === newline) {\n            continue;\n        }\n        let rows = splitRecords(text, delimiter, quote, PREVIEW_ROWS).filter((row) => !isBlankRow(row));\n        if (rows.length > 1 && !terminated) {\n            // the preview is a prefix of the file: its last row may be cut short\n            rows = rows.slice(0, -1);\n        }\n        if (rows.length === 0) {\n            continue;\n        }\n        let total = 0;\n        let delta = 0;\n        for (let i = 0; i < rows.length; i++) {\n            const count = rows[i].length;\n            total += count;\n            if (i > 0) {\n                delta += Math.abs(count - rows[i - 1].length);\n            }\n        }\n        const average = total / rows.length;\n        if (average > 1.99 && (delta < bestDelta || (delta === bestDelta && average > bestAverage))) {\n            best = delimiter;\n            bestDelta = delta;\n            bestAverage = average;\n        }\n    }\n    return best;\n}\n\n/**\n * Whether a parsed row is a blank line: one cell holding only whitespace.\n * @param row - the row\n * @returns true for a blank line\n */\nfunction isBlankRow(row: readonly string[]): boolean {\n    return row.length === 1 && row[0].trim().length === 0;\n}\n\n/**\n * Reads the records of one input. Iterate with `for await (const count of reader)`: each\n * iteration fills `reader.cells[0..count)` and `reader.quoted[0..count)` and sets `reader.line`\n * to the 1-based line the record started on. Blank lines (an empty unquoted record of one field)\n * are skipped. The arrays are reused between records. When the syntax gives no delimiter, the\n * first rows (PREVIEW_ROWS, or PREVIEW_CHARS characters) are buffered and the delimiter sniffed\n * from them before the first record is yielded.\n */\nexport class RecordReader implements AsyncIterable<number> {\n    /** The cells of the record most recently yielded; only the first `count` entries are valid. */\n    readonly cells: string[] = [];\n\n    /** Whether each cell of the record most recently yielded was quoted. */\n    readonly quoted: boolean[] = [];\n\n    /** The leading comment lines (without their line breaks), in order; empty without `syntax.comments`. */\n    readonly leadingComments: string[] = [];\n\n    private readonly input: ImportInput;\n\n    private readonly report: ImportReportBuilder;\n\n    private readonly readOptions: ReadOptions;\n\n    /** The delimiter (null while unsniffed), quote and comment characters. */\n    readonly syntax: RecordSyntax;\n\n    private delimiterText: string | null;\n\n    private recordLine = 0;\n\n    /**\n     * Create a reader; nothing is read until iteration starts.\n     * @param input - the input (or an async iterable of already decoded text chunks)\n     * @param report - the report decode and quoting errors are recorded in\n     * @param syntax - the delimiter (null to sniff) and quote (already checked)\n     * @param readOptions - cancellation and progress\n     */\n    constructor(input: ImportInput, report: ImportReportBuilder, syntax: RecordSyntax, readOptions: ReadOptions) {\n        this.input = input;\n        this.report = report;\n        this.readOptions = readOptions;\n        this.syntax = syntax;\n        this.delimiterText = syntax.delimiter;\n    }\n\n    /**\n     * The 1-based line the most recently yielded record started on.\n     * @returns the line number; 0 before the first record\n     */\n    get line(): number {\n        return this.recordLine;\n    }\n\n    /**\n     * The delimiter in use.\n     * @returns the given or sniffed delimiter; null before the sniff ran\n     */\n    get delimiter(): string | null {\n        return this.delimiterText;\n    }\n\n    /**\n     * Iterate the records.\n     * @yields the number of cells of each record\n     * @returns nothing\n     */\n    async *[Symbol.asyncIterator](): AsyncGenerator<number, void, undefined> {\n        const scanner = new RecordScanner(this);\n        for await (const chunk of this.chunks()) {\n            // the first chunk arrives after the sniff (or with the given delimiter)\n            scanner.setDelimiter((this.delimiterText ?? \",\").charCodeAt(0));\n            let i = 0;\n            while (i < chunk.length) {\n                i = scanner.scan(chunk, i);\n                if (scanner.ready) {\n                    scanner.ready = false;\n                    this.recordLine = scanner.recordLine;\n                    yield scanner.count;\n                    scanner.count = 0;\n                }\n            }\n            scanner.endChunk(chunk);\n        }\n        if (scanner.finish()) {\n            this.recordLine = scanner.recordLine;\n            yield scanner.count;\n        }\n    }\n\n    /**\n     * Record a text-after-quote error (fatal).\n     * @param line - the line the field started on\n     * @returns never\n     */\n    badQuote(line: number): never {\n        return this.report.fail(BAD_QUOTE_CODE, `line ${line}: text after the closing quote of a quoted field`, {\n            line,\n        });\n    }\n\n    /**\n     * Record an unclosed-quote error (fatal).\n     * @param line - the line the field started on\n     * @returns never\n     */\n    unclosedQuote(line: number): never {\n        return this.report.fail(UNCLOSED_QUOTE_CODE, `a quoted field starting on line ${line} is never closed`, {\n            line,\n        });\n    }\n\n    /**\n     * The decoded text chunks of the input; when the delimiter is to be sniffed, the first rows\n     * are buffered (bounded by PREVIEW_ROWS line breaks or PREVIEW_CHARS characters), the sniff\n     * runs, and the buffered text is yielded as one chunk before the rest streams through.\n     * @yields the text chunks\n     * @returns nothing\n     */\n    private async *chunks(): AsyncGenerator<string, void, undefined> {\n        const source = textChunks(this.input, this.report, this.readOptions);\n        if (this.delimiterText !== null) {\n            yield* source;\n            return;\n        }\n        const pieces: string[] = [];\n        let length = 0;\n        let breaks = 0;\n        let lastWasCr = false;\n        let sniffed = false;\n        for await (const chunk of source) {\n            if (sniffed) {\n                yield chunk;\n                continue;\n            }\n            pieces.push(chunk);\n            length += chunk.length;\n            for (let i = 0; i < chunk.length && breaks < PREVIEW_ROWS; i++) {\n                const c = chunk.charCodeAt(i);\n                if (c === LF) {\n                    if (!lastWasCr) {\n                        breaks++;\n                    }\n                } else if (c === CR) {\n                    breaks++;\n                }\n                lastWasCr = c === CR;\n            }\n            if (breaks < PREVIEW_ROWS && length < PREVIEW_CHARS) {\n                continue;\n            }\n            const text = pieces.join(\"\");\n            pieces.length = 0;\n            this.sniff(text);\n            sniffed = true;\n            yield text;\n        }\n        if (!sniffed) {\n            const text = pieces.join(\"\");\n            if (text.length > 0) {\n                this.sniff(text);\n                yield text;\n            }\n        }\n    }\n\n    /**\n     * Sniff the delimiter from the preview text and record it.\n     * @param text - the preview\n     */\n    private sniff(text: string): void {\n        const candidates = this.syntax.candidates ?? DELIMITER_CANDIDATES;\n        // the sniff is a heuristic over the first rows: a single chunk holding a huge quoted cell\n        // is capped so the candidate scans stay bounded\n        const body = stripLeadingComments(text.slice(0, PREVIEW_CHARS), this.syntax.comments ?? []);\n        this.delimiterText = sniffDelimiter(body, sniffNewline(body), candidates, this.syntax.quote) ?? candidates[0];\n    }\n}\n\n/**\n * The record state machine of RecordReader, resumable at every record end so the reader can\n * yield a record before the next one overwrites the shared cell arrays. Every state lives on the\n * instance; `scan()` runs the per-character loop over one chunk from a position and returns where\n * it stopped (after the character that completed a record, with `ready` set, or the chunk end).\n */\nclass RecordScanner {\n    /** Whether a record just completed. */\n    ready = false;\n\n    /** The cell count of the record in progress (the completed one while `ready`). */\n    count = 0;\n\n    /** The line the completed record started on (valid while `ready`, and after finish()). */\n    recordLine = 1;\n\n    /** The line the record in progress started on. */\n    private startLine = 1;\n\n    private readonly reader: RecordReader;\n\n    private readonly quoteCode: number;\n\n    private readonly quoteText: string;\n\n    private readonly commentCodes: readonly number[];\n\n    private delimiterCode = -1;\n\n    private state: State = START;\n\n    private field = \"\";\n\n    private fieldQuoted = false;\n\n    private line = 1;\n\n    private lastWasCr = false;\n\n    /** Whether no record has started yet (comment lines are recognised until one does). */\n    private leading: boolean;\n\n    private comment = \"\";\n\n    /** The start of the unconsumed run of the current chunk that belongs to the field or comment. */\n    private segment = 0;\n\n    /**\n     * Create a scanner over a reader's cell arrays.\n     * @param reader - the reader\n     */\n    constructor(reader: RecordReader) {\n        this.reader = reader;\n        this.quoteCode = reader.syntax.quote.charCodeAt(0);\n        this.quoteText = reader.syntax.quote;\n        this.commentCodes = (reader.syntax.comments ?? []).map((c) => c.charCodeAt(0));\n        this.leading = this.commentCodes.length > 0;\n    }\n\n    /**\n     * Set the delimiter once it is known (before the first chunk is scanned).\n     * @param code - the delimiter's character code\n     */\n    setDelimiter(code: number): void {\n        if (this.delimiterCode < 0) {\n            this.delimiterCode = code;\n        }\n    }\n\n    /**\n     * Scan one chunk from a position until a record completes or the chunk ends.\n     * @param chunk - the chunk\n     * @param from - where to start\n     * @returns the position after the last character consumed\n     */\n    scan(chunk: string, from: number): number {\n        const { quoteCode, delimiterCode } = this;\n        const n = chunk.length;\n        let i = from;\n        this.segment = from;\n        while (i < n) {\n            const c = chunk.charCodeAt(i);\n            if (this.state === COMMENT) {\n                if (c === LF || c === CR) {\n                    this.comment += chunk.slice(this.segment, i);\n                    this.reader.leadingComments.push(this.comment);\n                    this.comment = \"\";\n                    this.state = START;\n                    this.endLine(c);\n                    this.segment = i + 1;\n                }\n                i++;\n                continue;\n            }\n            if (this.state === QUOTED) {\n                // jump to the next quote; the run in between is field text whose line breaks are counted\n                const q = chunk.indexOf(this.quoteText, i);\n                const end = q < 0 ? n : q;\n                if (end > i) {\n                    this.countLineBreaks(chunk, i, end);\n                }\n                if (q < 0) {\n                    i = n;\n                    break;\n                }\n                this.field += chunk.slice(this.segment, q);\n                this.state = CLOSING;\n                this.lastWasCr = false;\n                i = q + 1;\n                continue;\n            }\n            if (this.state === CLOSING) {\n                if (c === quoteCode) {\n                    this.field += this.quoteText;\n                    this.segment = i + 1;\n                    this.state = QUOTED;\n                    this.lastWasCr = false;\n                    i++;\n                    continue;\n                }\n                if (c !== delimiterCode && c !== LF && c !== CR) {\n                    this.reader.badQuote(this.startLine);\n                }\n                this.state = AFTER_QUOTED;\n                this.segment = i;\n                // fall through to the delimiter / line-end handling below without consuming c\n            }\n            if (c === delimiterCode || c === LF || c === CR) {\n                if (c === LF && this.lastWasCr) {\n                    // the second half of a CRLF already ended the record\n                    this.lastWasCr = false;\n                    this.segment = i + 1;\n                    i++;\n                    continue;\n                }\n                if (this.state === UNQUOTED) {\n                    this.field += chunk.slice(this.segment, i);\n                }\n                if (c === delimiterCode) {\n                    this.push();\n                    this.state = START;\n                    this.lastWasCr = false;\n                    this.segment = i + 1;\n                    i++;\n                    continue;\n                }\n                const blank = this.state === START && this.count === 0;\n                if (!blank) {\n                    this.push();\n                    this.ready = true;\n                    this.recordLine = this.startLine;\n                }\n                this.state = START;\n                this.endLine(c);\n                this.segment = i + 1;\n                i++;\n                if (this.ready) {\n                    return i;\n                }\n                continue;\n            }\n            this.lastWasCr = false;\n            if (this.state === START) {\n                if (this.leading && this.count === 0 && this.commentCodes.includes(c)) {\n                    this.state = COMMENT;\n                    this.segment = i;\n                    i++;\n                    continue;\n                }\n                this.leading = false;\n                if (c === quoteCode) {\n                    this.state = QUOTED;\n                    this.fieldQuoted = true;\n                    this.segment = i + 1;\n                } else {\n                    this.state = UNQUOTED;\n                    this.segment = i;\n                }\n            }\n            i++;\n        }\n        return i;\n    }\n\n    /**\n     * Keep the tail of a chunk that belongs to an open field or comment.\n     * @param chunk - the chunk just scanned to its end\n     */\n    endChunk(chunk: string): void {\n        const n = chunk.length;\n        if (this.segment >= n) {\n            return;\n        }\n        if (this.state === COMMENT) {\n            this.comment += chunk.slice(this.segment, n);\n        } else if (this.state === QUOTED || this.state === UNQUOTED) {\n            this.field += chunk.slice(this.segment, n);\n        }\n        this.segment = n;\n    }\n\n    /**\n     * The end of the input: an open comment is kept, an open quote is fatal, an open record\n     * (no trailing line break) is completed.\n     * @returns true when a last record is ready\n     */\n    finish(): boolean {\n        const { state } = this;\n        if (state === COMMENT) {\n            this.reader.leadingComments.push(this.comment);\n            return false;\n        }\n        if (state === QUOTED) {\n            this.reader.unclosedQuote(this.startLine);\n        }\n        if (state !== START || this.count > 0) {\n            this.push();\n            this.recordLine = this.startLine;\n            return true;\n        }\n        return false;\n    }\n\n    /**\n     * Count the line breaks of a run of quoted text (a CRLF counts once, like the record loop).\n     * @param chunk - the chunk\n     * @param from - the first index of the run\n     * @param to - the index after the run\n     */\n    private countLineBreaks(chunk: string, from: number, to: number): void {\n        let { line, lastWasCr } = this;\n        for (let i = from; i < to; i++) {\n            const c = chunk.charCodeAt(i);\n            if (c === LF) {\n                if (!lastWasCr) {\n                    line++;\n                }\n                lastWasCr = false;\n            } else if (c === CR) {\n                line++;\n                lastWasCr = true;\n            } else {\n                lastWasCr = false;\n            }\n        }\n        this.line = line;\n        this.lastWasCr = lastWasCr;\n    }\n\n    /** Store the field in progress as the next cell. */\n    private push(): void {\n        const { reader, count } = this;\n        reader.cells[count] = this.field;\n        reader.quoted[count] = this.fieldQuoted;\n        this.count = count + 1;\n        this.field = \"\";\n        this.fieldQuoted = false;\n    }\n\n    /**\n     * Count a line break that ends a record or a comment line.\n     * @param c - the LF or CR\n     */\n    private endLine(c: number): void {\n        if (!(c === LF && this.lastWasCr)) {\n            this.line++;\n        }\n        this.startLine = this.line;\n        this.lastWasCr = c === CR;\n    }\n}\n\n/** What the CSV importer's reader needs besides the input. */\nexport interface CsvReaderOptions extends ReadOptions {\n    /** The delimiter; null or undefined sniffs it from the preview. */\n    readonly delimiter?: string | null | undefined;\n    /** The characters opening a leading comment line (RecordSyntax.comments); none by default. */\n    readonly comments?: readonly string[] | undefined;\n}\n\n/**\n * Records of a CSV input, one string array per row, streamed: iterate with\n * `for await (const row of reader)` and read `reader.line` for the 1-based line the row starts\n * on. The delimiter is known after the first row (`reader.delimiter`). Rows come out as fresh\n * arrays of cell texts exactly as written (no trimming, no typing); `reader.quoted` tells, for\n * the row just yielded, which cells were quoted (a quoted empty cell is the empty string, an\n * unquoted one is \"not set\"). Blank lines and whitespace-only single-cell lines are skipped.\n */\nexport class CsvRecordReader implements AsyncIterable<string[]> {\n    private readonly inner: RecordReader;\n\n    /**\n     * Create a reader; nothing is read until iteration starts.\n     * @param input - the input\n     * @param report - the report parse errors are recorded in\n     * @param options - the delimiter, cancellation and progress\n     */\n    constructor(input: ImportInput, report: ImportReportBuilder, options: CsvReaderOptions = {}) {\n        this.inner = new RecordReader(\n            input,\n            report,\n            { delimiter: options.delimiter ?? null, quote: '\"', comments: options.comments },\n            { signal: options.signal, onProgress: options.onProgress, encoding: options.encoding },\n        );\n    }\n\n    /**\n     * The leading comment lines the reader skipped (SNAP `#` lines, the KONECT `%` header), in order.\n     * @returns the lines without their line breaks\n     */\n    get leadingComments(): readonly string[] {\n        return this.inner.leadingComments;\n    }\n\n    /**\n     * The 1-based line the most recently yielded row starts on.\n     * @returns the line; 0 before the first row\n     */\n    get line(): number {\n        return this.inner.line;\n    }\n\n    /**\n     * The delimiter in use.\n     * @returns the given or sniffed delimiter; null before the sniff ran\n     */\n    get delimiter(): string | null {\n        return this.inner.delimiter;\n    }\n\n    /**\n     * Whether each cell of the row most recently yielded was quoted.\n     * @returns the flags (valid for the first `row.length` entries)\n     */\n    get quoted(): readonly boolean[] {\n        return this.inner.quoted;\n    }\n\n    /**\n     * Iterate the rows.\n     * @yields one row at a time as its cell texts\n     * @returns nothing\n     */\n    async *[Symbol.asyncIterator](): AsyncGenerator<string[], void, undefined> {\n        const { inner } = this;\n        for await (const count of inner) {\n            if (count === 1 && !inner.quoted[0] && inner.cells[0].trim().length === 0) {\n                continue;\n            }\n            yield inner.cells.slice(0, count);\n        }\n    }\n}\n"],"names":[],"mappings":";;AAuBO,MAAM,sBAAsB;AAG5B,MAAM,iBAAiB;AAG9B,MAAM,uBAA0C,OAAO,OAAO,CAAC,KAAK,KAAM,KAAK,KAAK,GAAG,CAAC;AAGxF,MAAM,eAAe;AAGrB,MAAM,gBAAgB,KAAK;AAE3B,MAAM,KAAK;AACX,MAAM,KAAK;AAGX,MAAM,QAAQ;AAEd,MAAM,WAAW;AAEjB,MAAM,SAAS;AAEf,MAAM,UAAU;AAEhB,MAAM,eAAe;AAErB,MAAM,UAAU;AA0BT,SAAS,kBAAkB,QAAoC;AAClE,aAAW,CAAC,QAAQ,KAAK,KAAK;AAAA,IAC1B,CAAC,aAAa,OAAO,SAAS;AAAA,IAC9B,CAAC,SAAS,OAAO,KAAK;AAAA,EAAA,GACd;AACR,QAAI,UAAU,MAAM;AAChB;AAAA,IACJ;AACA,QAAI,MAAM,WAAW,KAAK,UAAU,QAAQ,UAAU,MAAM;AACxD,YAAM,IAAI;AAAA,QACN;AAAA,QACA,UAAU,MAAM,KAAK,KAAK,UAAU,KAAK,CAAC;AAAA,QAC1C,EAAE,QAAQ,OAAO,MAAA;AAAA,MAAM;AAAA,IAE/B;AAAA,EACJ;AACA,MAAI,OAAO,cAAc,QAAQ,OAAO,cAAc,OAAO,OAAO;AAChE,UAAM,IAAI,iBAAiB,iBAAiB,2CAA2C;AAAA,MACnF,QAAQ;AAAA,MACR,OAAO,OAAO;AAAA,IAAA,CACjB;AAAA,EACL;AACA,SAAO;AACX;AAQO,SAAS,aAAa,MAA2B;AACpD,SAAO,CAAC,KAAK,SAAS,IAAI,KAAK,KAAK,SAAS,IAAI,IAAI,OAAO;AAChE;AAWA,SAAS,aAAa,MAAc,WAAmB,OAAe,SAA6B;AAC/F,QAAM,OAAmB,CAAA;AACzB,QAAM,gBAAgB,UAAU,WAAW,CAAC;AAC5C,QAAM,YAAY,MAAM,WAAW,CAAC;AACpC,MAAI,QAAkB,CAAA;AACtB,MAAI,QAAe;AACnB,MAAI,UAAU;AACd,MAAI,QAAQ;AACZ,QAAM,IAAI,KAAK;AACf,WAAS,IAAI,GAAG,IAAI,KAAK,KAAK,SAAS,SAAS,KAAK;AACjD,UAAM,IAAI,KAAK,WAAW,CAAC;AAC3B,QAAI,UAAU,QAAQ;AAClB,UAAI,MAAM,WAAW;AACjB,iBAAS,KAAK,MAAM,SAAS,CAAC;AAC9B,gBAAQ;AAAA,MACZ;AACA;AAAA,IACJ;AACA,QAAI,UAAU,SAAS;AACnB,UAAI,MAAM,WAAW;AACjB,iBAAS;AACT,kBAAU,IAAI;AACd,gBAAQ;AACR;AAAA,MACJ;AACA,cAAQ;AACR,gBAAU;AAAA,IACd;AACA,QAAI,MAAM,iBAAiB,MAAM,MAAM,MAAM,IAAI;AAC7C,UAAI,UAAU,YAAY,UAAU,cAAc;AAC9C,iBAAS,KAAK,MAAM,SAAS,CAAC;AAAA,MAClC;AACA,UAAI,MAAM,eAAe;AACrB,cAAM,KAAK,KAAK;AAChB,gBAAQ;AACR,gBAAQ;AAAA,MACZ,OAAO;AACH,YAAI,UAAU,SAAS,MAAM,SAAS,GAAG;AACrC,gBAAM,KAAK,KAAK;AAChB,eAAK,KAAK,KAAK;AACf,kBAAQ,CAAA;AACR,kBAAQ;AAAA,QACZ;AACA,gBAAQ;AACR,YAAI,MAAM,MAAM,KAAK,WAAW,IAAI,CAAC,MAAM,IAAI;AAC3C;AAAA,QACJ;AAAA,MACJ;AACA,gBAAU,IAAI;AACd;AAAA,IACJ;AACA,QAAI,UAAU,OAAO;AACjB,UAAI,MAAM,WAAW;AACjB,gBAAQ;AACR,kBAAU,IAAI;AAAA,MAClB,OAAO;AACH,gBAAQ;AACR,kBAAU;AAAA,MACd;AAAA,IACJ;AAAA,EACJ;AACA,MAAI,KAAK,SAAS,YAAY,UAAU,SAAS,MAAM,SAAS,IAAI;AAChE,QAAI,UAAU,YAAY,UAAU,gBAAgB,UAAU,QAAQ;AAClE,eAAS,KAAK,MAAM,SAAS,CAAC;AAAA,IAClC;AACA,UAAM,KAAK,KAAK;AAChB,SAAK,KAAK,KAAK;AAAA,EACnB;AACA,SAAO;AACX;AASA,SAAS,qBAAqB,MAAc,UAAqC;AAC7E,MAAI,SAAS,WAAW,GAAG;AACvB,WAAO;AAAA,EACX;AACA,MAAI,KAAK;AACT,SAAO,KAAK,KAAK,UAAU,SAAS,SAAS,KAAK,EAAE,CAAC,GAAG;AACpD,UAAM,KAAK,KAAK,QAAQ,MAAM,EAAE;AAChC,UAAM,KAAK,KAAK,QAAQ,MAAM,EAAE;AAChC,UAAM,MAAM,KAAK,IAAI,KAAK,IAAI,WAAW,IAAI,KAAK,IAAI,WAAW,EAAE;AACnE,QAAI,QAAQ,UAAU;AAClB,aAAO;AAAA,IACX;AACA,SAAK,MAAM;AACX,QAAI,KAAK,KAAK,CAAC,MAAM,QAAQ,KAAK,EAAE,MAAM,MAAM;AAC5C;AAAA,IACJ;AAAA,EACJ;AACA,SAAO,OAAO,IAAI,OAAO,KAAK,MAAM,EAAE;AAC1C;AAcO,SAAS,eACZ,MACA,UAAuB,MACvB,aAAgC,sBAChC,QAAQ,KACK;AACb,MAAI,OAAsB;AAC1B,MAAI,YAAY;AAChB,MAAI,cAAc;AAClB,QAAM,aAAa,KAAK,SAAS,IAAI,KAAK,KAAK,SAAS,IAAI;AAC5D,aAAW,aAAa,YAAY;AAChC,QAAI,cAAc,SAAS,cAAc,SAAS;AAC9C;AAAA,IACJ;AACA,QAAI,OAAO,aAAa,MAAM,WAAW,OAAO,YAAY,EAAE,OAAO,CAAC,QAAQ,CAAC,WAAW,GAAG,CAAC;AAC9F,QAAI,KAAK,SAAS,KAAK,CAAC,YAAY;AAEhC,aAAO,KAAK,MAAM,GAAG,EAAE;AAAA,IAC3B;AACA,QAAI,KAAK,WAAW,GAAG;AACnB;AAAA,IACJ;AACA,QAAI,QAAQ;AACZ,QAAI,QAAQ;AACZ,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AAClC,YAAM,QAAQ,KAAK,CAAC,EAAE;AACtB,eAAS;AACT,UAAI,IAAI,GAAG;AACP,iBAAS,KAAK,IAAI,QAAQ,KAAK,IAAI,CAAC,EAAE,MAAM;AAAA,MAChD;AAAA,IACJ;AACA,UAAM,UAAU,QAAQ,KAAK;AAC7B,QAAI,UAAU,SAAS,QAAQ,aAAc,UAAU,aAAa,UAAU,cAAe;AACzF,aAAO;AACP,kBAAY;AACZ,oBAAc;AAAA,IAClB;AAAA,EACJ;AACA,SAAO;AACX;AAOA,SAAS,WAAW,KAAiC;AACjD,SAAO,IAAI,WAAW,KAAK,IAAI,CAAC,EAAE,OAAO,WAAW;AACxD;AAUO,MAAM,aAA8C;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA8BvD,YAAY,OAAoB,QAA6B,QAAsB,aAA0B;AA5B7G,SAAS,QAAkB,CAAA;AAG3B,SAAS,SAAoB,CAAA;AAG7B,SAAS,kBAA4B,CAAA;AAarC,SAAQ,aAAa;AAUjB,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,cAAc;AACnB,SAAK,SAAS;AACd,SAAK,gBAAgB,OAAO;AAAA,EAChC;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,OAAe;AACf,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC3B,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,QAAQ,OAAO,aAAa,IAA6C;AACrE,UAAM,UAAU,IAAI,cAAc,IAAI;AACtC,qBAAiB,SAAS,KAAK,UAAU;AAErC,cAAQ,cAAc,KAAK,iBAAiB,KAAK,WAAW,CAAC,CAAC;AAC9D,UAAI,IAAI;AACR,aAAO,IAAI,MAAM,QAAQ;AACrB,YAAI,QAAQ,KAAK,OAAO,CAAC;AACzB,YAAI,QAAQ,OAAO;AACf,kBAAQ,QAAQ;AAChB,eAAK,aAAa,QAAQ;AAC1B,gBAAM,QAAQ;AACd,kBAAQ,QAAQ;AAAA,QACpB;AAAA,MACJ;AACA,cAAQ,SAAS,KAAK;AAAA,IAC1B;AACA,QAAI,QAAQ,UAAU;AAClB,WAAK,aAAa,QAAQ;AAC1B,YAAM,QAAQ;AAAA,IAClB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAS,MAAqB;AAC1B,WAAO,KAAK,OAAO,KAAK,gBAAgB,QAAQ,IAAI,oDAAoD;AAAA,MACpG;AAAA,IAAA,CACH;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAAqB;AAC/B,WAAO,KAAK,OAAO,KAAK,qBAAqB,mCAAmC,IAAI,oBAAoB;AAAA,MACpG;AAAA,IAAA,CACH;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAe,SAAkD;AAC7D,UAAM,SAAS,WAAW,KAAK,OAAO,KAAK,QAAQ,KAAK,WAAW;AACnE,QAAI,KAAK,kBAAkB,MAAM;AAC7B,aAAO;AACP;AAAA,IACJ;AACA,UAAM,SAAmB,CAAA;AACzB,QAAI,SAAS;AACb,QAAI,SAAS;AACb,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,qBAAiB,SAAS,QAAQ;AAC9B,UAAI,SAAS;AACT,cAAM;AACN;AAAA,MACJ;AACA,aAAO,KAAK,KAAK;AACjB,gBAAU,MAAM;AAChB,eAAS,IAAI,GAAG,IAAI,MAAM,UAAU,SAAS,cAAc,KAAK;AAC5D,cAAM,IAAI,MAAM,WAAW,CAAC;AAC5B,YAAI,MAAM,IAAI;AACV,cAAI,CAAC,WAAW;AACZ;AAAA,UACJ;AAAA,QACJ,WAAW,MAAM,IAAI;AACjB;AAAA,QACJ;AACA,oBAAY,MAAM;AAAA,MACtB;AACA,UAAI,SAAS,gBAAgB,SAAS,eAAe;AACjD;AAAA,MACJ;AACA,YAAM,OAAO,OAAO,KAAK,EAAE;AAC3B,aAAO,SAAS;AAChB,WAAK,MAAM,IAAI;AACf,gBAAU;AACV,YAAM;AAAA,IACV;AACA,QAAI,CAAC,SAAS;AACV,YAAM,OAAO,OAAO,KAAK,EAAE;AAC3B,UAAI,KAAK,SAAS,GAAG;AACjB,aAAK,MAAM,IAAI;AACf,cAAM;AAAA,MACV;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,MAAM,MAAoB;AAC9B,UAAM,aAAa,KAAK,OAAO,cAAc;AAG7C,UAAM,OAAO,qBAAqB,KAAK,MAAM,GAAG,aAAa,GAAG,KAAK,OAAO,YAAY,CAAA,CAAE;AAC1F,SAAK,gBAAgB,eAAe,MAAM,aAAa,IAAI,GAAG,YAAY,KAAK,OAAO,KAAK,KAAK,WAAW,CAAC;AAAA,EAChH;AACJ;AAQA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA;AAAA,EA6ChB,YAAY,QAAsB;AA3ClC,SAAA,QAAQ;AAGR,SAAA,QAAQ;AAGR,SAAA,aAAa;AAGb,SAAQ,YAAY;AAUpB,SAAQ,gBAAgB;AAExB,SAAQ,QAAe;AAEvB,SAAQ,QAAQ;AAEhB,SAAQ,cAAc;AAEtB,SAAQ,OAAO;AAEf,SAAQ,YAAY;AAKpB,SAAQ,UAAU;AAGlB,SAAQ,UAAU;AAOd,SAAK,SAAS;AACd,SAAK,YAAY,OAAO,OAAO,MAAM,WAAW,CAAC;AACjD,SAAK,YAAY,OAAO,OAAO;AAC/B,SAAK,gBAAgB,OAAO,OAAO,YAAY,CAAA,GAAI,IAAI,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC;AAC7E,SAAK,UAAU,KAAK,aAAa,SAAS;AAAA,EAC9C;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,aAAa,MAAoB;AAC7B,QAAI,KAAK,gBAAgB,GAAG;AACxB,WAAK,gBAAgB;AAAA,IACzB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,KAAK,OAAe,MAAsB;AACtC,UAAM,EAAE,WAAW,cAAA,IAAkB;AACrC,UAAM,IAAI,MAAM;AAChB,QAAI,IAAI;AACR,SAAK,UAAU;AACf,WAAO,IAAI,GAAG;AACV,YAAM,IAAI,MAAM,WAAW,CAAC;AAC5B,UAAI,KAAK,UAAU,SAAS;AACxB,YAAI,MAAM,MAAM,MAAM,IAAI;AACtB,eAAK,WAAW,MAAM,MAAM,KAAK,SAAS,CAAC;AAC3C,eAAK,OAAO,gBAAgB,KAAK,KAAK,OAAO;AAC7C,eAAK,UAAU;AACf,eAAK,QAAQ;AACb,eAAK,QAAQ,CAAC;AACd,eAAK,UAAU,IAAI;AAAA,QACvB;AACA;AACA;AAAA,MACJ;AACA,UAAI,KAAK,UAAU,QAAQ;AAEvB,cAAM,IAAI,MAAM,QAAQ,KAAK,WAAW,CAAC;AACzC,cAAM,MAAM,IAAI,IAAI,IAAI;AACxB,YAAI,MAAM,GAAG;AACT,eAAK,gBAAgB,OAAO,GAAG,GAAG;AAAA,QACtC;AACA,YAAI,IAAI,GAAG;AACP,cAAI;AACJ;AAAA,QACJ;AACA,aAAK,SAAS,MAAM,MAAM,KAAK,SAAS,CAAC;AACzC,aAAK,QAAQ;AACb,aAAK,YAAY;AACjB,YAAI,IAAI;AACR;AAAA,MACJ;AACA,UAAI,KAAK,UAAU,SAAS;AACxB,YAAI,MAAM,WAAW;AACjB,eAAK,SAAS,KAAK;AACnB,eAAK,UAAU,IAAI;AACnB,eAAK,QAAQ;AACb,eAAK,YAAY;AACjB;AACA;AAAA,QACJ;AACA,YAAI,MAAM,iBAAiB,MAAM,MAAM,MAAM,IAAI;AAC7C,eAAK,OAAO,SAAS,KAAK,SAAS;AAAA,QACvC;AACA,aAAK,QAAQ;AACb,aAAK,UAAU;AAAA,MAEnB;AACA,UAAI,MAAM,iBAAiB,MAAM,MAAM,MAAM,IAAI;AAC7C,YAAI,MAAM,MAAM,KAAK,WAAW;AAE5B,eAAK,YAAY;AACjB,eAAK,UAAU,IAAI;AACnB;AACA;AAAA,QACJ;AACA,YAAI,KAAK,UAAU,UAAU;AACzB,eAAK,SAAS,MAAM,MAAM,KAAK,SAAS,CAAC;AAAA,QAC7C;AACA,YAAI,MAAM,eAAe;AACrB,eAAK,KAAA;AACL,eAAK,QAAQ;AACb,eAAK,YAAY;AACjB,eAAK,UAAU,IAAI;AACnB;AACA;AAAA,QACJ;AACA,cAAM,QAAQ,KAAK,UAAU,SAAS,KAAK,UAAU;AACrD,YAAI,CAAC,OAAO;AACR,eAAK,KAAA;AACL,eAAK,QAAQ;AACb,eAAK,aAAa,KAAK;AAAA,QAC3B;AACA,aAAK,QAAQ;AACb,aAAK,QAAQ,CAAC;AACd,aAAK,UAAU,IAAI;AACnB;AACA,YAAI,KAAK,OAAO;AACZ,iBAAO;AAAA,QACX;AACA;AAAA,MACJ;AACA,WAAK,YAAY;AACjB,UAAI,KAAK,UAAU,OAAO;AACtB,YAAI,KAAK,WAAW,KAAK,UAAU,KAAK,KAAK,aAAa,SAAS,CAAC,GAAG;AACnE,eAAK,QAAQ;AACb,eAAK,UAAU;AACf;AACA;AAAA,QACJ;AACA,aAAK,UAAU;AACf,YAAI,MAAM,WAAW;AACjB,eAAK,QAAQ;AACb,eAAK,cAAc;AACnB,eAAK,UAAU,IAAI;AAAA,QACvB,OAAO;AACH,eAAK,QAAQ;AACb,eAAK,UAAU;AAAA,QACnB;AAAA,MACJ;AACA;AAAA,IACJ;AACA,WAAO;AAAA,EACX;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,SAAS,OAAqB;AAC1B,UAAM,IAAI,MAAM;AAChB,QAAI,KAAK,WAAW,GAAG;AACnB;AAAA,IACJ;AACA,QAAI,KAAK,UAAU,SAAS;AACxB,WAAK,WAAW,MAAM,MAAM,KAAK,SAAS,CAAC;AAAA,IAC/C,WAAW,KAAK,UAAU,UAAU,KAAK,UAAU,UAAU;AACzD,WAAK,SAAS,MAAM,MAAM,KAAK,SAAS,CAAC;AAAA,IAC7C;AACA,SAAK,UAAU;AAAA,EACnB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAkB;AACd,UAAM,EAAE,UAAU;AAClB,QAAI,UAAU,SAAS;AACnB,WAAK,OAAO,gBAAgB,KAAK,KAAK,OAAO;AAC7C,aAAO;AAAA,IACX;AACA,QAAI,UAAU,QAAQ;AAClB,WAAK,OAAO,cAAc,KAAK,SAAS;AAAA,IAC5C;AACA,QAAI,UAAU,SAAS,KAAK,QAAQ,GAAG;AACnC,WAAK,KAAA;AACL,WAAK,aAAa,KAAK;AACvB,aAAO;AAAA,IACX;AACA,WAAO;AAAA,EACX;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQQ,gBAAgB,OAAe,MAAc,IAAkB;AACnE,QAAI,EAAE,MAAM,UAAA,IAAc;AAC1B,aAAS,IAAI,MAAM,IAAI,IAAI,KAAK;AAC5B,YAAM,IAAI,MAAM,WAAW,CAAC;AAC5B,UAAI,MAAM,IAAI;AACV,YAAI,CAAC,WAAW;AACZ;AAAA,QACJ;AACA,oBAAY;AAAA,MAChB,WAAW,MAAM,IAAI;AACjB;AACA,oBAAY;AAAA,MAChB,OAAO;AACH,oBAAY;AAAA,MAChB;AAAA,IACJ;AACA,SAAK,OAAO;AACZ,SAAK,YAAY;AAAA,EACrB;AAAA;AAAA,EAGQ,OAAa;AACjB,UAAM,EAAE,QAAQ,MAAA,IAAU;AAC1B,WAAO,MAAM,KAAK,IAAI,KAAK;AAC3B,WAAO,OAAO,KAAK,IAAI,KAAK;AAC5B,SAAK,QAAQ,QAAQ;AACrB,SAAK,QAAQ;AACb,SAAK,cAAc;AAAA,EACvB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,QAAQ,GAAiB;AAC7B,QAAI,EAAE,MAAM,MAAM,KAAK,YAAY;AAC/B,WAAK;AAAA,IACT;AACA,SAAK,YAAY,KAAK;AACtB,SAAK,YAAY,MAAM;AAAA,EAC3B;AACJ;AAkBO,MAAM,gBAAmD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS5D,YAAY,OAAoB,QAA6B,UAA4B,CAAA,GAAI;AACzF,SAAK,QAAQ,IAAI;AAAA,MACb;AAAA,MACA;AAAA,MACA,EAAE,WAAW,QAAQ,aAAa,MAAM,OAAO,KAAK,UAAU,QAAQ,SAAA;AAAA,MACtE,EAAE,QAAQ,QAAQ,QAAQ,YAAY,QAAQ,YAAY,UAAU,QAAQ,SAAA;AAAA,IAAS;AAAA,EAE7F;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,kBAAqC;AACrC,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,OAAe;AACf,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC3B,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,SAA6B;AAC7B,WAAO,KAAK,MAAM;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,QAAQ,OAAO,aAAa,IAA+C;AACvE,UAAM,EAAE,UAAU;AAClB,qBAAiB,SAAS,OAAO;AAC7B,UAAI,UAAU,KAAK,CAAC,MAAM,OAAO,CAAC,KAAK,MAAM,MAAM,CAAC,EAAE,KAAA,EAAO,WAAW,GAAG;AACvE;AAAA,MACJ;AACA,YAAM,MAAM,MAAM,MAAM,GAAG,KAAK;AAAA,IACpC;AAAA,EACJ;AACJ;"}

package/dist/chunks/{text-CajMdVFy.js → text-Dr0Ifpag.js}

@@ -177,7 +177,7 @@ function kindOfValue(value) {
   if (typeof value === "string") {
     return "string";
   }
-  return Number.isInteger(value) && value >= -2147483648 && value <= 2147483647 && !Object.is(value, -0) ? "i32" : "f64";
+  return Number.isInteger(value) && value >= -2147483648 && value <= 2147483647 ? "i32" : "f64";
 }
 export {
   TextCellWriter as T,
@@ -186,4 +186,4 @@ export {
   inferTextDtype as i,
   parseTextCell as p
 };
-//# sourceMappingURL=text-CajMdVFy.js.map
+//# sourceMappingURL=text-Dr0Ifpag.js.map

package/dist/chunks/{text-CajMdVFy.js.map → text-Dr0Ifpag.js.map}

@@ -1 +1 @@
-{"version":3,"file":"text-CajMdVFy.js","sources":["../../src/common/text.ts"],"sourcesContent":["/**\n * The fixed lexical grammar of design section 5.1 for untyped text sources (CSV cells, GML\n * values, DOT strings, Pajek tokens), reproduced from the core's reference implementation so the\n * two agree (invariant I15). Text importers parse each cell with parseTextCell() and push the JS\n * value; the sink's per-column inference then widens `bool -> i32 -> f64 -> string` and never per\n * cell, so a column that saw `1` and then `\"01\"` becomes string for all rows.\n *\n * - `bool` is exactly `true` / `false` (case-sensitive);\n * - `i32` is `/^-?(0|[1-9][0-9]*)$/` within `[-2^31, 2^31)`; `0` / `1` are i32, never bool;\n * - `f64` is a decimal or exponent literal accepted by `Number()` that is not empty, not\n *   whitespace, not `Infinity` / `NaN`, not a hex / octal / binary form, has no leading zeros in\n *   its integer part, and whose value is finite (an integer literal outside i32 range is f64);\n * - everything else is `string`.\n */\n\nimport { type ColumnHandle, GraphFormatError, type GraphSink, INVALID_INDEX } from \"@graphty/graph-format\";\n\nimport { type ImportReportBuilder } from \"./report.js\";\n\nconst I32_TEXT = /^-?(0|[1-9][0-9]*)$/;\nconst F64_TEXT = /^[+-]?((0|[1-9][0-9]*)(\\.[0-9]*)?|\\.[0-9]+)([eE][+-]?[0-9]+)?$/;\nconst I32_MIN = -2147483648;\nconst I32_MAX = 2147483647;\n\n/**\n * The dtype a text cell parses as under the design section 5.1 grammar.\n * Consumed by the per-format importers and exporters under src/formats.\n * @public\n */\nexport type TextDtype = \"bool\" | \"i32\" | \"f64\" | \"string\";\n\n/**\n * Classify one text cell by the fixed grammar.\n * @param text - the cell text, exactly as read (no trimming)\n * @returns bool, i32, f64 or string\n */\nexport function inferTextDtype(text: string): TextDtype {\n    if (text === \"true\" || text === \"false\") {\n        return \"bool\";\n    }\n    if (I32_TEXT.test(text)) {\n        const n = Number(text);\n        if (n >= I32_MIN && n <= I32_MAX) {\n            return \"i32\";\n        }\n        return Number.isFinite(n) ? \"f64\" : \"string\";\n    }\n    if (F64_TEXT.test(text) && Number.isFinite(Number(text))) {\n        return \"f64\";\n    }\n    return \"string\";\n}\n\n/**\n * Parse one text cell into the JS value its grammar class implies: a boolean for bool, a number for\n * i32 / f64, the text itself for string. The sink infers the column dtype from the value.\n * @param text - the cell text, exactly as read\n * @returns the value\n */\nexport function parseTextCell(text: string): boolean | number | string {\n    switch (inferTextDtype(text)) {\n        case \"bool\":\n            return text === \"true\";\n        case \"i32\":\n        case \"f64\":\n            return Number(text);\n        case \"string\":\n            return text;\n        default:\n            return text;\n    }\n}\n\n/**\n * Whether a text cell is a number under the f64 grammar (an i32 or f64 literal).\n * @param text - the cell text\n * @returns true when parseTextCell(text) returns a number\n */\nexport function isNumericText(text: string): boolean {\n    const dtype = inferTextDtype(text);\n    return dtype === \"i32\" || dtype === \"f64\";\n}\n\n/**\n * Whether a numeric cell text is the canonical spelling of its value (`String(value) === text`),\n * so that the sink's own re-formatting of the value would reproduce it.\n * @param text - the cell text\n * @param value - the parsed value\n * @returns true when the lexical form survives a widening to string\n */\nfunction isCanonicalNumberText(text: string, value: number): boolean {\n    return String(value) === text;\n}\n\n/**\n * The widening rank of a text dtype in the design section 5.1 order.\n * @param dtype - the text dtype\n * @returns 0 for bool, 1 for i32, 2 for f64, 3 for string\n */\nfunction textDtypeRank(dtype: TextDtype): number {\n    switch (dtype) {\n        case \"bool\":\n            return 0;\n        case \"i32\":\n            return 1;\n        case \"f64\":\n            return 2;\n        case \"string\":\n            return 3;\n        default: {\n            const name: string = dtype;\n            throw new GraphFormatError(\"E_UNSUPPORTED\", `unknown text dtype ${name}`, { dtype: name });\n        }\n    }\n}\n\n/**\n * The inferred-column writer of the untyped text formats (CSV, DOT, Pajek; design section 5.1):\n * parses every cell by the fixed grammar and pushes the scalar, and keeps the COLUMN's dtype the\n * one the grammar implies rather than the one the values happen to imply:\n *\n * - a column whose cells are all `2.0`-style f64 text is widened to f64 through the sink's\n *   widening call even though every value is integral (the values alone would infer i32);\n * - a numeric cell whose text is not the canonical spelling of its value (`1e5`, `-0`, `1.0`) is\n *   remembered, and when a later cell widens the column to string the original texts are written\n *   back, so the lexical form is never rewritten by the widening.\n *\n * A sink without the optional widening call keeps the value-inferred dtype and the writer records\n * one W_WIDENING_UNSUPPORTED warning per column.\n * Consumed by the per-format importers under src/formats.\n * @public\n */\nexport class TextCellWriter {\n    /** The column name in the sink. */\n    readonly name: string;\n\n    private readonly domain: \"node\" | \"edge\";\n\n    private readonly sink: GraphSink;\n\n    private readonly report: ImportReportBuilder;\n\n    private handle: ColumnHandle = INVALID_INDEX as ColumnHandle;\n\n    /** The widest text dtype seen (the column's dtype under the 5.1 grammar). */\n    private textDtype: TextDtype | null = null;\n\n    /** The widest dtype the pushed values imply (what the sink inferred on its own). */\n    private valueDtype: TextDtype | null = null;\n\n    private readonly keptRows: number[] = [];\n\n    private readonly keptTexts: string[] = [];\n\n    /**\n     * Create a writer; the column is declared by the sink's inference on the first write.\n     * @param name - the column name\n     * @param domain - node or edge\n     * @param sink - the sink\n     * @param report - the report the widening warning is recorded in\n     */\n    constructor(name: string, domain: \"node\" | \"edge\", sink: GraphSink, report: ImportReportBuilder) {\n        this.name = name;\n        this.domain = domain;\n        this.sink = sink;\n        this.report = report;\n    }\n\n    /**\n     * The column handle once the first cell was written.\n     * @returns the handle, or INVALID_INDEX before the first write\n     */\n    get column(): ColumnHandle {\n        return this.handle;\n    }\n\n    /**\n     * Write one cell text.\n     * @param row - the node or edge index\n     * @param text - the cell text, exactly as read\n     */\n    write(row: number, text: string): void {\n        const kind = inferTextDtype(text);\n        const value = parseTextCell(text);\n        const wasString = this.textDtype === \"string\";\n        const textDtype =\n            this.textDtype === null || textDtypeRank(kind) > textDtypeRank(this.textDtype) ? kind : this.textDtype;\n        // a string column keeps every cell's lexical form; a numeric text is never re-spelled. The\n        // sink may refuse the value (a declared column of another dtype): the state advances only\n        // once the cell is written, so a refused cell leaves the writer as it was.\n        this.set(row, textDtype === \"string\" ? text : value);\n        this.textDtype = textDtype;\n        const valueKind = kindOfValue(value);\n        if (this.valueDtype === null || textDtypeRank(valueKind) > textDtypeRank(this.valueDtype)) {\n            this.valueDtype = valueKind;\n        }\n        if (this.textDtype === \"string\") {\n            if (!wasString && this.keptRows.length > 0) {\n                // the sink just widened the column to string from the values; restore the texts\n                // whose lexical form the values did not carry\n                for (let i = 0; i < this.keptRows.length; i++) {\n                    this.set(this.keptRows[i], this.keptTexts[i]);\n                }\n                this.keptRows.length = 0;\n                this.keptTexts.length = 0;\n            }\n            return;\n        }\n        if (typeof value === \"number\" && !isCanonicalNumberText(text, value)) {\n            this.keptRows.push(row);\n            this.keptTexts.push(text);\n        }\n        if (this.textDtype === \"f64\" && this.valueDtype !== \"f64\") {\n            this.widen(\"f64\");\n        }\n    }\n\n    /**\n     * Write a value (the parsed scalar, or a text into a widened column).\n     * @param row - the row\n     * @param value - the value\n     */\n    private set(row: number, value: unknown): void {\n        if (this.handle === INVALID_INDEX) {\n            if (this.domain === \"node\") {\n                this.sink.setNodeValue(this.name, row, value);\n                this.handle = this.sink.nodeColumn(this.name);\n            } else {\n                this.sink.setEdgeValue(this.name, row, value);\n                this.handle = this.sink.edgeColumn(this.name);\n            }\n            return;\n        }\n        if (this.domain === \"node\") {\n            this.sink.setNodeValue(this.handle, row, value);\n        } else {\n            this.sink.setEdgeValue(this.handle, row, value);\n        }\n    }\n\n    /**\n     * Widen the column to the dtype the text grammar implies, through the sink's optional call.\n     * @param dtype - the dtype\n     */\n    private widen(dtype: \"f64\"): void {\n        const { sink } = this;\n        const supported =\n            this.domain === \"node\" ? sink.widenNodeColumn !== undefined : sink.widenEdgeColumn !== undefined;\n        if (!supported) {\n            this.report.warnOnce(\n                \"coercion\",\n                WIDENING_UNSUPPORTED_CODE,\n                `column \"${this.name}\" holds ${dtype} text but the sink cannot widen an inferred column; it keeps the value-inferred dtype`,\n                { element: this.name },\n            );\n            this.valueDtype = dtype;\n            return;\n        }\n        try {\n            if (this.domain === \"node\") {\n                sink.widenNodeColumn?.(this.handle, dtype);\n            } else {\n                sink.widenEdgeColumn?.(this.handle, dtype);\n            }\n            this.valueDtype = dtype;\n        } catch (err) {\n            if (!(err instanceof GraphFormatError) || err.code !== \"E_COLUMN_TYPE\") {\n                throw err;\n            }\n            // a caller's declared column of the name: its dtype stands\n            this.valueDtype = dtype;\n        }\n    }\n}\n\n/** Issue code: the sink has no widening call, so a text column keeps the dtype its values imply. */\nexport const WIDENING_UNSUPPORTED_CODE = \"W_WIDENING_UNSUPPORTED\";\n\n/**\n * The text dtype a parsed cell value implies on its own (what the sink's inference sees).\n * @param value - the parsed value\n * @returns the dtype\n */\nfunction kindOfValue(value: boolean | number | string): TextDtype {\n    if (typeof value === \"boolean\") {\n        return \"bool\";\n    }\n    if (typeof value === \"string\") {\n        return \"string\";\n    }\n    return Number.isInteger(value) && value >= -2147483648 && value <= 2147483647 && !Object.is(value, -0)\n        ? \"i32\"\n        : \"f64\";\n}\n"],"names":[],"mappings":";AAmBA,MAAM,WAAW;AACjB,MAAM,WAAW;AACjB,MAAM,UAAU;AAChB,MAAM,UAAU;AAcT,SAAS,eAAe,MAAyB;AACpD,MAAI,SAAS,UAAU,SAAS,SAAS;AACrC,WAAO;AAAA,EACX;AACA,MAAI,SAAS,KAAK,IAAI,GAAG;AACrB,UAAM,IAAI,OAAO,IAAI;AACrB,QAAI,KAAK,WAAW,KAAK,SAAS;AAC9B,aAAO;AAAA,IACX;AACA,WAAO,OAAO,SAAS,CAAC,IAAI,QAAQ;AAAA,EACxC;AACA,MAAI,SAAS,KAAK,IAAI,KAAK,OAAO,SAAS,OAAO,IAAI,CAAC,GAAG;AACtD,WAAO;AAAA,EACX;AACA,SAAO;AACX;AAQO,SAAS,cAAc,MAAyC;AACnE,UAAQ,eAAe,IAAI,GAAA;AAAA,IACvB,KAAK;AACD,aAAO,SAAS;AAAA,IACpB,KAAK;AAAA,IACL,KAAK;AACD,aAAO,OAAO,IAAI;AAAA,IACtB,KAAK;AACD,aAAO;AAAA,IACX;AACI,aAAO;AAAA,EAAA;AAEnB;AAOO,SAAS,cAAc,MAAuB;AACjD,QAAM,QAAQ,eAAe,IAAI;AACjC,SAAO,UAAU,SAAS,UAAU;AACxC;AASA,SAAS,sBAAsB,MAAc,OAAwB;AACjE,SAAO,OAAO,KAAK,MAAM;AAC7B;AAOA,SAAS,cAAc,OAA0B;AAC7C,UAAQ,OAAA;AAAA,IACJ,KAAK;AACD,aAAO;AAAA,IACX,KAAK;AACD,aAAO;AAAA,IACX,KAAK;AACD,aAAO;AAAA,IACX,KAAK;AACD,aAAO;AAAA,IACX,SAAS;AACL,YAAM,OAAe;AACrB,YAAM,IAAI,iBAAiB,iBAAiB,sBAAsB,IAAI,IAAI,EAAE,OAAO,MAAM;AAAA,IAC7F;AAAA,EAAA;AAER;AAkBO,MAAM,eAAe;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA6BxB,YAAY,MAAc,QAAyB,MAAiB,QAA6B;AAnBjG,SAAQ,SAAuB;AAG/B,SAAQ,YAA8B;AAGtC,SAAQ,aAA+B;AAEvC,SAAiB,WAAqB,CAAA;AAEtC,SAAiB,YAAsB,CAAA;AAUnC,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,OAAO;AACZ,SAAK,SAAS;AAAA,EAClB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,SAAuB;AACvB,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,KAAa,MAAoB;AACnC,UAAM,OAAO,eAAe,IAAI;AAChC,UAAM,QAAQ,cAAc,IAAI;AAChC,UAAM,YAAY,KAAK,cAAc;AACrC,UAAM,YACF,KAAK,cAAc,QAAQ,cAAc,IAAI,IAAI,cAAc,KAAK,SAAS,IAAI,OAAO,KAAK;AAIjG,SAAK,IAAI,KAAK,cAAc,WAAW,OAAO,KAAK;AACnD,SAAK,YAAY;AACjB,UAAM,YAAY,YAAY,KAAK;AACnC,QAAI,KAAK,eAAe,QAAQ,cAAc,SAAS,IAAI,cAAc,KAAK,UAAU,GAAG;AACvF,WAAK,aAAa;AAAA,IACtB;AACA,QAAI,KAAK,cAAc,UAAU;AAC7B,UAAI,CAAC,aAAa,KAAK,SAAS,SAAS,GAAG;AAGxC,iBAAS,IAAI,GAAG,IAAI,KAAK,SAAS,QAAQ,KAAK;AAC3C,eAAK,IAAI,KAAK,SAAS,CAAC,GAAG,KAAK,UAAU,CAAC,CAAC;AAAA,QAChD;AACA,aAAK,SAAS,SAAS;AACvB,aAAK,UAAU,SAAS;AAAA,MAC5B;AACA;AAAA,IACJ;AACA,QAAI,OAAO,UAAU,YAAY,CAAC,sBAAsB,MAAM,KAAK,GAAG;AAClE,WAAK,SAAS,KAAK,GAAG;AACtB,WAAK,UAAU,KAAK,IAAI;AAAA,IAC5B;AACA,QAAI,KAAK,cAAc,SAAS,KAAK,eAAe,OAAO;AACvD,WAAK,MAAM,KAAK;AAAA,IACpB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOQ,IAAI,KAAa,OAAsB;AAC3C,QAAI,KAAK,WAAW,eAAe;AAC/B,UAAI,KAAK,WAAW,QAAQ;AACxB,aAAK,KAAK,aAAa,KAAK,MAAM,KAAK,KAAK;AAC5C,aAAK,SAAS,KAAK,KAAK,WAAW,KAAK,IAAI;AAAA,MAChD,OAAO;AACH,aAAK,KAAK,aAAa,KAAK,MAAM,KAAK,KAAK;AAC5C,aAAK,SAAS,KAAK,KAAK,WAAW,KAAK,IAAI;AAAA,MAChD;AACA;AAAA,IACJ;AACA,QAAI,KAAK,WAAW,QAAQ;AACxB,WAAK,KAAK,aAAa,KAAK,QAAQ,KAAK,KAAK;AAAA,IAClD,OAAO;AACH,WAAK,KAAK,aAAa,KAAK,QAAQ,KAAK,KAAK;AAAA,IAClD;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,MAAM,OAAoB;AAC9B,UAAM,EAAE,SAAS;AACjB,UAAM,YACF,KAAK,WAAW,SAAS,KAAK,oBAAoB,SAAY,KAAK,oBAAoB;AAC3F,QAAI,CAAC,WAAW;AACZ,WAAK,OAAO;AAAA,QACR;AAAA,QACA;AAAA,QACA,WAAW,KAAK,IAAI,WAAW,KAAK;AAAA,QACpC,EAAE,SAAS,KAAK,KAAA;AAAA,MAAK;AAEzB,WAAK,aAAa;AAClB;AAAA,IACJ;AACA,QAAI;AACA,UAAI,KAAK,WAAW,QAAQ;AACxB,aAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,MAC7C,OAAO;AACH,aAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,MAC7C;AACA,WAAK,aAAa;AAAA,IACtB,SAAS,KAAK;AACV,UAAI,EAAE,eAAe,qBAAqB,IAAI,SAAS,iBAAiB;AACpE,cAAM;AAAA,MACV;AAEA,WAAK,aAAa;AAAA,IACtB;AAAA,EACJ;AACJ;AAGO,MAAM,4BAA4B;AAOzC,SAAS,YAAY,OAA6C;AAC9D,MAAI,OAAO,UAAU,WAAW;AAC5B,WAAO;AAAA,EACX;AACA,MAAI,OAAO,UAAU,UAAU;AAC3B,WAAO;AAAA,EACX;AACA,SAAO,OAAO,UAAU,KAAK,KAAK,SAAS,eAAe,SAAS,cAAc,CAAC,OAAO,GAAG,OAAO,EAAE,IAC/F,QACA;AACV;"}
+{"version":3,"file":"text-Dr0Ifpag.js","sources":["../../src/common/text.ts"],"sourcesContent":["/**\n * The fixed lexical grammar of design section 5.1 for untyped text sources (CSV cells, GML\n * values, DOT strings, Pajek tokens), reproduced from the core's reference implementation so the\n * two agree (invariant I15). Text importers parse each cell with parseTextCell() and push the JS\n * value; the sink's per-column inference then widens `bool -> i32 -> f64 -> string` and never per\n * cell, so a column that saw `1` and then `\"01\"` becomes string for all rows.\n *\n * - `bool` is exactly `true` / `false` (case-sensitive);\n * - `i32` is `/^-?(0|[1-9][0-9]*)$/` within `[-2^31, 2^31)`; `0` / `1` are i32, never bool;\n * - `f64` is a decimal or exponent literal accepted by `Number()` that is not empty, not\n *   whitespace, not `Infinity` / `NaN`, not a hex / octal / binary form, has no leading zeros in\n *   its integer part, and whose value is finite (an integer literal outside i32 range is f64);\n * - everything else is `string`.\n */\n\nimport { type ColumnHandle, GraphFormatError, type GraphSink, INVALID_INDEX } from \"@graphty/graph-format\";\n\nimport { type ImportReportBuilder } from \"./report.js\";\n\nconst I32_TEXT = /^-?(0|[1-9][0-9]*)$/;\nconst F64_TEXT = /^[+-]?((0|[1-9][0-9]*)(\\.[0-9]*)?|\\.[0-9]+)([eE][+-]?[0-9]+)?$/;\nconst I32_MIN = -2147483648;\nconst I32_MAX = 2147483647;\n\n/**\n * The dtype a text cell parses as under the design section 5.1 grammar.\n * Consumed by the per-format importers and exporters under src/formats.\n * @public\n */\nexport type TextDtype = \"bool\" | \"i32\" | \"f64\" | \"string\";\n\n/**\n * Classify one text cell by the fixed grammar.\n * @param text - the cell text, exactly as read (no trimming)\n * @returns bool, i32, f64 or string\n */\nexport function inferTextDtype(text: string): TextDtype {\n    if (text === \"true\" || text === \"false\") {\n        return \"bool\";\n    }\n    if (I32_TEXT.test(text)) {\n        const n = Number(text);\n        if (n >= I32_MIN && n <= I32_MAX) {\n            return \"i32\";\n        }\n        return Number.isFinite(n) ? \"f64\" : \"string\";\n    }\n    if (F64_TEXT.test(text) && Number.isFinite(Number(text))) {\n        return \"f64\";\n    }\n    return \"string\";\n}\n\n/**\n * Parse one text cell into the JS value its grammar class implies: a boolean for bool, a number for\n * i32 / f64, the text itself for string. The sink infers the column dtype from the value.\n * @param text - the cell text, exactly as read\n * @returns the value\n */\nexport function parseTextCell(text: string): boolean | number | string {\n    switch (inferTextDtype(text)) {\n        case \"bool\":\n            return text === \"true\";\n        case \"i32\":\n        case \"f64\":\n            return Number(text);\n        case \"string\":\n            return text;\n        default:\n            return text;\n    }\n}\n\n/**\n * Whether a text cell is a number under the f64 grammar (an i32 or f64 literal).\n * @param text - the cell text\n * @returns true when parseTextCell(text) returns a number\n */\nexport function isNumericText(text: string): boolean {\n    const dtype = inferTextDtype(text);\n    return dtype === \"i32\" || dtype === \"f64\";\n}\n\n/**\n * Whether a numeric cell text is the canonical spelling of its value (`String(value) === text`),\n * so that the sink's own re-formatting of the value would reproduce it.\n * @param text - the cell text\n * @param value - the parsed value\n * @returns true when the lexical form survives a widening to string\n */\nfunction isCanonicalNumberText(text: string, value: number): boolean {\n    return String(value) === text;\n}\n\n/**\n * The widening rank of a text dtype in the design section 5.1 order.\n * @param dtype - the text dtype\n * @returns 0 for bool, 1 for i32, 2 for f64, 3 for string\n */\nfunction textDtypeRank(dtype: TextDtype): number {\n    switch (dtype) {\n        case \"bool\":\n            return 0;\n        case \"i32\":\n            return 1;\n        case \"f64\":\n            return 2;\n        case \"string\":\n            return 3;\n        default: {\n            const name: string = dtype;\n            throw new GraphFormatError(\"E_UNSUPPORTED\", `unknown text dtype ${name}`, { dtype: name });\n        }\n    }\n}\n\n/**\n * The inferred-column writer of the untyped text formats (CSV, DOT, Pajek; design section 5.1):\n * parses every cell by the fixed grammar and pushes the scalar, and keeps the COLUMN's dtype the\n * one the grammar implies rather than the one the values happen to imply:\n *\n * - a column whose cells are all `2.0`-style f64 text is widened to f64 through the sink's\n *   widening call even though every value is integral (the values alone would infer i32);\n * - a numeric cell whose text is not the canonical spelling of its value (`1e5`, `-0`, `1.0`) is\n *   remembered, and when a later cell widens the column to string the original texts are written\n *   back, so the lexical form is never rewritten by the widening.\n *\n * A sink without the optional widening call keeps the value-inferred dtype and the writer records\n * one W_WIDENING_UNSUPPORTED warning per column.\n * Consumed by the per-format importers under src/formats.\n * @public\n */\nexport class TextCellWriter {\n    /** The column name in the sink. */\n    readonly name: string;\n\n    private readonly domain: \"node\" | \"edge\";\n\n    private readonly sink: GraphSink;\n\n    private readonly report: ImportReportBuilder;\n\n    private handle: ColumnHandle = INVALID_INDEX as ColumnHandle;\n\n    /** The widest text dtype seen (the column's dtype under the 5.1 grammar). */\n    private textDtype: TextDtype | null = null;\n\n    /** The widest dtype the pushed values imply (what the sink inferred on its own). */\n    private valueDtype: TextDtype | null = null;\n\n    private readonly keptRows: number[] = [];\n\n    private readonly keptTexts: string[] = [];\n\n    /**\n     * Create a writer; the column is declared by the sink's inference on the first write.\n     * @param name - the column name\n     * @param domain - node or edge\n     * @param sink - the sink\n     * @param report - the report the widening warning is recorded in\n     */\n    constructor(name: string, domain: \"node\" | \"edge\", sink: GraphSink, report: ImportReportBuilder) {\n        this.name = name;\n        this.domain = domain;\n        this.sink = sink;\n        this.report = report;\n    }\n\n    /**\n     * The column handle once the first cell was written.\n     * @returns the handle, or INVALID_INDEX before the first write\n     */\n    get column(): ColumnHandle {\n        return this.handle;\n    }\n\n    /**\n     * Write one cell text.\n     * @param row - the node or edge index\n     * @param text - the cell text, exactly as read\n     */\n    write(row: number, text: string): void {\n        const kind = inferTextDtype(text);\n        const value = parseTextCell(text);\n        const wasString = this.textDtype === \"string\";\n        const textDtype =\n            this.textDtype === null || textDtypeRank(kind) > textDtypeRank(this.textDtype) ? kind : this.textDtype;\n        // a string column keeps every cell's lexical form; a numeric text is never re-spelled. The\n        // sink may refuse the value (a declared column of another dtype): the state advances only\n        // once the cell is written, so a refused cell leaves the writer as it was.\n        this.set(row, textDtype === \"string\" ? text : value);\n        this.textDtype = textDtype;\n        const valueKind = kindOfValue(value);\n        if (this.valueDtype === null || textDtypeRank(valueKind) > textDtypeRank(this.valueDtype)) {\n            this.valueDtype = valueKind;\n        }\n        if (this.textDtype === \"string\") {\n            if (!wasString && this.keptRows.length > 0) {\n                // the sink just widened the column to string from the values; restore the texts\n                // whose lexical form the values did not carry\n                for (let i = 0; i < this.keptRows.length; i++) {\n                    this.set(this.keptRows[i], this.keptTexts[i]);\n                }\n                this.keptRows.length = 0;\n                this.keptTexts.length = 0;\n            }\n            return;\n        }\n        if (typeof value === \"number\" && !isCanonicalNumberText(text, value)) {\n            this.keptRows.push(row);\n            this.keptTexts.push(text);\n        }\n        if (this.textDtype === \"f64\" && this.valueDtype !== \"f64\") {\n            this.widen(\"f64\");\n        }\n    }\n\n    /**\n     * Write a value (the parsed scalar, or a text into a widened column).\n     * @param row - the row\n     * @param value - the value\n     */\n    private set(row: number, value: unknown): void {\n        if (this.handle === INVALID_INDEX) {\n            if (this.domain === \"node\") {\n                this.sink.setNodeValue(this.name, row, value);\n                this.handle = this.sink.nodeColumn(this.name);\n            } else {\n                this.sink.setEdgeValue(this.name, row, value);\n                this.handle = this.sink.edgeColumn(this.name);\n            }\n            return;\n        }\n        if (this.domain === \"node\") {\n            this.sink.setNodeValue(this.handle, row, value);\n        } else {\n            this.sink.setEdgeValue(this.handle, row, value);\n        }\n    }\n\n    /**\n     * Widen the column to the dtype the text grammar implies, through the sink's optional call.\n     * @param dtype - the dtype\n     */\n    private widen(dtype: \"f64\"): void {\n        const { sink } = this;\n        const supported =\n            this.domain === \"node\" ? sink.widenNodeColumn !== undefined : sink.widenEdgeColumn !== undefined;\n        if (!supported) {\n            this.report.warnOnce(\n                \"coercion\",\n                WIDENING_UNSUPPORTED_CODE,\n                `column \"${this.name}\" holds ${dtype} text but the sink cannot widen an inferred column; it keeps the value-inferred dtype`,\n                { element: this.name },\n            );\n            this.valueDtype = dtype;\n            return;\n        }\n        try {\n            if (this.domain === \"node\") {\n                sink.widenNodeColumn?.(this.handle, dtype);\n            } else {\n                sink.widenEdgeColumn?.(this.handle, dtype);\n            }\n            this.valueDtype = dtype;\n        } catch (err) {\n            if (!(err instanceof GraphFormatError) || err.code !== \"E_COLUMN_TYPE\") {\n                throw err;\n            }\n            // a caller's declared column of the name: its dtype stands\n            this.valueDtype = dtype;\n        }\n    }\n}\n\n/** Issue code: the sink has no widening call, so a text column keeps the dtype its values imply. */\nexport const WIDENING_UNSUPPORTED_CODE = \"W_WIDENING_UNSUPPORTED\";\n\n/**\n * The text dtype a parsed cell value implies on its own (what the sink's inference sees). It must\n * be graph-format's inference rule exactly: -0 is an integer there, so a column holding only\n * `-0.0` is inferred i32 and must be widened to f64, or the sign is lost.\n * @param value - the parsed value\n * @returns the dtype\n */\nfunction kindOfValue(value: boolean | number | string): TextDtype {\n    if (typeof value === \"boolean\") {\n        return \"bool\";\n    }\n    if (typeof value === \"string\") {\n        return \"string\";\n    }\n    return Number.isInteger(value) && value >= -2147483648 && value <= 2147483647 ? \"i32\" : \"f64\";\n}\n"],"names":[],"mappings":";AAmBA,MAAM,WAAW;AACjB,MAAM,WAAW;AACjB,MAAM,UAAU;AAChB,MAAM,UAAU;AAcT,SAAS,eAAe,MAAyB;AACpD,MAAI,SAAS,UAAU,SAAS,SAAS;AACrC,WAAO;AAAA,EACX;AACA,MAAI,SAAS,KAAK,IAAI,GAAG;AACrB,UAAM,IAAI,OAAO,IAAI;AACrB,QAAI,KAAK,WAAW,KAAK,SAAS;AAC9B,aAAO;AAAA,IACX;AACA,WAAO,OAAO,SAAS,CAAC,IAAI,QAAQ;AAAA,EACxC;AACA,MAAI,SAAS,KAAK,IAAI,KAAK,OAAO,SAAS,OAAO,IAAI,CAAC,GAAG;AACtD,WAAO;AAAA,EACX;AACA,SAAO;AACX;AAQO,SAAS,cAAc,MAAyC;AACnE,UAAQ,eAAe,IAAI,GAAA;AAAA,IACvB,KAAK;AACD,aAAO,SAAS;AAAA,IACpB,KAAK;AAAA,IACL,KAAK;AACD,aAAO,OAAO,IAAI;AAAA,IACtB,KAAK;AACD,aAAO;AAAA,IACX;AACI,aAAO;AAAA,EAAA;AAEnB;AAOO,SAAS,cAAc,MAAuB;AACjD,QAAM,QAAQ,eAAe,IAAI;AACjC,SAAO,UAAU,SAAS,UAAU;AACxC;AASA,SAAS,sBAAsB,MAAc,OAAwB;AACjE,SAAO,OAAO,KAAK,MAAM;AAC7B;AAOA,SAAS,cAAc,OAA0B;AAC7C,UAAQ,OAAA;AAAA,IACJ,KAAK;AACD,aAAO;AAAA,IACX,KAAK;AACD,aAAO;AAAA,IACX,KAAK;AACD,aAAO;AAAA,IACX,KAAK;AACD,aAAO;AAAA,IACX,SAAS;AACL,YAAM,OAAe;AACrB,YAAM,IAAI,iBAAiB,iBAAiB,sBAAsB,IAAI,IAAI,EAAE,OAAO,MAAM;AAAA,IAC7F;AAAA,EAAA;AAER;AAkBO,MAAM,eAAe;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA6BxB,YAAY,MAAc,QAAyB,MAAiB,QAA6B;AAnBjG,SAAQ,SAAuB;AAG/B,SAAQ,YAA8B;AAGtC,SAAQ,aAA+B;AAEvC,SAAiB,WAAqB,CAAA;AAEtC,SAAiB,YAAsB,CAAA;AAUnC,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,OAAO;AACZ,SAAK,SAAS;AAAA,EAClB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,SAAuB;AACvB,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,KAAa,MAAoB;AACnC,UAAM,OAAO,eAAe,IAAI;AAChC,UAAM,QAAQ,cAAc,IAAI;AAChC,UAAM,YAAY,KAAK,cAAc;AACrC,UAAM,YACF,KAAK,cAAc,QAAQ,cAAc,IAAI,IAAI,cAAc,KAAK,SAAS,IAAI,OAAO,KAAK;AAIjG,SAAK,IAAI,KAAK,cAAc,WAAW,OAAO,KAAK;AACnD,SAAK,YAAY;AACjB,UAAM,YAAY,YAAY,KAAK;AACnC,QAAI,KAAK,eAAe,QAAQ,cAAc,SAAS,IAAI,cAAc,KAAK,UAAU,GAAG;AACvF,WAAK,aAAa;AAAA,IACtB;AACA,QAAI,KAAK,cAAc,UAAU;AAC7B,UAAI,CAAC,aAAa,KAAK,SAAS,SAAS,GAAG;AAGxC,iBAAS,IAAI,GAAG,IAAI,KAAK,SAAS,QAAQ,KAAK;AAC3C,eAAK,IAAI,KAAK,SAAS,CAAC,GAAG,KAAK,UAAU,CAAC,CAAC;AAAA,QAChD;AACA,aAAK,SAAS,SAAS;AACvB,aAAK,UAAU,SAAS;AAAA,MAC5B;AACA;AAAA,IACJ;AACA,QAAI,OAAO,UAAU,YAAY,CAAC,sBAAsB,MAAM,KAAK,GAAG;AAClE,WAAK,SAAS,KAAK,GAAG;AACtB,WAAK,UAAU,KAAK,IAAI;AAAA,IAC5B;AACA,QAAI,KAAK,cAAc,SAAS,KAAK,eAAe,OAAO;AACvD,WAAK,MAAM,KAAK;AAAA,IACpB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOQ,IAAI,KAAa,OAAsB;AAC3C,QAAI,KAAK,WAAW,eAAe;AAC/B,UAAI,KAAK,WAAW,QAAQ;AACxB,aAAK,KAAK,aAAa,KAAK,MAAM,KAAK,KAAK;AAC5C,aAAK,SAAS,KAAK,KAAK,WAAW,KAAK,IAAI;AAAA,MAChD,OAAO;AACH,aAAK,KAAK,aAAa,KAAK,MAAM,KAAK,KAAK;AAC5C,aAAK,SAAS,KAAK,KAAK,WAAW,KAAK,IAAI;AAAA,MAChD;AACA;AAAA,IACJ;AACA,QAAI,KAAK,WAAW,QAAQ;AACxB,WAAK,KAAK,aAAa,KAAK,QAAQ,KAAK,KAAK;AAAA,IAClD,OAAO;AACH,WAAK,KAAK,aAAa,KAAK,QAAQ,KAAK,KAAK;AAAA,IAClD;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,MAAM,OAAoB;AAC9B,UAAM,EAAE,SAAS;AACjB,UAAM,YACF,KAAK,WAAW,SAAS,KAAK,oBAAoB,SAAY,KAAK,oBAAoB;AAC3F,QAAI,CAAC,WAAW;AACZ,WAAK,OAAO;AAAA,QACR;AAAA,QACA;AAAA,QACA,WAAW,KAAK,IAAI,WAAW,KAAK;AAAA,QACpC,EAAE,SAAS,KAAK,KAAA;AAAA,MAAK;AAEzB,WAAK,aAAa;AAClB;AAAA,IACJ;AACA,QAAI;AACA,UAAI,KAAK,WAAW,QAAQ;AACxB,aAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,MAC7C,OAAO;AACH,aAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,MAC7C;AACA,WAAK,aAAa;AAAA,IACtB,SAAS,KAAK;AACV,UAAI,EAAE,eAAe,qBAAqB,IAAI,SAAS,iBAAiB;AACpE,cAAM;AAAA,MACV;AAEA,WAAK,aAAa;AAAA,IACtB;AAAA,EACJ;AACJ;AAGO,MAAM,4BAA4B;AASzC,SAAS,YAAY,OAA6C;AAC9D,MAAI,OAAO,UAAU,WAAW;AAC5B,WAAO;AAAA,EACX;AACA,MAAI,OAAO,UAAU,UAAU;AAC3B,WAAO;AAAA,EACX;AACA,SAAO,OAAO,UAAU,KAAK,KAAK,SAAS,eAAe,SAAS,aAAa,QAAQ;AAC5F;"}