@podlite/schema 0.0.38 → 0.0.39

package/lib/plugin-tables.js

@@ -9,13 +9,271 @@ Object.defineProperty(exports, "__esModule", { value: true });
  * and the remaining content is treated as the definition for the term.
  */
 const makeTransformer_1 = __importDefault(require("./helpers/makeTransformer"));
+const config_1 = __importDefault(require("./helpers/config"));
 function flattenDeep(arr) {
     return arr.reduce((acc, val) => (Array.isArray(val) ? acc.concat(flattenDeep(val)) : acc.concat(val)), []);
 }
+// ─── CSV source for =table (spec §1672) ────────────────────────────────────
+// Parse CSV content per RFC 4180: comma delimiter, `"`-quoted fields with
+// `""` as embedded-quote escape. Supports LF or CRLF line endings. Leading
+// indentation typical of =data block bodies is stripped from each line.
+function parseCsv(text) {
+    const rows = [];
+    let field = '';
+    let row = [];
+    let inQuote = false;
+    let i = 0;
+    while (i < text.length) {
+        const c = text[i];
+        if (inQuote) {
+            if (c === '"' && text[i + 1] === '"') {
+                field += '"';
+                i += 2;
+                continue;
+            }
+            if (c === '"') {
+                inQuote = false;
+                i++;
+                continue;
+            }
+            field += c;
+            i++;
+            continue;
+        }
+        if (c === '"' && field === '') {
+            inQuote = true;
+            i++;
+            continue;
+        }
+        if (c === ',') {
+            row.push(field);
+            field = '';
+            i++;
+            continue;
+        }
+        if (c === '\r') {
+            i++;
+            continue;
+        }
+        if (c === '\n') {
+            row.push(field);
+            rows.push(row);
+            row = [];
+            field = '';
+            i++;
+            continue;
+        }
+        field += c;
+        i++;
+    }
+    if (field !== '' || row.length > 0) {
+        row.push(field);
+        rows.push(row);
+    }
+    // Drop rows that are entirely blank (empty content after trim).
+    return rows.filter(r => !(r.length === 1 && r[0].trim() === ''));
+}
+// Parse a TSV (tab-separated values) blob. Unlike CSV, TSV has no quoting
+// mechanism — fields are split strictly on tabs and `"` is a literal
+// character. Tabs and newlines inside fields are not representable in TSV.
+function parseTsv(text) {
+    const lines = text.split(/\r?\n/);
+    const rows = lines.map(line => line.split('\t'));
+    return rows.filter(r => !(r.length === 1 && r[0].trim() === ''));
+}
+// Locate the first `=data` block with a matching `:key` attribute anywhere
+// in the document tree.
+function findDataBlockByKey(tree, key) {
+    let found = null;
+    const walk = node => {
+        if (found)
+            return;
+        if (Array.isArray(node)) {
+            node.forEach(walk);
+            return;
+        }
+        if (!node || typeof node !== 'object')
+            return;
+        if (node.type === 'block' && node.name === 'data') {
+            const attrs = (0, config_1.default)(node, {});
+            if (attrs.getFirstValue('key') === key) {
+                found = node;
+                return;
+            }
+        }
+        if (Array.isArray(node.content))
+            node.content.forEach(walk);
+    };
+    walk(tree);
+    return found;
+}
+function extractDataText(dataNode) {
+    if (!dataNode || !Array.isArray(dataNode.content))
+        return '';
+    const verbatim = dataNode.content.find(c => c && c.type === 'verbatim');
+    return verbatim && typeof verbatim.value === 'string' ? verbatim.value : '';
+}
+// Detect whether the =table body is a single-line source reference like
+// `data:key` or `file:path` (spec §1672). Returns the parsed reference or
+// null if the body is ordinary table content.
+function detectSourceReference(tableNode) {
+    const texts = [];
+    const walker = (0, makeTransformer_1.default)({
+        'row:text': r => {
+            texts.push(r.value);
+            return r;
+        },
+        'head:text': h => {
+            texts.push(h.value);
+            return h;
+        },
+    });
+    walker(tableNode, {});
+    const joined = texts.join('\n').trim();
+    const lines = joined
+        .split(/\r?\n/)
+        .map(l => l.trim())
+        .filter(l => l.length > 0);
+    if (lines.length !== 1)
+        return null;
+    const m = lines[0].match(/^(data|file):(\S+)$/);
+    if (!m)
+        return null;
+    return { scheme: m[1], target: m[2] };
+}
+function buildCellBlock(text) {
+    return {
+        name: 'cell',
+        type: 'block',
+        margin: '',
+        content: [{ type: 'text', value: text }],
+    };
+}
+function buildRowBlock(cells, isHeader) {
+    const block = {
+        name: 'row',
+        type: 'block',
+        margin: '',
+        content: cells,
+    };
+    if (isHeader) {
+        block.config = [{ name: 'header', value: true, type: 'boolean' }];
+    }
+    return block;
+}
+// Convert parsed CSV/TSV rows to plain `=row`/`=cell` blocks. The spec is
+// silent on how to mark a header row in `=table data:<key>` references, so
+// no row receives `:header` here. Authors who need a header row can use a
+// structured table with explicit `=begin row :header`, or a Markdown GFM
+// table with a separator line.
+function csvToTableContent(csvRows) {
+    return csvRows.map(row => {
+        const cells = row.map(v => buildCellBlock(v.trim()));
+        return buildRowBlock(cells, false);
+    });
+}
+// ─── Error recovery (design notes Rules 2-4) ───────────────────────────────
+// Rule 2 — table-level cell count validation. Pad short rows with empty
+// cells; truncate long rows. Emit a warning whenever a row is changed.
+// Expected count is taken from the `:header` row if present, otherwise from
+// the row with the maximum cell count.
+//
+// Skipped when any cell uses `:colspan` or `:rowspan`: a spanning cell
+// occupies multiple columns, so naive cell counting would misreport row
+// width and drop legitimate spanned cells.
+function normalizeCellCounts(tableNode, source = 'table') {
+    if (!tableNode || !Array.isArray(tableNode.content))
+        return tableNode;
+    const rows = tableNode.content.filter(c => c && c.type === 'block' && c.name === 'row');
+    if (rows.length === 0)
+        return tableNode;
+    const cellsOf = row => (Array.isArray(row.content) ? row.content.filter(c => c && c.name === 'cell') : []);
+    const isHeaderRow = row => Array.isArray(row.config) && row.config.some(a => a.name === 'header' && a.value === true);
+    const cellHasSpan = cell => Array.isArray(cell.config) && cell.config.some(a => a.name === 'colspan' || a.name === 'rowspan');
+    const rowHasSpan = row => cellsOf(row).some(cellHasSpan);
+    if (rows.some(rowHasSpan))
+        return tableNode;
+    const headerRow = rows.find(isHeaderRow);
+    const expected = headerRow ? cellsOf(headerRow).length : Math.max(...rows.map(r => cellsOf(r).length));
+    if (expected === 0)
+        return tableNode;
+    let mutated = false;
+    const newContent = tableNode.content.map(child => {
+        if (!child || child.type !== 'block' || child.name !== 'row')
+            return child;
+        const cells = cellsOf(child);
+        if (cells.length === expected)
+            return child;
+        if (cells.length < expected) {
+            const padding = [];
+            for (let i = cells.length; i < expected; i++)
+                padding.push(buildCellBlock(''));
+            console.warn(`[${source}] row has ${cells.length} cells, expected ${expected} — padded with ${padding.length} empty`);
+            mutated = true;
+            return { ...child, content: [...child.content, ...padding] };
+        }
+        // cells.length > expected → truncate
+        const dropped = cells.length - expected;
+        console.warn(`[${source}] row has ${cells.length} cells, expected ${expected} — truncated ${dropped}`);
+        mutated = true;
+        // Keep non-cell entries (e.g. blanklines) and the first `expected` cells
+        let keepCells = expected;
+        const trimmed = [];
+        for (const c of child.content || []) {
+            if (c && c.type === 'block' && c.name === 'cell') {
+                if (keepCells > 0) {
+                    trimmed.push(c);
+                    keepCells--;
+                }
+            }
+            else {
+                trimmed.push(c);
+            }
+        }
+        return { ...child, content: trimmed };
+    });
+    return mutated ? { ...tableNode, content: newContent } : tableNode;
+}
+// Rule 3 — mixed separator detection (text-mode only). Inspects each line
+// for visible separators (`|` / `+`) surrounded by whitespace; lines without
+// any visible separator fall back to whitespace separation. Warns when more
+// than one separator type is observed within a single table.
+function detectMixedSeparators(lines) {
+    const seen = new Set();
+    for (const line of lines) {
+        if (!line || typeof line !== 'string')
+            continue;
+        if (/\s\|\s/.test(line))
+            seen.add('pipe');
+        else if (/\s\+\s/.test(line))
+            seen.add('plus');
+        else if (line.trim().length > 0)
+            seen.add('whitespace');
+        if (seen.size > 1)
+            break;
+    }
+    if (seen.size > 1) {
+        console.warn(`[table] mixed separator types detected: ${Array.from(seen).join(', ')} — recommend a single style`);
+    }
+}
+// Rule 4 — replace =table with =code block. Used when a referenced =data
+// source has a non-CSV mime type: the spec mandates the source be rendered
+// as a code block so the user can still see the underlying content.
+function buildCodeFromDataBlock(tableNode, dataBlock) {
+    return {
+        type: 'block',
+        name: 'code',
+        margin: tableNode.margin || '',
+        content: Array.isArray(dataBlock.content) ? dataBlock.content : [],
+        config: Array.isArray(tableNode.config) ? tableNode.config : [],
+    };
+}
 /**
  *  Helpers section
  */
-// run cb in symbols pair
+// Bit-mask helper used by the positional column extractor for multi-line
+// rows. Compares two same-length digit strings character by character via
+// `cb`, returning the joined result.
 const strbin = (str1, str2, cb) => {
     let res = [];
     for (let i = 0; str1.length > i; i++) {
@@ -23,62 +281,109 @@ const strbin = (str1, str2, cb) => {
     }
     return res.join('');
 };
-// Create mask for extract columns
+// Build a unified column-position mask across a set of lines. Used as a
+// fallback when a row spans multiple lines (continuation lines that align
+// content by character position rather than by separator). Returns a binary
+// string where `0` runs mark column ranges and `1` runs mark gaps.
 const makeMask = (lines, separators) => {
-    // calculate template length
     const tmplLength = Math.max(...[...lines, ...separators].map(s => s.length));
-    // make bin mask for each string
     const masks = lines.map(str => {
-        /** make mask for each line
-                    '        The Shoveller | Eddie Stevens   | King Arthur\'s singing shovel',
-                    '0000000011111111111110001111111111111000001111111111111111111111111111' ]
-                    then  not(mask) ... then  & masks
-                   */
-        // enlarge string to tmplLength
-        let tstr = str + ' '.repeat(tmplLength - str.length);
-        let mask = [];
+        const tstr = str + ' '.repeat(tmplLength - str.length);
+        const mask = [];
         const re = /\s+[+|\s]\s/g;
         let match;
         while ((match = re.exec(tstr)) != null) {
             const tmpMask = '1'.repeat(match.index) + '0'.repeat(match[0].length);
             mask.push(tmpMask + '1'.repeat(tmplLength - tmpMask.length));
         }
-        return mask.reduce((a, b) => {
-            return strbin(a, b, (i1, i2) => i1 & i2);
-        }, '1'.repeat(tmplLength));
+        return mask.reduce((a, b) => strbin(a, b, (i1, i2) => i1 & i2), '1'.repeat(tmplLength));
     });
-    // make result mask
     const inverted = masks.map(m => strbin(m, '', i1 => (i1 == 0 ? 1 : 0)));
-    const columnTemplate = inverted.reduce((a, b) => {
-        return strbin(a, b, (i1, i2) => i1 & i2);
-    }, '1'.repeat(tmplLength));
-    return columnTemplate;
+    return inverted.reduce((a, b) => strbin(a, b, (i1, i2) => i1 & i2), '1'.repeat(tmplLength));
 };
+// Apply a column-position template to a multi-line text block, extracting
+// per-column substrings and aggregating across lines (continuation lines
+// append to the cell at the same column position).
 const extractColumnsByTemplate = (text, template) => {
-    const lines = flattenDeep(text
-        .split(/\n/) // split each row by eol
-        .filter(str => str.length > 0));
+    const lines = flattenDeep(text.split(/\n/).filter(s => s.length > 0));
     const cols = lines.map(line => {
         const re = /((1+|0+))/g;
-        let columns = [];
+        const columns = [];
         let match;
         while ((match = re.exec(template)) != null) {
-            if (match[0][0] == 1)
+            if (match[0][0] == '1')
                 continue;
-            const s = line.substring(match.index, match.index + match[0].length);
-            columns.push(s);
+            columns.push(line.substring(match.index, match.index + match[0].length));
         }
         return columns;
     });
-    let result = [];
-    result = cols.reduce((a, b) => {
+    const result = [];
+    cols.reduce((a, b) => {
         for (let i = 0; i < b.length; i++) {
             a[i] = (a[i] === undefined ? '' : a[i]) + ' ' + b[i];
         }
         return a;
-    }, []);
+    }, result);
     return result;
 };
+const detectLineSeparator = (line) => {
+    if (/(?:^|\s)\|(?:\s|$)/.test(line))
+        return 'pipe';
+    if (/(?:^|\s)\+(?:\s|$)/.test(line))
+        return 'plus';
+    return 'whitespace';
+};
+const trimEdgeEmpty = (cells) => {
+    // A leading `|` produces an empty cell at index 0; a trailing `|` does the
+    // same at the end. Drop those edge artifacts. Empty cells in the middle of
+    // the row are preserved.
+    let start = 0;
+    let end = cells.length;
+    if (cells[start] === '')
+        start++;
+    if (end > start && cells[end - 1] === '')
+        end--;
+    return cells.slice(start, end);
+};
+const splitLineByPipe = (line) => trimEdgeEmpty(line.split(/\s*\|\s*/).map(c => c.trim()));
+const splitLineByPlus = (line) => trimEdgeEmpty(line.split(/\s*\+\s*/).map(c => c.trim()));
+const splitLineByWhitespace = (line) => line
+    .trim()
+    .split(/\s{2,}/)
+    .filter(c => c !== '');
+const splitLineCells = (line) => {
+    const trimmed = line.trim();
+    if (trimmed === '')
+        return [];
+    const kind = detectLineSeparator(line);
+    if (kind === 'pipe')
+        return splitLineByPipe(line);
+    if (kind === 'plus')
+        return splitLineByPlus(line);
+    return splitLineByWhitespace(line);
+};
+// Convert a row's raw text (which may span multiple lines) to cell values.
+// Multi-line rows: take the separator kind from the first non-blank line and
+// apply it to every line of the row, then aggregate column-wise so that a
+// continuation line (e.g. wrapped cell content) appends to the cell from the
+// previous line in the same column.
+const rowToCells = (rowValue) => {
+    const lines = rowValue.split(/\r?\n/).filter(l => l.trim() !== '');
+    if (lines.length === 0)
+        return [];
+    if (lines.length === 1)
+        return splitLineCells(lines[0]);
+    const kind = detectLineSeparator(lines[0]);
+    const splitFn = kind === 'pipe' ? splitLineByPipe : kind === 'plus' ? splitLineByPlus : splitLineByWhitespace;
+    const lineCells = lines.map(splitFn);
+    const maxCols = Math.max(...lineCells.map(c => c.length));
+    const merged = [];
+    for (let i = 0; i < maxCols; i++) {
+        const parts = lineCells.map(line => line[i] ?? '').filter(p => p !== '');
+        merged.push(parts.join(' '));
+    }
+    return merged;
+};
 /**
  *  Main transforms
  */
@@ -102,14 +407,43 @@ const isStructured = tableNode => Array.isArray(tableNode.content) &&
 exports.default = () => tree => {
     const transformer = (0, makeTransformer_1.default)({
         table: node => {
-            // structured mode: transform row children (wrap implicit cells)
+            // CSV/data source reference (spec §1672):
+            //   =table data:<key>      →  resolve =data block with :key<key>
+            //   =table file:<path>     →  defer to host reader (not implemented here)
+            const ref = detectSourceReference(node);
+            if (ref && ref.scheme === 'data') {
+                const dataBlock = findDataBlockByKey(tree, ref.target);
+                if (!dataBlock) {
+                    // Rule 4: source not found → empty table (still a =table block)
+                    console.warn(`[table] no =data block found for data:${ref.target} — rendered as empty`);
+                    return { ...node, content: [] };
+                }
+                const mimeType = (0, config_1.default)(dataBlock, {}).getFirstValue('mime-type');
+                const isCsv = mimeType === 'text/csv';
+                const isTsv = mimeType === 'text/tab-separated-values';
+                if (isCsv || isTsv) {
+                    const text = extractDataText(dataBlock);
+                    const rows = isCsv ? parseCsv(text) : parseTsv(text);
+                    if (rows.length === 0) {
+                        console.warn(`[table] ${isCsv ? 'CSV' : 'TSV'} parse produced no rows for data:${ref.target} — rendered as empty`);
+                        return { ...node, content: [] };
+                    }
+                    const filledNode = { ...node, content: csvToTableContent(rows) };
+                    return normalizeCellCounts(filledNode, `table data:${ref.target}`);
+                }
+                // Rule 4: source not tabular → render as code block so content remains visible
+                console.warn(`[table] =data :key<${ref.target}> has non-tabular mime-type ${mimeType || '(none)'} — rendered as =code`);
+                return buildCodeFromDataBlock(node, dataBlock);
+            }
+            // structured mode: transform row children (wrap implicit cells), then
+            // apply Rule 2 cell count normalization.
             if (isStructured(node)) {
                 const transformedContent = (node.content || []).map(c => {
                     if (c && c.name === 'row')
                         return wrapImplicitCells(c);
                     return c;
                 });
-                return { ...node, content: transformedContent };
+                return normalizeCellCounts({ ...node, content: transformedContent }, 'table');
             }
             let rows = [];
             const collectValues = row => {
@@ -134,6 +468,8 @@ exports.default = () => tree => {
             // split each row into lines
             const lines = flattenDeep(rows.map(splitToLines));
             const separators = flattenDeep(seps.map(splitToLines));
+            // Rule 3: warn on mixed separator types within a single table
+            detectMixedSeparators(lines);
             // collect text rows
             let textRows = [];
             (0, makeTransformer_1.default)({
@@ -141,33 +477,48 @@ exports.default = () => tree => {
                     textRows.push(row.value);
                 },
             })(node);
-            const columnTemplate = makeMask(lines, separators);
             const makeBlock = (name, content, extra = {}) => {
                 return { ...extra, name, type: 'block', content: Array.isArray(content) ? content : [content] };
             };
             const makeRow = cells => makeBlock('row', cells);
             const makeHeaderRow = cells => makeBlock('row', cells, { config: [{ name: 'header', value: true, type: 'boolean' }] });
             const makeCell = text => makeBlock('cell', { type: 'text', value: text });
-            // make columns
+            // Routing: per-line separator detection (Rule 1) is used only when a
+            // line with a visible separator (`|` or `+`) coexists with one that
+            // has only whitespace separation — the Scenario 2 case where the
+            // legacy positional mask collapses columns. Tables with a uniform
+            // separator, or with both visible kinds (pipe + plus) handled by the
+            // shared mask, fall back to the legacy positional template, which
+            // preserves continuation-line alignment in multi-line rows and keeps
+            // byte-for-byte AST/HTML output stable.
+            const columnTemplate = makeMask(lines, separators);
+            const seenSeparatorKinds = new Set(lines.map(detectLineSeparator));
+            const hasVisible = seenSeparatorKinds.has('pipe') || seenSeparatorKinds.has('plus');
+            const hasWhitespace = seenSeparatorKinds.has('whitespace');
+            const useMixedSplitting = hasVisible && hasWhitespace;
+            const splitToCells = (rowValue) => {
+                if (useMixedSplitting) {
+                    const rowLines = rowValue.split(/\r?\n/).filter(l => l.trim() !== '');
+                    if (rowLines.length <= 1)
+                        return rowToCells(rowValue);
+                }
+                return extractColumnsByTemplate(rowValue, columnTemplate);
+            };
             const res = (0, makeTransformer_1.default)({
                 'row:text': row => {
                     if (textRows.length == 1) {
-                        // split each text row into lines
+                        // No separator blocks: each line of the only text row becomes its own row
                         const textRowsLines = flattenDeep([row.value].map(splitToLines));
-                        return textRowsLines.map(rowValue => {
-                            const cols = extractColumnsByTemplate(rowValue, columnTemplate);
-                            return makeRow(cols.map(makeCell));
-                        });
+                        if (useMixedSplitting) {
+                            return textRowsLines.map(line => makeRow(splitLineCells(line).map(makeCell)));
+                        }
+                        return textRowsLines.map(line => makeRow(extractColumnsByTemplate(line, columnTemplate).map(makeCell)));
                     }
-                    const cols = extractColumnsByTemplate(row.value, columnTemplate);
-                    return makeRow(cols.map(makeCell));
-                },
-                'head:text': head => {
-                    const cols = extractColumnsByTemplate(head.value, columnTemplate);
-                    return makeHeaderRow(cols.map(makeCell));
+                    return makeRow(splitToCells(row.value).map(makeCell));
                 },
+                'head:text': head => makeHeaderRow(splitToCells(head.value).map(makeCell)),
             })(node);
-            return res;
+            return normalizeCellCounts(res, 'table');
         },
     });
     return transformer(tree, {});

package/lib/selectors.d.ts

@@ -0,0 +1,18 @@
+import { PodliteDocument, PodNode } from './index';
+/**
+ * Minimal corpus item that the selector engine needs. Concrete consumers
+ * (publisher's `publishRecord`, editor preview, etc.) supply richer
+ * objects; only `file` and `node` are read here.
+ */
+export declare type SelectorDoc = {
+    file: string;
+    node: PodNode | PodliteDocument;
+};
+export declare type ParsedSelector = {
+    scheme?: string;
+    document?: string;
+    anchor?: string;
+    blockFilters: string[];
+};
+export declare const parseSelector: (selector: string) => ParsedSelector | undefined;
+export declare const runSelector: <T extends SelectorDoc>(selector: string, docs: T[]) => PodNode[] | T[];