@graphty/graph-io 0.3.9 → 0.3.11

package/dist/csv.js

@@ -1,8 +1,8 @@
-import { a$ as DICT_SAMPLE_ROWS, a_ as DictHeuristic, X as ROLE_TAKEN_CODE$1, o as ID_MERGED_CODE$1, i as DUPLICATE_EDGE_ID_CODE$1, k as DUPLICATE_NODE_CODE$1, y as MISSING_ID_CODE$1, M as MISSING_ENDPOINT_CODE$1, m as EMPTY_INPUT_CODE$1, aF as resolveImportOptions, I as ImportReportBuilder, aC as reportSinkOptions, aD as reportUnusedOptions, l as DirectionResolver, v as IdCoercer, a as throwIfAborted, aA as parseWeightText, aL as uniqueColumnName, U as RENAMED_CODE, ai as declareResolved, ac as capabilities, L as LOSS, ax as joinText, ak as encodeChunks, ad as checkCapabilities, aE as resolveExportOptions, az as pairFolding, aJ as countMixedEdges, al as explicitWeights, aY as joinListText, aq as formatInteger, am as formatDecimal, ab as canonicalId, z as MIXED_DIRECTION_CODE, g as DIRECTION_FORCED_CODE, h as DIRECTION_REFUSED_CODE, Y as SINK_OPTION_CODE, P as OPTION_IGNORED_CODE, C as COLUMN_RENAMED_CODE, a5 as UNKNOWN_ENCODING_CODE, n as ENCODING_FALLBACK_CODE, s as INVALID_ENCODING_CODE, u as INVALID_UTF8_CODE } from "./chunks/writer-BdMak4_J.js";
+import { a$ as DICT_SAMPLE_ROWS, a_ as DictHeuristic, X as ROLE_TAKEN_CODE$1, o as ID_MERGED_CODE$1, i as DUPLICATE_EDGE_ID_CODE$1, k as DUPLICATE_NODE_CODE$1, y as MISSING_ID_CODE$1, M as MISSING_ENDPOINT_CODE$1, m as EMPTY_INPUT_CODE$1, aF as resolveImportOptions, I as ImportReportBuilder, aC as reportSinkOptions, aD as reportUnusedOptions, l as DirectionResolver, v as IdCoercer, a as throwIfAborted, aA as parseWeightText, aL as uniqueColumnName, U as RENAMED_CODE, ai as declareResolved, ac as capabilities, L as LOSS, ax as joinText, ak as encodeChunks, ad as checkCapabilities, aE as resolveExportOptions, az as pairFolding, aJ as countMixedEdges, al as explicitWeights, aY as joinListText, aq as formatInteger, am as formatDecimal, ab as canonicalId, z as MIXED_DIRECTION_CODE, g as DIRECTION_FORCED_CODE, h as DIRECTION_REFUSED_CODE, Y as SINK_OPTION_CODE, P as OPTION_IGNORED_CODE, C as COLUMN_RENAMED_CODE, a5 as UNKNOWN_ENCODING_CODE, n as ENCODING_FALLBACK_CODE, s as INVALID_ENCODING_CODE, u as INVALID_UTF8_CODE } from "./chunks/writer-BtWpUaiH.js";
 import { i as inferTextDtype, T as TextCellWriter, p as parseTextCell, W as WIDENING_UNSUPPORTED_CODE } from "./chunks/text-Dr0Ifpag.js";
 import { GraphFormatError, INVALID_INDEX } from "@graphty/graph-format";
-import { s as sniffNewline, a as sniffDelimiter, C as CsvRecordReader, B as BAD_QUOTE_CODE, U as UNCLOSED_QUOTE_CODE } from "./chunks/records-BKSMowhR.js";
-import { n as quoteCsvCell } from "./chunks/escape-BjmIaFFo.js";
+import { s as sniffNewline, a as sniffDelimiter, C as CsvRecordReader, D as DELIMITER_CANDIDATES, B as BAD_QUOTE_CODE, U as UNCLOSED_QUOTE_CODE } from "./chunks/records-IHsCfv7s.js";
+import { m as quoteCsvCell } from "./chunks/escape-D1f9-cwf.js";
 const SOURCE_NAMES = Object.freeze([
   "source",
   "src",
@@ -277,7 +277,7 @@ const ID_MERGED_CODE = ID_MERGED_CODE$1;
 const COLUMN_MISSING_CODE = "W_CSV_COLUMN_MISSING";
 const ROLE_TAKEN_CODE = ROLE_TAKEN_CODE$1;
 const DUPLICATE_EDGE_ID_CODE = DUPLICATE_EDGE_ID_CODE$1;
-const TABLE_MODES = /* @__PURE__ */ new Set(["edges", "nodes", "auto"]);
+const TABLE_MODES = /* @__PURE__ */ new Set(["edges", "nodes", "adjacency", "auto"]);
 const USED_OPTIONS = /* @__PURE__ */ new Set([
   "ids",
   "addMissingNodes",
@@ -330,7 +330,7 @@ function resolveCsvOptions(options) {
     });
   }
   if (o.table !== void 0 && !TABLE_MODES.has(o.table)) {
-    throw new GraphFormatError("E_UNSUPPORTED", 'option table: expected "edges", "nodes" or "auto"', {
+    throw new GraphFormatError("E_UNSUPPORTED", 'option table: expected "edges", "nodes", "adjacency" or "auto"', {
       option: "table",
       found: o.table
     });
@@ -341,6 +341,28 @@ function resolveCsvOptions(options) {
   if (o.typeColumn !== null) {
     checkColumnRef("typeColumn", o.typeColumn);
   }
+  if (o.table === "adjacency") {
+    for (const name of ["sourceColumn", "targetColumn", "typeColumn", "idColumn"]) {
+      if (o[name] !== void 0) {
+        throw new GraphFormatError("E_UNSUPPORTED", `option ${name}: an adjacency table has no columns`, {
+          option: name,
+          found: o[name]
+        });
+      }
+    }
+    if (o.rowNumberIds === true) {
+      throw new GraphFormatError("E_UNSUPPORTED", "option rowNumberIds: an adjacency table names its nodes", {
+        option: "rowNumberIds",
+        found: o.rowNumberIds
+      });
+    }
+  }
+  if (o.rowNumberIds !== void 0 && typeof o.rowNumberIds !== "boolean") {
+    throw new GraphFormatError("E_UNSUPPORTED", "option rowNumberIds: expected a boolean", {
+      option: "rowNumberIds",
+      found: o.rowNumberIds
+    });
+  }
   return {
     delimiter: o.delimiter ?? null,
     header: o.header ?? "auto",
@@ -349,7 +371,8 @@ function resolveCsvOptions(options) {
     targetColumn: o.targetColumn ?? null,
     typeColumn: o.typeColumn,
     idColumn: o.idColumn ?? null,
-    nodes: o.nodes ?? null
+    nodes: o.nodes ?? null,
+    rowNumberIds: o.rowNumberIds ?? false
   };
 }
 function checkColumnRef(name, value) {
@@ -417,6 +440,25 @@ function parseKind(text) {
       return null;
   }
 }
+function splitNeighbour(text) {
+  const colon = text.lastIndexOf(":");
+  if (colon < 0) {
+    return { id: text, weight: void 0 };
+  }
+  const suffix = text.slice(colon + 1);
+  if (suffix.length === 0) {
+    return { id: text.slice(0, colon), weight: void 0 };
+  }
+  const weight = numberOrNull(suffix);
+  return weight === null ? { id: text, weight: void 0 } : { id: text.slice(0, colon), weight };
+}
+function numberOrNull(text) {
+  try {
+    return parseWeightText(text) ?? null;
+  } catch {
+    return null;
+  }
+}
 class TableReader {
   /**
    * Create a table reader.
@@ -436,9 +478,12 @@ class TableReader {
     this.where = { line: null, element: null };
     this.state = state;
     this.kind = kind;
+    this.rowNumbers = state.csv.rowNumberIds && (kind === "nodes" || kind === "auto" && state.csv.nodes === null);
     const readerOptions = {
       delimiter: state.csv.delimiter,
       comments: COMMENT_CHARS,
+      // field counts say nothing about an adjacency table, whose rows vary in width
+      skipQuoteErrors: kind === "adjacency",
       signal: state.common.signal,
       onProgress: progress ? state.common.onProgress : null,
       encoding: state.common.encoding
@@ -463,12 +508,20 @@ class TableReader {
   async readRows(iterator) {
     const { report } = this.state;
     const first = await iterator.next();
+    if (first.done && this.kind === "adjacency") {
+      return;
+    }
     const firstRow = first.done ? report.fail(EMPTY_INPUT_CODE, "the input is empty: no header row and no records") : first.value;
     const firstLine = this.reader.line;
     const firstQuoted = this.reader.quoted.slice(0, firstRow.length);
     const pending = [];
     let header;
-    const { header: mode } = this.state.csv;
+    let mode = this.state.csv.header;
+    if (mode === "auto" && this.kind === "adjacency") {
+      mode = false;
+    } else if (mode === "auto" && this.rowNumbers) {
+      mode = true;
+    }
     if (mode === "auto") {
       const second = await iterator.next();
       const secondRow = second.done ? null : second.value;
@@ -527,6 +580,9 @@ class TableReader {
    */
   resolvePlan(names, header, line) {
     const { csv, report } = this.state;
+    if (this.kind === "adjacency") {
+      return { kind: "adjacency", names: [] };
+    }
     const width = names.length;
     let source = -1;
     let target = -1;
@@ -565,7 +621,7 @@ class TableReader {
       );
     }
     const idResolves = header ? findColumn(names, ID_NAMES) >= 0 : width >= 1;
-    if (this.kind === "auto" && csv.idColumn === null && !idResolves) {
+    if (this.kind === "auto" && csv.idColumn === null && !idResolves && !this.rowNumbers) {
       report.fail(
         NO_ENDPOINT_COLUMNS_CODE,
         `no source / target columns and no id column in the header (${shown}); the input is neither an edge table nor a node table`,
@@ -674,6 +730,10 @@ class TableReader {
         id = -1;
         break;
       default:
+        if (idColumn < 0 && this.rowNumbers) {
+          id = -1;
+          break;
+        }
         if (idColumn < 0) {
           report.fail(
             NO_ID_COLUMN_CODE,
@@ -695,7 +755,8 @@ class TableReader {
         attributes.push(i);
       }
     }
-    return { kind: "nodes", names, width: names.length, id, label, attributes };
+    const rowNumber = id < 0 && common.nodeIdFrom !== "index";
+    return { kind: "nodes", names, width: names.length, id, rowNumber, label, attributes };
   }
   /**
    * Declare the role columns and create a writer per attribute column.
@@ -703,6 +764,9 @@ class TableReader {
    */
   prepareColumns(line) {
     const plan = this.requirePlan();
+    if (plan.kind === "adjacency") {
+      return;
+    }
     const { sink, report } = this.state;
     const domain = plan.kind === "edges" ? "edge" : "node";
     const origin = { format: "csv" };
@@ -752,8 +816,72 @@ class TableReader {
     this.dataRows++;
     if (plan.kind === "edges") {
       this.processEdgeRow(plan, row, quoted, line);
-    } else {
+    } else if (plan.kind === "nodes") {
       this.processNodeRow(plan, row, quoted, line);
+    } else {
+      this.processAdjacencyRow(row, quoted, line);
+    }
+  }
+  /**
+   * Push one adjacency row: the node, then one edge per set neighbour cell, in row order.
+   * @param row - the cells
+   * @param quoted - whether each cell was quoted
+   * @param line - the row's line
+   */
+  processAdjacencyRow(row, quoted, line) {
+    const { report, sink, resolver, common } = this.state;
+    const { counts } = report;
+    const { where } = this;
+    where.line = line;
+    where.element = null;
+    let neighbours = 0;
+    for (let k = 1; k < row.length; k++) {
+      if (!isUnset(row[k], quoted[k])) {
+        neighbours++;
+      }
+    }
+    if (isUnset(row[0], quoted[0])) {
+      report.error("missing-value", MISSING_ID_CODE, `line ${line}: blank node cell`, { line });
+      counts.skippedNodes++;
+      counts.skippedEdges += neighbours;
+      return;
+    }
+    const kind = this.state.commentDirected ?? common.defaultDirected ? "directed" : "undirected";
+    if (!this.state.headerSet) {
+      this.state.headerSet = true;
+      resolver.setHeader(kind !== "undirected", where);
+    }
+    let source;
+    try {
+      where.element = row[0];
+      source = this.coerce(row[0]);
+      if (sink.indexOf(source) === INVALID_INDEX) {
+        counts.nodes++;
+      }
+      sink.addNode(source);
+    } catch (err) {
+      report.recordError(err, where);
+      counts.skippedNodes++;
+      counts.skippedEdges += neighbours;
+      return;
+    }
+    for (let k = 1; k < row.length; k++) {
+      if (isUnset(row[k], quoted[k])) {
+        continue;
+      }
+      where.element = row[k];
+      try {
+        const cell = splitNeighbour(row[k]);
+        const target = this.coerce(cell.id);
+        const targetNew = sink.indexOf(target) === INVALID_INDEX;
+        const before = sink.edgeCount;
+        resolver.addEdge(source, target, kind, common.weightFrom === null ? void 0 : cell.weight, where);
+        counts.edges += sink.edgeCount - before;
+        counts.nodes += targetNew ? 1 : 0;
+      } catch (err) {
+        report.recordError(err, where);
+        counts.skippedEdges++;
+      }
     }
   }
   /**
@@ -884,7 +1012,7 @@ class TableReader {
     where.element = idText;
     let index;
     try {
-      const id = plan.id >= 0 ? this.coerce(idText) : ordinal;
+      const id = plan.id >= 0 || plan.rowNumber ? this.coerce(idText) : ordinal;
       if (sink.indexOf(id) !== INVALID_INDEX) {
         report.warning(
           "merged",
@@ -1069,7 +1197,9 @@ const CSV_LOSS = Object.freeze({
   /** The edge table carries no node without an edge: isolated nodes vanish on re-import. */
   ISOLATED_NODES: "W_CSV_ISOLATED_NODES",
   /** The edge table lists nodes by first appearance; the node order (and indices) change on re-import. */
-  NODE_ORDER: "W_CSV_NODE_ORDER"
+  NODE_ORDER: "W_CSV_NODE_ORDER",
+  /** An adjacency table holds no edge column but the weight: edge ids, labels and attributes are not written. */
+  EDGE_COLUMNS: "W_CSV_EDGE_COLUMNS"
 });
 const CSV_CAPABILITIES = capabilities({
   mixedDirection: true,
@@ -1102,7 +1232,8 @@ const SKIPPED_ROLES = /* @__PURE__ */ new Set([
   "timestamp",
   "timestamps",
   "spells",
-  "open"
+  "open",
+  "spellsOpen"
 ]);
 function resolveCsvExportOptions(options) {
   const o = options ?? {};
@@ -1112,8 +1243,8 @@ function resolveCsvExportOptions(options) {
       found: o.dialect
     });
   }
-  if (o.table !== void 0 && o.table !== "edges" && o.table !== "nodes") {
-    throw new GraphFormatError("E_UNSUPPORTED", 'option table: expected "edges" or "nodes"', {
+  if (o.table !== void 0 && o.table !== "edges" && o.table !== "nodes" && o.table !== "adjacency") {
+    throw new GraphFormatError("E_UNSUPPORTED", 'option table: expected "edges", "nodes" or "adjacency"', {
       option: "table",
       found: o.table
     });
@@ -1131,6 +1262,12 @@ function resolveCsvExportOptions(options) {
       found: o.newline
     });
   }
+  if (o.table === "adjacency" && o.header === true) {
+    throw new GraphFormatError("E_UNSUPPORTED", "option header: an adjacency table has no header row", {
+      option: "header",
+      found: o.header
+    });
+  }
   if (o.header !== void 0 && typeof o.header !== "boolean") {
     throw new GraphFormatError("E_UNSUPPORTED", "option header: expected a boolean", {
       option: "header",
@@ -1255,19 +1392,20 @@ function planExport(snapshot, options, values) {
       edgeRows.push(e);
     }
   }
-  if (csv.dialect.type === null) {
+  if (csv.dialect.type === null || csv.table === "adjacency") {
+    const where = csv.table === "adjacency" ? "an adjacency table" : "the generic dialect";
     const mixed = countMixedEdges(snapshot);
     if (!snapshot.directed) {
       note(
         CSV_LOSS.DIRECTION_DROPPED,
-        `the generic dialect has no direction column; ${snapshot.edgeCount} undirected edge(s) read back as directed unless the importer is told otherwise`,
+        `${where} has no direction column; ${snapshot.edgeCount} undirected edge(s) read back as directed unless the importer is told otherwise`,
         null,
         snapshot.edgeCount
       );
     } else if (mixed > 0) {
       note(
         CSV_LOSS.DIRECTION_DROPPED,
-        `the generic dialect has no direction column; ${mixed} undirected edge(s) of a mixed graph read back as one directed edge each`,
+        `${where} has no direction column; ${mixed} undirected edge(s) of a mixed graph read back as one directed edge each`,
         null,
         mixed
       );
@@ -1306,6 +1444,17 @@ function planExport(snapshot, options, values) {
     edgeLabel,
     csv.table === "edges" ? note : null
   );
+  if (csv.table === "adjacency") {
+    const dropped = edgeColumns.length + (edgeId === null ? 0 : 1) + (edgeLabel === null ? 0 : 1);
+    if (dropped > 0) {
+      note(
+        CSV_LOSS.EDGE_COLUMNS,
+        `${dropped} edge column(s) are not written: an adjacency table holds the weight only`,
+        null,
+        dropped
+      );
+    }
+  }
   if (csv.table === "edges" && values) {
     checkRoleColumn(edgeId, "id", EDGE_ID_NAMES, "edge", edgeRows, note);
     checkRoleColumn(edgeLabel, "label", LABEL_NAMES, "edge", edgeRows, note);
@@ -1556,8 +1705,80 @@ function typeText(snapshot, plan, e) {
   }
   return "Directed";
 }
+function neighbourCell(id, weight) {
+  if (weight !== null) {
+    return `${id}:${weight}`;
+  }
+  return id.includes(":") ? `${id}:` : id;
+}
+function* adjacencyLines(snapshot, plan) {
+  const { delimiter, newline } = plan.csv;
+  const forced = (text) => `"${text.replace(/"/g, '""')}"`;
+  const quote = (text) => DELIMITER_CANDIDATES.some((c) => text.includes(c)) ? forced(text) : quoteCsvCell(text, delimiter);
+  const first = (text) => /^[#%]/.test(text) ? forced(text) : quote(text);
+  const list = snapshot.edgeList();
+  const { edgeRows, weights } = plan;
+  const seen = new Uint8Array(snapshot.nodeCount);
+  let next = 0;
+  const advance = () => {
+    while (next < seen.length && seen[next] === 1) {
+      next++;
+    }
+  };
+  function* singles(upTo) {
+    for (let i = next; i <= upTo; i++) {
+      if (seen[i] === 0) {
+        seen[i] = 1;
+        yield first(plan.idText(i)) + newline;
+      }
+    }
+    advance();
+  }
+  let r = 0;
+  while (r < edgeRows.length) {
+    const source = list.src[edgeRows[r]];
+    let end = r;
+    while (end < edgeRows.length && list.src[edgeRows[end]] === source) {
+      end++;
+    }
+    const fresh = [];
+    let highest = -1;
+    let inOrder = true;
+    const mark = (v) => {
+      if (seen[v] === 0) {
+        seen[v] = 1;
+        inOrder &&= v === next + fresh.length;
+        fresh.push(v);
+        highest = Math.max(highest, v);
+      }
+    };
+    mark(source);
+    for (let k = r; k < end; k++) {
+      mark(list.dst[edgeRows[k]]);
+    }
+    if (!inOrder) {
+      for (const v of fresh) {
+        seen[v] = 0;
+      }
+      yield* singles(highest);
+    }
+    advance();
+    const cells = [first(plan.idText(source))];
+    for (let k = r; k < end; k++) {
+      const e = edgeRows[k];
+      cells.push(quote(neighbourCell(plan.idText(list.dst[e]), weights.weighted ? weights.text(e) : null)));
+    }
+    yield cells.join(delimiter) + newline;
+    r = end;
+  }
+  yield* singles(snapshot.nodeCount - 1);
+}
 function* lines(snapshot, plan) {
   const { csv } = plan;
+  if (csv.table === "adjacency") {
+    yield* adjacencyLines(snapshot, plan);
+    return;
+  }
   const { delimiter, newline } = csv;
   const quote = (text) => quoteCsvCell(text, delimiter);
   const cell = (column, row) => {