feat: html table extractor

extract/html-table/README.md

@@ -0,0 +1,61 @@
+<!-- START_INFOCARD -->
+
+# @flatfile/plugin-extract-html-table
+
+This plugin provides HTML table extraction capabilities for Flatfile. It parses HTML files and extracts structured data from tables, handling complex layouts and nested tables.
+
+**Event Type:** `listener.on('file:created')`
+
+**Supported File Types:** `.html`
+
+<!-- END_INFOCARD -->
+
+## Features
+
+- Extracts table structure, including headers and cell data
+- Handles nested tables and complex table layouts
+- Handles colspan and rowspan attributes (configurable)
+- Supports nested tables up to a configurable depth
+- Converts extracted data into a structured format
+- Provides error handling for malformed HTML or table structures
+- Debug mode for detailed logging
+
+## Parameters
+
+#### `options` - `object` - (optional)
+
+- `handleColspan` - `boolean` - (optional): Determines how to handle colspan. Default is true.
+- `handleRowspan` - `boolean` - (optional): Determines how to handle rowspan. Default is true.
+- `maxDepth` - `number` - (optional): Maximum depth for nested tables. Default is 3.
+- `debug` - `boolean` - (optional): Enables debug logging. Default is false.
+
+## API Calls
+
+- `api.files.download`
+- `api.files.update`
+
+## Usage
+
+**install**
+```bash
+npm install @flatfile/plugin-extract-html-table
+```
+
+**import**
+```javascript
+import { HTMLTableExtractor } from '@flatfile/plugin-extract-html-table';
+```
+
+**listener.js**
+```javascript
+const listener = new FlatfileListener();
+
+listener.use(
+  HTMLTableExtractor({
+    handleColspan: true,
+    handleRowspan: true,
+    maxDepth: 3,
+    debug: false
+  })
+);
+```

extract/html-table/jest.config.js

@@ -0,0 +1,16 @@
+module.exports = {
+  testEnvironment: 'node',
+
+  transform: {
+    '^.+\\.tsx?$': 'ts-jest',
+  },
+  setupFiles: ['../../test/dotenv-config.js'],
+  setupFilesAfterEnv: [
+    '../../test/betterConsoleLog.js',
+    '../../test/unit.cleanup.js',
+  ],
+  testTimeout: 60_000,
+  globalSetup: '../../test/setup-global.js',
+  forceExit: true,
+  passWithNoTests: true,
+}

extract/html-table/package.json

@@ -0,0 +1,55 @@
+{
+  "name": "@flatfile/plugin-extract-html-table",
+  "version": "1.0.0",
+  "description": "A Flatfile plugin for extracting table data from HTML files",
+  "main": "./dist/index.cjs",
+  "module": "./dist/index.mjs",
+  "types": "./dist/index.d.ts",
+  "browser": {
+    "./dist/index.cjs": "./dist/index.browser.cjs",
+    "./dist/index.mjs": "./dist/index.browser.mjs"
+  },
+  "exports": {
+    "types": "./dist/index.d.ts",
+    "node": {
+      "import": "./dist/index.mjs",
+      "require": "./dist/index.cjs"
+    },
+    "browser": {
+      "require": "./dist/index.browser.cjs",
+      "import": "./dist/index.browser.mjs"
+    },
+    "default": "./dist/index.mjs"
+  },
+  "source": "./src/index.ts",
+  "files": ["dist/**"],
+  "scripts": {
+    "build": "rollup -c",
+    "build:watch": "rollup -c --watch",
+    "build:prod": "NODE_ENV=production rollup -c",
+    "check": "tsc ./**/*.ts --noEmit --esModuleInterop",
+    "test": "jest src/*.spec.ts --detectOpenHandles",
+    "test:unit": "jest src/*.spec.ts --testPathIgnorePatterns=.*\\.e2e\\.spec\\.ts$ --detectOpenHandles",
+    "test:e2e": "jest src/*.e2e.spec.ts --detectOpenHandles"
+  },
+  "keywords": ["flatfile", "flatfile-plugins", "category-extractors", "html", "table-extractor"],
+  "author": "Flatfile",
+  "license": "ISC",
+  "dependencies": {
+    "@flatfile/util-extractor": "^2.1.5",
+    "node-html-parser": "^6.1.13"
+  },
+  "devDependencies": {
+    "@flatfile/rollup-config": "^0.1.1"
+  },
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/FlatFilers/flatfile-plugins.git",
+    "directory": "extract/html-table"
+  },
+  "browserslist": [
+    "> 0.5%",
+    "last 2 versions",
+    "not dead"
+  ]
+}

extract/html-table/rollup.config.mjs

@@ -0,0 +1,5 @@
+import { buildConfig } from '@flatfile/rollup-config';
+
+const config = buildConfig({});
+
+export default config;

extract/html-table/samples/complex_table.html

@@ -0,0 +1,51 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <title>Complex Table Example</title>
+</head>
+<body>
+    <h1>Quarterly Sales Report</h1>
+    <table border="1">
+        <tr>
+            <th rowspan="2">Product Category</th>
+            <th colspan="4">Quarterly Sales</th>
+            <th rowspan="2">Total</th>
+        </tr>
+        <tr>
+            <th>Q1</th>
+            <th>Q2</th>
+            <th>Q3</th>
+            <th>Q4</th>
+        </tr>
+        <tr>
+            <td>Electronics</td>
+            <td>$50,000</td>
+            <td>$65,000</td>
+            <td>$70,000</td>
+            <td>$90,000</td>
+            <td>$275,000</td>
+        </tr>
+        <tr>
+            <td>Clothing</td>
+            <td>$35,000</td>
+            <td>$40,000</td>
+            <td>$55,000</td>
+            <td>$80,000</td>
+            <td>$210,000</td>
+        </tr>
+        <tr>
+            <td>Home & Garden</td>
+            <td>$25,000</td>
+            <td>$30,000</td>
+            <td>$40,000</td>
+            <td>$45,000</td>
+            <td>$140,000</td>
+        </tr>
+        <tr>
+            <td colspan="5">Total Sales</td>
+            <td>$625,000</td>
+        </tr>
+    </table>
+</body>
+</html>

extract/html-table/samples/multiple_tables.html

@@ -0,0 +1,78 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <title>Multiple Tables Example</title>
+</head>
+<body>
+    <h1>Company Data</h1>
+
+    <h2>Employee Information</h2>
+    <table>
+        <tr>
+            <th>Name</th>
+            <th>Position</th>
+            <th>Department</th>
+        </tr>
+        <tr>
+            <td>John Doe</td>
+            <td>Software Engineer</td>
+            <td>IT</td>
+        </tr>
+        <tr>
+            <td>Jane Smith</td>
+            <td>Marketing Specialist</td>
+            <td>Marketing</td>
+        </tr>
+    </table>
+
+    <h2>Department Budget</h2>
+    <table>
+        <tr>
+            <th>Department</th>
+            <th>Budget</th>
+            <th>Expenses</th>
+        </tr>
+        <tr>
+            <td>IT</td>
+            <td>$500,000</td>
+            <td>
+                <table>
+                    <tr>
+                        <th>Category</th>
+                        <th>Amount</th>
+                    </tr>
+                    <tr>
+                        <td>Hardware</td>
+                        <td>$200,000</td>
+                    </tr>
+                    <tr>
+                        <td>Software</td>
+                        <td>$150,000</td>
+                    </tr>
+                </table>
+            </td>
+        </tr>
+        <tr>
+            <td>Marketing</td>
+            <td>$300,000</td>
+            <td>
+                <table>
+                    <tr>
+                        <th>Category</th>
+                        <th>Amount</th>
+                    </tr>
+                    <tr>
+                        <td>Advertising</td>
+                        <td>$150,000</td>
+                    </tr>
+                    <tr>
+                        <td>Events</td>
+                        <td>$100,000</td>
+                    </tr>
+                </table>
+            </td>
+        </tr>
+    </table>
+</body>
+</html>

extract/html-table/samples/simple_table.html

@@ -0,0 +1,32 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <title>Simple Table Example</title>
+</head>
+<body>
+    <h1>Employee Information</h1>
+    <table>
+        <tr>
+            <th>Name</th>
+            <th>Position</th>
+            <th>Department</th>
+        </tr>
+        <tr>
+            <td>John Doe</td>
+            <td>Software Engineer</td>
+            <td>IT</td>
+        </tr>
+        <tr>
+            <td>Jane Smith</td>
+            <td>Marketing Specialist</td>
+            <td>Marketing</td>
+        </tr>
+        <tr>
+            <td>Bob Johnson</td>
+            <td>HR Manager</td>
+            <td>Human Resources</td>
+        </tr>
+    </table>
+</body>
+</html>

extract/html-table/src/index.ts

@@ -0,0 +1,15 @@
+import { Extractor } from '@flatfile/util-extractor'
+import { parseBuffer } from './parser'
+
+export interface HTMLTableExtractorOptions {
+  handleColspan?: boolean
+  handleRowspan?: boolean
+  maxDepth?: number
+  debug?: boolean
+}
+
+export const HTMLTableExtractor = (options: HTMLTableExtractorOptions = {}) => {
+  return Extractor('html', 'html-tables', parseBuffer, options)
+}
+
+export const htmlTableParser = parseBuffer