kirograph 0.16.0 → 0.17.0

package/dist/data/parsers/jsonl.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../../src/data/parsers/jsonl.ts"],
+  "sourcesContent": ["/**\n * JSONL/NDJSON Streaming Parser\n *\n * Handles: .jsonl, .ndjson\n * Parses one JSON object per line without loading the full file.\n */\n\nimport * as fs from 'fs';\nimport * as readline from 'readline';\nimport type { DataFormatParser, ParsedRow } from '../types';\n\nconst BATCH_SIZE = 10_000;\n\nexport const jsonlParser: DataFormatParser = {\n  name: 'jsonl',\n  extensions: ['.jsonl', '.ndjson'],\n  isAvailable: () => true,\n\n  async parse(filePath, opts) {\n    const stream = fs.createReadStream(filePath, { encoding: 'utf8' });\n    const rl = readline.createInterface({ input: stream, crlfDelay: Infinity });\n\n    const columnSet = new Set<string>();\n    let totalRows = 0;\n    let batch: ParsedRow[] = [];\n\n    for await (const line of rl) {\n      if (line.trim() === '') continue;\n      if (totalRows >= opts.maxRows) break;\n\n      try {\n        const obj = JSON.parse(line);\n        if (typeof obj !== 'object' || obj === null || Array.isArray(obj)) continue;\n\n        const row: ParsedRow = {};\n        for (const [key, value] of Object.entries(obj)) {\n          columnSet.add(key);\n          if (value === null || value === undefined) {\n            row[key] = null;\n          } else if (typeof value === 'object') {\n            row[key] = JSON.stringify(value);\n          } else {\n            row[key] = value as string | number | boolean;\n          }\n        }\n\n        batch.push(row);\n        totalRows++;\n\n        if (batch.length >= BATCH_SIZE) {\n          opts.onBatch(batch, [...columnSet]);\n          batch = [];\n        }\n      } catch {\n        // Skip malformed lines\n      }\n    }\n\n    const columns = [...columnSet];\n    if (batch.length > 0) {\n      opts.onBatch(batch, columns);\n    }\n\n    return { columns, totalRows };\n  },\n};\n"],
+  "mappings": ";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAOA,SAAoB;AACpB,eAA0B;AAG1B,MAAM,aAAa;AAEZ,MAAM,cAAgC;AAAA,EAC3C,MAAM;AAAA,EACN,YAAY,CAAC,UAAU,SAAS;AAAA,EAChC,aAAa,MAAM;AAAA,EAEnB,MAAM,MAAM,UAAU,MAAM;AAC1B,UAAM,SAAS,GAAG,iBAAiB,UAAU,EAAE,UAAU,OAAO,CAAC;AACjE,UAAM,KAAK,SAAS,gBAAgB,EAAE,OAAO,QAAQ,WAAW,SAAS,CAAC;AAE1E,UAAM,YAAY,oBAAI,IAAY;AAClC,QAAI,YAAY;AAChB,QAAI,QAAqB,CAAC;AAE1B,qBAAiB,QAAQ,IAAI;AAC3B,UAAI,KAAK,KAAK,MAAM,GAAI;AACxB,UAAI,aAAa,KAAK,QAAS;AAE/B,UAAI;AACF,cAAM,MAAM,KAAK,MAAM,IAAI;AAC3B,YAAI,OAAO,QAAQ,YAAY,QAAQ,QAAQ,MAAM,QAAQ,GAAG,EAAG;AAEnE,cAAM,MAAiB,CAAC;AACxB,mBAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,GAAG,GAAG;AAC9C,oBAAU,IAAI,GAAG;AACjB,cAAI,UAAU,QAAQ,UAAU,QAAW;AACzC,gBAAI,GAAG,IAAI;AAAA,UACb,WAAW,OAAO,UAAU,UAAU;AACpC,gBAAI,GAAG,IAAI,KAAK,UAAU,KAAK;AAAA,UACjC,OAAO;AACL,gBAAI,GAAG,IAAI;AAAA,UACb;AAAA,QACF;AAEA,cAAM,KAAK,GAAG;AACd;AAEA,YAAI,MAAM,UAAU,YAAY;AAC9B,eAAK,QAAQ,OAAO,CAAC,GAAG,SAAS,CAAC;AAClC,kBAAQ,CAAC;AAAA,QACX;AAAA,MACF,QAAQ;AAAA,MAER;AAAA,IACF;AAEA,UAAM,UAAU,CAAC,GAAG,SAAS;AAC7B,QAAI,MAAM,SAAS,GAAG;AACpB,WAAK,QAAQ,OAAO,OAAO;AAAA,IAC7B;AAEA,WAAO,EAAE,SAAS,UAAU;AAAA,EAC9B;AACF;",
+  "names": []
+}

package/dist/data/parsers/parquet.js

@@ -0,0 +1,95 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var parquet_exports = {};
+__export(parquet_exports, {
+  parquetParser: () => parquetParser
+});
+module.exports = __toCommonJS(parquet_exports);
+const BATCH_SIZE = 1e4;
+let parquetModule = null;
+let parquetChecked = false;
+function tryLoadParquet() {
+  if (parquetChecked) return parquetModule !== null;
+  parquetChecked = true;
+  try {
+    parquetModule = require("parquetjs-lite");
+    return true;
+  } catch {
+    try {
+      parquetModule = require("parquetjs");
+      return true;
+    } catch {
+      return false;
+    }
+  }
+}
+const parquetParser = {
+  name: "parquet",
+  extensions: [".parquet"],
+  isAvailable() {
+    return tryLoadParquet();
+  },
+  async parse(filePath, opts) {
+    if (!tryLoadParquet()) {
+      return { columns: [], totalRows: 0 };
+    }
+    const reader = await parquetModule.ParquetReader.openFile(filePath);
+    const schema = reader.getSchema();
+    const columns = Object.keys(schema.fields);
+    const cursor = reader.getCursor();
+    let totalRows = 0;
+    let batch = [];
+    let record;
+    while ((record = await cursor.next()) && totalRows < opts.maxRows) {
+      const row = {};
+      for (const col of columns) {
+        const val = record[col];
+        if (val === null || val === void 0) {
+          row[col] = null;
+        } else if (val instanceof Date) {
+          row[col] = val.toISOString();
+        } else if (Buffer.isBuffer(val)) {
+          row[col] = val.toString("utf8");
+        } else if (typeof val === "bigint") {
+          row[col] = Number(val);
+        } else if (typeof val === "object") {
+          row[col] = JSON.stringify(val);
+        } else {
+          row[col] = val;
+        }
+      }
+      batch.push(row);
+      totalRows++;
+      if (batch.length >= BATCH_SIZE) {
+        opts.onBatch(batch, columns);
+        batch = [];
+      }
+    }
+    if (batch.length > 0) {
+      opts.onBatch(batch, columns);
+    }
+    await reader.close();
+    return { columns, totalRows };
+  }
+};
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  parquetParser
+});
+//# sourceMappingURL=parquet.js.map

package/dist/data/parsers/parquet.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../../src/data/parsers/parquet.ts"],
+  "sourcesContent": ["/**\n * Parquet Parser (Optional Dependency)\n *\n * Handles: .parquet\n * Requires: parquetjs-lite package (optional dependency)\n * Gracefully returns isAvailable()=false if not installed.\n */\n\nimport type { DataFormatParser, ParsedRow } from '../types';\n\nconst BATCH_SIZE = 10_000;\n\nlet parquetModule: any = null;\nlet parquetChecked = false;\n\nfunction tryLoadParquet(): boolean {\n  if (parquetChecked) return parquetModule !== null;\n  parquetChecked = true;\n  try {\n    parquetModule = require('parquetjs-lite');\n    return true;\n  } catch {\n    try {\n      parquetModule = require('parquetjs');\n      return true;\n    } catch {\n      return false;\n    }\n  }\n}\n\nexport const parquetParser: DataFormatParser = {\n  name: 'parquet',\n  extensions: ['.parquet'],\n\n  isAvailable(): boolean {\n    return tryLoadParquet();\n  },\n\n  async parse(filePath, opts) {\n    if (!tryLoadParquet()) {\n      return { columns: [], totalRows: 0 };\n    }\n\n    const reader = await parquetModule.ParquetReader.openFile(filePath);\n    const schema = reader.getSchema();\n    const columns = Object.keys(schema.fields);\n\n    const cursor = reader.getCursor();\n    let totalRows = 0;\n    let batch: ParsedRow[] = [];\n    let record: any;\n\n    while ((record = await cursor.next()) && totalRows < opts.maxRows) {\n      const row: ParsedRow = {};\n      for (const col of columns) {\n        const val = record[col];\n        if (val === null || val === undefined) {\n          row[col] = null;\n        } else if (val instanceof Date) {\n          row[col] = val.toISOString();\n        } else if (Buffer.isBuffer(val)) {\n          row[col] = val.toString('utf8');\n        } else if (typeof val === 'bigint') {\n          row[col] = Number(val);\n        } else if (typeof val === 'object') {\n          row[col] = JSON.stringify(val);\n        } else {\n          row[col] = val;\n        }\n      }\n\n      batch.push(row);\n      totalRows++;\n\n      if (batch.length >= BATCH_SIZE) {\n        opts.onBatch(batch, columns);\n        batch = [];\n      }\n    }\n\n    if (batch.length > 0) {\n      opts.onBatch(batch, columns);\n    }\n\n    await reader.close();\n    return { columns, totalRows };\n  },\n};\n"],
+  "mappings": ";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAUA,MAAM,aAAa;AAEnB,IAAI,gBAAqB;AACzB,IAAI,iBAAiB;AAErB,SAAS,iBAA0B;AACjC,MAAI,eAAgB,QAAO,kBAAkB;AAC7C,mBAAiB;AACjB,MAAI;AACF,oBAAgB,QAAQ,gBAAgB;AACxC,WAAO;AAAA,EACT,QAAQ;AACN,QAAI;AACF,sBAAgB,QAAQ,WAAW;AACnC,aAAO;AAAA,IACT,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEO,MAAM,gBAAkC;AAAA,EAC7C,MAAM;AAAA,EACN,YAAY,CAAC,UAAU;AAAA,EAEvB,cAAuB;AACrB,WAAO,eAAe;AAAA,EACxB;AAAA,EAEA,MAAM,MAAM,UAAU,MAAM;AAC1B,QAAI,CAAC,eAAe,GAAG;AACrB,aAAO,EAAE,SAAS,CAAC,GAAG,WAAW,EAAE;AAAA,IACrC;AAEA,UAAM,SAAS,MAAM,cAAc,cAAc,SAAS,QAAQ;AAClE,UAAM,SAAS,OAAO,UAAU;AAChC,UAAM,UAAU,OAAO,KAAK,OAAO,MAAM;AAEzC,UAAM,SAAS,OAAO,UAAU;AAChC,QAAI,YAAY;AAChB,QAAI,QAAqB,CAAC;AAC1B,QAAI;AAEJ,YAAQ,SAAS,MAAM,OAAO,KAAK,MAAM,YAAY,KAAK,SAAS;AACjE,YAAM,MAAiB,CAAC;AACxB,iBAAW,OAAO,SAAS;AACzB,cAAM,MAAM,OAAO,GAAG;AACtB,YAAI,QAAQ,QAAQ,QAAQ,QAAW;AACrC,cAAI,GAAG,IAAI;AAAA,QACb,WAAW,eAAe,MAAM;AAC9B,cAAI,GAAG,IAAI,IAAI,YAAY;AAAA,QAC7B,WAAW,OAAO,SAAS,GAAG,GAAG;AAC/B,cAAI,GAAG,IAAI,IAAI,SAAS,MAAM;AAAA,QAChC,WAAW,OAAO,QAAQ,UAAU;AAClC,cAAI,GAAG,IAAI,OAAO,GAAG;AAAA,QACvB,WAAW,OAAO,QAAQ,UAAU;AAClC,cAAI,GAAG,IAAI,KAAK,UAAU,GAAG;AAAA,QAC/B,OAAO;AACL,cAAI,GAAG,IAAI;AAAA,QACb;AAAA,MACF;AAEA,YAAM,KAAK,GAAG;AACd;AAEA,UAAI,MAAM,UAAU,YAAY;AAC9B,aAAK,QAAQ,OAAO,OAAO;AAC3B,gBAAQ,CAAC;AAAA,MACX;AAAA,IACF;AAEA,QAAI,MAAM,SAAS,GAAG;AACpB,WAAK,QAAQ,OAAO,OAAO;AAAA,IAC7B;AAEA,UAAM,OAAO,MAAM;AACnB,WAAO,EAAE,SAAS,UAAU;AAAA,EAC9B;AACF;",
+  "names": []
+}

package/dist/data/profiler.js

@@ -0,0 +1,182 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var profiler_exports = {};
+__export(profiler_exports, {
+  ColumnProfiler: () => ColumnProfiler
+});
+module.exports = __toCommonJS(profiler_exports);
+function detectType(value) {
+  if (value === null || value === void 0) return "null";
+  if (typeof value === "boolean") return "boolean";
+  if (typeof value === "number") {
+    return Number.isInteger(value) ? "integer" : "float";
+  }
+  if (typeof value === "string") {
+    if (/^\d{4}-\d{2}-\d{2}/.test(value) && !isNaN(Date.parse(value))) return "date";
+    return "string";
+  }
+  return "string";
+}
+function inferFinalType(types, totalCount) {
+  const nonNull = new Map(types);
+  nonNull.delete("null");
+  if (nonNull.size === 0) return "null";
+  if (nonNull.size === 1) return [...nonNull.keys()][0];
+  if (nonNull.size === 2 && nonNull.has("integer") && nonNull.has("float")) return "float";
+  for (const [type, count] of nonNull) {
+    if (count / totalCount > 0.8) return type;
+  }
+  return "mixed";
+}
+class ColumnProfiler {
+  constructor() {
+    this.stats = /* @__PURE__ */ new Map();
+    this.columnOrder = [];
+  }
+  /**
+   * Process a batch of rows.
+   */
+  addBatch(rows, columns) {
+    for (let i = 0; i < columns.length; i++) {
+      const col = columns[i];
+      if (!this.stats.has(col)) {
+        this.columnOrder.push(col);
+        this.stats.set(col, {
+          name: col,
+          position: this.columnOrder.length - 1,
+          types: /* @__PURE__ */ new Map(),
+          nullCount: 0,
+          totalCount: 0,
+          distinctValues: /* @__PURE__ */ new Set(),
+          numericSum: 0,
+          numericCount: 0,
+          minValue: null,
+          maxValue: null,
+          samples: []
+        });
+      }
+    }
+    for (const row of rows) {
+      for (const col of columns) {
+        const stat = this.stats.get(col);
+        const value = row[col];
+        stat.totalCount++;
+        if (value === null || value === void 0) {
+          stat.nullCount++;
+          stat.types.set("null", (stat.types.get("null") ?? 0) + 1);
+          continue;
+        }
+        const type = detectType(value);
+        stat.types.set(type, (stat.types.get(type) ?? 0) + 1);
+        const strVal = String(value);
+        if (stat.distinctValues.size < 1e4) {
+          stat.distinctValues.add(strVal);
+        }
+        if (typeof value === "number" && !isNaN(value)) {
+          stat.numericSum += value;
+          stat.numericCount++;
+          const numStr = String(value);
+          if (stat.minValue === null || value < parseFloat(stat.minValue)) stat.minValue = numStr;
+          if (stat.maxValue === null || value > parseFloat(stat.maxValue)) stat.maxValue = numStr;
+        } else if (type === "string" || type === "date") {
+          if (stat.minValue === null || strVal < stat.minValue) stat.minValue = strVal;
+          if (stat.maxValue === null || strVal > stat.maxValue) stat.maxValue = strVal;
+        }
+        if (stat.samples.length < 5 && !stat.samples.includes(strVal) && strVal.length <= 100) {
+          stat.samples.push(strVal);
+        }
+      }
+    }
+  }
+  /**
+   * Finalize and return column profiles.
+   */
+  finalize(datasetId) {
+    const now = Date.now();
+    const profiles = [];
+    for (const col of this.columnOrder) {
+      const stat = this.stats.get(col);
+      const inferredType = inferFinalType(stat.types, stat.totalCount);
+      const nullPct = stat.totalCount > 0 ? stat.nullCount / stat.totalCount : 0;
+      const meanValue = stat.numericCount > 0 ? stat.numericSum / stat.numericCount : null;
+      profiles.push({
+        id: `${datasetId}::${col}#column`,
+        datasetId,
+        name: col,
+        position: stat.position,
+        inferredType,
+        nullable: stat.nullCount > 0,
+        nullCount: stat.nullCount,
+        nullPct,
+        cardinality: stat.distinctValues.size,
+        minValue: stat.minValue,
+        maxValue: stat.maxValue,
+        meanValue,
+        sampleValues: stat.samples,
+        summary: generateColumnSummary(col, inferredType, stat, nullPct),
+        updatedAt: now
+      });
+    }
+    return profiles;
+  }
+}
+function generateColumnSummary(name, type, stat, nullPct) {
+  const parts = [];
+  const cardinality = stat.distinctValues.size;
+  const total = stat.totalCount;
+  if (type === "integer" || type === "float") {
+    parts.push(`Numeric (${type})`);
+    if (stat.minValue != null && stat.maxValue != null) {
+      parts.push(`range ${stat.minValue}\u2013${stat.maxValue}`);
+    }
+  } else if (type === "boolean") {
+    parts.push("Boolean flag");
+  } else if (type === "date") {
+    parts.push("Date/timestamp");
+    if (stat.minValue && stat.maxValue) {
+      parts.push(`from ${stat.minValue} to ${stat.maxValue}`);
+    }
+  } else if (type === "string") {
+    if (cardinality <= 10 && total > 20) {
+      parts.push(`Categorical (${cardinality} values)`);
+    } else if (cardinality === total && total > 10) {
+      parts.push("Unique identifier");
+    } else {
+      parts.push("Text");
+    }
+  } else {
+    parts.push(type);
+  }
+  if (cardinality === 1 && total > 1) {
+    parts.push("constant");
+  } else if (total > 0 && cardinality / total > 0.95 && total > 10) {
+    parts.push("near-unique");
+  }
+  if (nullPct > 0.5) {
+    parts.push(`mostly null (${(nullPct * 100).toFixed(0)}%)`);
+  } else if (nullPct > 0.1) {
+    parts.push(`some nulls (${(nullPct * 100).toFixed(0)}%)`);
+  }
+  return parts.join(", ");
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  ColumnProfiler
+});
+//# sourceMappingURL=profiler.js.map

package/dist/data/profiler.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/data/profiler.ts"],
+  "sourcesContent": ["/**\n * Column Profiler\n *\n * Analyzes parsed rows to produce column profiles:\n * - Type inference (string, integer, float, boolean, date, null, mixed)\n * - Null count and percentage\n * - Cardinality (distinct values)\n * - Min/max values\n * - Mean (for numeric columns)\n * - Sample values (up to 5)\n */\n\nimport type { DataColumn, InferredType, ParsedRow } from './types';\n\ninterface ColumnStats {\n  name: string;\n  position: number;\n  types: Map<string, number>;  // type \u2192 count\n  nullCount: number;\n  totalCount: number;\n  distinctValues: Set<string>;\n  numericSum: number;\n  numericCount: number;\n  minValue: string | null;\n  maxValue: string | null;\n  samples: string[];\n}\n\nfunction detectType(value: any): string {\n  if (value === null || value === undefined) return 'null';\n  if (typeof value === 'boolean') return 'boolean';\n  if (typeof value === 'number') {\n    return Number.isInteger(value) ? 'integer' : 'float';\n  }\n  if (typeof value === 'string') {\n    // Try to detect dates\n    if (/^\\d{4}-\\d{2}-\\d{2}/.test(value) && !isNaN(Date.parse(value))) return 'date';\n    return 'string';\n  }\n  return 'string';\n}\n\nfunction inferFinalType(types: Map<string, number>, totalCount: number): InferredType {\n  // Remove null from consideration\n  const nonNull = new Map(types);\n  nonNull.delete('null');\n\n  if (nonNull.size === 0) return 'null';\n  if (nonNull.size === 1) return [...nonNull.keys()][0] as InferredType;\n\n  // integer + float \u2192 float\n  if (nonNull.size === 2 && nonNull.has('integer') && nonNull.has('float')) return 'float';\n\n  // If one type dominates (>80%), use it\n  for (const [type, count] of nonNull) {\n    if (count / totalCount > 0.8) return type as InferredType;\n  }\n\n  return 'mixed';\n}\n\n/**\n * Profile columns from a batch of rows.\n * Call repeatedly with batches, then call finalize() to get the final profiles.\n */\nexport class ColumnProfiler {\n  private stats = new Map<string, ColumnStats>();\n  private columnOrder: string[] = [];\n\n  /**\n   * Process a batch of rows.\n   */\n  addBatch(rows: ParsedRow[], columns: string[]): void {\n    // Initialize stats for new columns\n    for (let i = 0; i < columns.length; i++) {\n      const col = columns[i];\n      if (!this.stats.has(col)) {\n        this.columnOrder.push(col);\n        this.stats.set(col, {\n          name: col,\n          position: this.columnOrder.length - 1,\n          types: new Map(),\n          nullCount: 0,\n          totalCount: 0,\n          distinctValues: new Set(),\n          numericSum: 0,\n          numericCount: 0,\n          minValue: null,\n          maxValue: null,\n          samples: [],\n        });\n      }\n    }\n\n    // Process each row\n    for (const row of rows) {\n      for (const col of columns) {\n        const stat = this.stats.get(col)!;\n        const value = row[col];\n        stat.totalCount++;\n\n        if (value === null || value === undefined) {\n          stat.nullCount++;\n          stat.types.set('null', (stat.types.get('null') ?? 0) + 1);\n          continue;\n        }\n\n        const type = detectType(value);\n        stat.types.set(type, (stat.types.get(type) ?? 0) + 1);\n\n        // Track distinct values (cap at 10000 to avoid memory issues)\n        const strVal = String(value);\n        if (stat.distinctValues.size < 10_000) {\n          stat.distinctValues.add(strVal);\n        }\n\n        // Numeric stats\n        if (typeof value === 'number' && !isNaN(value)) {\n          stat.numericSum += value;\n          stat.numericCount++;\n          const numStr = String(value);\n          if (stat.minValue === null || value < parseFloat(stat.minValue)) stat.minValue = numStr;\n          if (stat.maxValue === null || value > parseFloat(stat.maxValue)) stat.maxValue = numStr;\n        } else if (type === 'string' || type === 'date') {\n          if (stat.minValue === null || strVal < stat.minValue) stat.minValue = strVal;\n          if (stat.maxValue === null || strVal > stat.maxValue) stat.maxValue = strVal;\n        }\n\n        // Collect samples (up to 5 unique)\n        if (stat.samples.length < 5 && !stat.samples.includes(strVal) && strVal.length <= 100) {\n          stat.samples.push(strVal);\n        }\n      }\n    }\n  }\n\n  /**\n   * Finalize and return column profiles.\n   */\n  finalize(datasetId: string): DataColumn[] {\n    const now = Date.now();\n    const profiles: DataColumn[] = [];\n\n    for (const col of this.columnOrder) {\n      const stat = this.stats.get(col)!;\n      const inferredType = inferFinalType(stat.types, stat.totalCount);\n      const nullPct = stat.totalCount > 0 ? stat.nullCount / stat.totalCount : 0;\n      const meanValue = stat.numericCount > 0 ? stat.numericSum / stat.numericCount : null;\n\n      profiles.push({\n        id: `${datasetId}::${col}#column`,\n        datasetId,\n        name: col,\n        position: stat.position,\n        inferredType,\n        nullable: stat.nullCount > 0,\n        nullCount: stat.nullCount,\n        nullPct,\n        cardinality: stat.distinctValues.size,\n        minValue: stat.minValue,\n        maxValue: stat.maxValue,\n        meanValue,\n        sampleValues: stat.samples,\n        summary: generateColumnSummary(col, inferredType, stat, nullPct),\n        updatedAt: now,\n      });\n    }\n\n    return profiles;\n  }\n}\n\n/**\n * Generate a natural-language summary for a column based on its profile.\n */\nfunction generateColumnSummary(\n  name: string,\n  type: InferredType,\n  stat: ColumnStats,\n  nullPct: number,\n): string {\n  const parts: string[] = [];\n  const cardinality = stat.distinctValues.size;\n  const total = stat.totalCount;\n\n  // Type description\n  if (type === 'integer' || type === 'float') {\n    parts.push(`Numeric (${type})`);\n    if (stat.minValue != null && stat.maxValue != null) {\n      parts.push(`range ${stat.minValue}\u2013${stat.maxValue}`);\n    }\n  } else if (type === 'boolean') {\n    parts.push('Boolean flag');\n  } else if (type === 'date') {\n    parts.push('Date/timestamp');\n    if (stat.minValue && stat.maxValue) {\n      parts.push(`from ${stat.minValue} to ${stat.maxValue}`);\n    }\n  } else if (type === 'string') {\n    if (cardinality <= 10 && total > 20) {\n      parts.push(`Categorical (${cardinality} values)`);\n    } else if (cardinality === total && total > 10) {\n      parts.push('Unique identifier');\n    } else {\n      parts.push('Text');\n    }\n  } else {\n    parts.push(type);\n  }\n\n  // Cardinality insight\n  if (cardinality === 1 && total > 1) {\n    parts.push('constant');\n  } else if (total > 0 && cardinality / total > 0.95 && total > 10) {\n    parts.push('near-unique');\n  }\n\n  // Null info\n  if (nullPct > 0.5) {\n    parts.push(`mostly null (${(nullPct * 100).toFixed(0)}%)`);\n  } else if (nullPct > 0.1) {\n    parts.push(`some nulls (${(nullPct * 100).toFixed(0)}%)`);\n  }\n\n  return parts.join(', ');\n}\n"],
+  "mappings": ";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AA4BA,SAAS,WAAW,OAAoB;AACtC,MAAI,UAAU,QAAQ,UAAU,OAAW,QAAO;AAClD,MAAI,OAAO,UAAU,UAAW,QAAO;AACvC,MAAI,OAAO,UAAU,UAAU;AAC7B,WAAO,OAAO,UAAU,KAAK,IAAI,YAAY;AAAA,EAC/C;AACA,MAAI,OAAO,UAAU,UAAU;AAE7B,QAAI,qBAAqB,KAAK,KAAK,KAAK,CAAC,MAAM,KAAK,MAAM,KAAK,CAAC,EAAG,QAAO;AAC1E,WAAO;AAAA,EACT;AACA,SAAO;AACT;AAEA,SAAS,eAAe,OAA4B,YAAkC;AAEpF,QAAM,UAAU,IAAI,IAAI,KAAK;AAC7B,UAAQ,OAAO,MAAM;AAErB,MAAI,QAAQ,SAAS,EAAG,QAAO;AAC/B,MAAI,QAAQ,SAAS,EAAG,QAAO,CAAC,GAAG,QAAQ,KAAK,CAAC,EAAE,CAAC;AAGpD,MAAI,QAAQ,SAAS,KAAK,QAAQ,IAAI,SAAS,KAAK,QAAQ,IAAI,OAAO,EAAG,QAAO;AAGjF,aAAW,CAAC,MAAM,KAAK,KAAK,SAAS;AACnC,QAAI,QAAQ,aAAa,IAAK,QAAO;AAAA,EACvC;AAEA,SAAO;AACT;AAMO,MAAM,eAAe;AAAA,EAArB;AACL,SAAQ,QAAQ,oBAAI,IAAyB;AAC7C,SAAQ,cAAwB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA,EAKjC,SAAS,MAAmB,SAAyB;AAEnD,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,YAAM,MAAM,QAAQ,CAAC;AACrB,UAAI,CAAC,KAAK,MAAM,IAAI,GAAG,GAAG;AACxB,aAAK,YAAY,KAAK,GAAG;AACzB,aAAK,MAAM,IAAI,KAAK;AAAA,UAClB,MAAM;AAAA,UACN,UAAU,KAAK,YAAY,SAAS;AAAA,UACpC,OAAO,oBAAI,IAAI;AAAA,UACf,WAAW;AAAA,UACX,YAAY;AAAA,UACZ,gBAAgB,oBAAI,IAAI;AAAA,UACxB,YAAY;AAAA,UACZ,cAAc;AAAA,UACd,UAAU;AAAA,UACV,UAAU;AAAA,UACV,SAAS,CAAC;AAAA,QACZ,CAAC;AAAA,MACH;AAAA,IACF;AAGA,eAAW,OAAO,MAAM;AACtB,iBAAW,OAAO,SAAS;AACzB,cAAM,OAAO,KAAK,MAAM,IAAI,GAAG;AAC/B,cAAM,QAAQ,IAAI,GAAG;AACrB,aAAK;AAEL,YAAI,UAAU,QAAQ,UAAU,QAAW;AACzC,eAAK;AACL,eAAK,MAAM,IAAI,SAAS,KAAK,MAAM,IAAI,MAAM,KAAK,KAAK,CAAC;AACxD;AAAA,QACF;AAEA,cAAM,OAAO,WAAW,KAAK;AAC7B,aAAK,MAAM,IAAI,OAAO,KAAK,MAAM,IAAI,IAAI,KAAK,KAAK,CAAC;AAGpD,cAAM,SAAS,OAAO,KAAK;AAC3B,YAAI,KAAK,eAAe,OAAO,KAAQ;AACrC,eAAK,eAAe,IAAI,MAAM;AAAA,QAChC;AAGA,YAAI,OAAO,UAAU,YAAY,CAAC,MAAM,KAAK,GAAG;AAC9C,eAAK,cAAc;AACnB,eAAK;AACL,gBAAM,SAAS,OAAO,KAAK;AAC3B,cAAI,KAAK,aAAa,QAAQ,QAAQ,WAAW,KAAK,QAAQ,EAAG,MAAK,WAAW;AACjF,cAAI,KAAK,aAAa,QAAQ,QAAQ,WAAW,KAAK,QAAQ,EAAG,MAAK,WAAW;AAAA,QACnF,WAAW,SAAS,YAAY,SAAS,QAAQ;AAC/C,cAAI,KAAK,aAAa,QAAQ,SAAS,KAAK,SAAU,MAAK,WAAW;AACtE,cAAI,KAAK,aAAa,QAAQ,SAAS,KAAK,SAAU,MAAK,WAAW;AAAA,QACxE;AAGA,YAAI,KAAK,QAAQ,SAAS,KAAK,CAAC,KAAK,QAAQ,SAAS,MAAM,KAAK,OAAO,UAAU,KAAK;AACrF,eAAK,QAAQ,KAAK,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,SAAS,WAAiC;AACxC,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,WAAyB,CAAC;AAEhC,eAAW,OAAO,KAAK,aAAa;AAClC,YAAM,OAAO,KAAK,MAAM,IAAI,GAAG;AAC/B,YAAM,eAAe,eAAe,KAAK,OAAO,KAAK,UAAU;AAC/D,YAAM,UAAU,KAAK,aAAa,IAAI,KAAK,YAAY,KAAK,aAAa;AACzE,YAAM,YAAY,KAAK,eAAe,IAAI,KAAK,aAAa,KAAK,eAAe;AAEhF,eAAS,KAAK;AAAA,QACZ,IAAI,GAAG,SAAS,KAAK,GAAG;AAAA,QACxB;AAAA,QACA,MAAM;AAAA,QACN,UAAU,KAAK;AAAA,QACf;AAAA,QACA,UAAU,KAAK,YAAY;AAAA,QAC3B,WAAW,KAAK;AAAA,QAChB;AAAA,QACA,aAAa,KAAK,eAAe;AAAA,QACjC,UAAU,KAAK;AAAA,QACf,UAAU,KAAK;AAAA,QACf;AAAA,QACA,cAAc,KAAK;AAAA,QACnB,SAAS,sBAAsB,KAAK,cAAc,MAAM,OAAO;AAAA,QAC/D,WAAW;AAAA,MACb,CAAC;AAAA,IACH;AAEA,WAAO;AAAA,EACT;AACF;AAKA,SAAS,sBACP,MACA,MACA,MACA,SACQ;AACR,QAAM,QAAkB,CAAC;AACzB,QAAM,cAAc,KAAK,eAAe;AACxC,QAAM,QAAQ,KAAK;AAGnB,MAAI,SAAS,aAAa,SAAS,SAAS;AAC1C,UAAM,KAAK,YAAY,IAAI,GAAG;AAC9B,QAAI,KAAK,YAAY,QAAQ,KAAK,YAAY,MAAM;AAClD,YAAM,KAAK,SAAS,KAAK,QAAQ,SAAI,KAAK,QAAQ,EAAE;AAAA,IACtD;AAAA,EACF,WAAW,SAAS,WAAW;AAC7B,UAAM,KAAK,cAAc;AAAA,EAC3B,WAAW,SAAS,QAAQ;AAC1B,UAAM,KAAK,gBAAgB;AAC3B,QAAI,KAAK,YAAY,KAAK,UAAU;AAClC,YAAM,KAAK,QAAQ,KAAK,QAAQ,OAAO,KAAK,QAAQ,EAAE;AAAA,IACxD;AAAA,EACF,WAAW,SAAS,UAAU;AAC5B,QAAI,eAAe,MAAM,QAAQ,IAAI;AACnC,YAAM,KAAK,gBAAgB,WAAW,UAAU;AAAA,IAClD,WAAW,gBAAgB,SAAS,QAAQ,IAAI;AAC9C,YAAM,KAAK,mBAAmB;AAAA,IAChC,OAAO;AACL,YAAM,KAAK,MAAM;AAAA,IACnB;AAAA,EACF,OAAO;AACL,UAAM,KAAK,IAAI;AAAA,EACjB;AAGA,MAAI,gBAAgB,KAAK,QAAQ,GAAG;AAClC,UAAM,KAAK,UAAU;AAAA,EACvB,WAAW,QAAQ,KAAK,cAAc,QAAQ,QAAQ,QAAQ,IAAI;AAChE,UAAM,KAAK,aAAa;AAAA,EAC1B;AAGA,MAAI,UAAU,KAAK;AACjB,UAAM,KAAK,iBAAiB,UAAU,KAAK,QAAQ,CAAC,CAAC,IAAI;AAAA,EAC3D,WAAW,UAAU,KAAK;AACxB,UAAM,KAAK,gBAAgB,UAAU,KAAK,QAAQ,CAAC,CAAC,IAAI;AAAA,EAC1D;AAEA,SAAO,MAAM,KAAK,IAAI;AACxB;",
+  "names": []
+}