🕸 🕷 puppeteer-table-parser

Library to make parsing website tables much easier! When you are using puppeteer for scrapping websites and web application, you will find out that parsing tables consistently is not that easy. This library brings you abstraction between puppeteer and page context.

This library solves the following issues:

• ✨ Parsing columns by their name.
• ✨ Respect the defined order of columns.
• ✨ Appending custom columns with custom data.
• ✨ Custom sanitization of data in cells.
• ✨ Group and Aggregate data by your own function.
• ✨ Merge data from two independent tables into one structure.
• ✨ Handles invalid HTML structure.
• ✨ Retrieve results as CSV or array of plain JS objects.
• ✨ And much more!

Installation

yarn add puppeteer-table-parser

npm install puppeteer-table-parser

// CommonJS
const { tableParser } = require('puppeteer-table-parser')

// ESM / Typescript
import { tableParser } from 'puppeteer-table-parser'

API

interface ParserSettings {
  selector: string; // CSS selector
  allowedColNames: Record<string, string>; // key = input name, value = output name)

  headerRowsSelector?: string | null; // (default: 'thead tr', null ignores table's header selection)
  headerRowsCellSelector?: string; // (default: 'td,th')
  bodyRowsSelector?: string;  // (default: 'tbody tr')
  bodyRowsCellSelector?: string;  // (default: 'td')
  cellTextSource?: CellTextSource; // (default: 'innerText') see "Cell text source" below
  reverseTraversal?: boolean // (default: false)
  temporaryColNames?: string[]; // (default: []) 
  extraCols?: ExtraCol[]; // (default: [])
  withHeader?: boolean; // (default: true)
  csvSeparator?: string; // (default: ';')
  newLine?: string; // (default: '\n')
  rowValidationPolicy?: RowValidationPolicy; // (default: 'NON_EMPTY')
  groupBy?: {
    cols: string[];
    handler?: (rows: string[][], getColumnIndex: GetColumnIndexType) => string[];
  }
  rowValidator: (
    row: string[],
    getColumnIndex: GetColumnIndexType,
    rowIndex: number,
    rows: Readonly<string[][]>,
  ) => boolean;
  rowTransform?: (row: string[], getColumnIndex: GetColumnIndexType) => void;
  asArray?: boolean; // (default: false)
  rowValuesAsArray?: boolean; // (default: false)
  rowValuesAsObject?: boolean; // (default: false)
  colFilter?: (elText: string[], index: number) => string; // (default: (txt: string) => txt.join(' '))
  colParser?: (value: string, formattedIndex: number, getColumnIndex: GetColumnIndexType) => string; // (default: (txt: string) => txt.trim())
  optionalColNames?: string[]; // (default: [])
  excludedColumns?: (rows: string[][], getColumnIndex: GetColumnIndexType) => string[]; // (default: undefined)
};

Parsing workflow

1. Find table(s) by provided CSS selector.
2. Find associated columns by applying colFilter on their text and verify their count.
3. Filter rows based on rowValidationPolicy
4. Add extra columns specified in extraCols property in settings.
5. Run rowValidator function for every table row.
6. Run colParser for every cell in a row.
7. Run rowTransform function for each row.
8. Group results into buckets (groupBy.cols) property and pick the aggregated rows.
9. Add processed row to a temp array result.
10. Run excludedColumns and exclude retrieved columns from the rows and the header.
11. Add header row (if withHeader property is true).
12. Merge partial results and return them.

Performance: cell text source

By default each body cell is read with HTMLElement.innerText. innerText reflects the rendered text, but reading it forces the browser to perform a synchronous layout/reflow for every cell — which dominates parse time on large tables.

Setting cellTextSource: CellTextSource.TEXT_CONTENT switches the body-cell read to Node.textContent, which does not trigger layout. On large tables this is substantially faster (measured ~15–40% lower parse time on a 35 000 × 12 table, ~22% median), with identical output on whitespace-clean tables.

import { tableParser, CellTextSource } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: 'table',
  allowedColNames: { 'Car Name': 'car' },
  cellTextSource: CellTextSource.TEXT_CONTENT, // faster; see caveats below
})

When output can differ. textContent returns the raw source text, so it diverges from innerText on three cases (all of which survive the default .trim() colParser):

1. Interior whitespace runs — innerText collapses foo   bar to foo bar; textContent preserves the run.
2. <br> elements — innerText renders them as a newline; textContent emits nothing.
3. display:none descendants — innerText omits hidden text; textContent includes it.

Keep the default innerText if your cells contain any of the above and you rely on the rendered representation. For the common case of plain-text data cells, textContent is a safe, faster drop-in.

Examples

All data came from the HTML page, which you can find in test/assets/1.html.

Basic example (the simple table where we want to parse three columns without editing)

import { tableParser } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: 'table',
  allowedColNames: {
    'Car Name': 'car',
    'Horse Powers': 'hp',
    'Manufacture Year': 'year',
  },
});

car;hp;year
Audi S5;332;2015
Alfa Romeo Giulia;500;2020
BMW X3;215;2017
Skoda Octavia;120;2012

Basic example with custom column name parsing:

import { tableParser } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: 'table',
  colFilter: (value: string[]) => {
    return value.join(' ').replace(' ', '-').toLowerCase();
  },
  colParser: (value: string) => {
    return value.trim();
  },
  allowedColNames: {
    'car-name': 'car',
    'horse-powers': 'hp',
    'manufacture-year': 'year',
  },
})

car;hp;year
Audi S5;332;2015
Alfa Romeo Giulia;500;2020
BMW X3;215;2017
Skoda Octavia;120;2012

Basic example with row validation and using temporary column.

import { tableParser } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: 'table',
  allowedColNames: {
    'Car Name': 'car',
    'Manufacture Year': 'year',
    'Horse Powers': 'hp',
  },
  temporaryColNames: ['Horse Powers'],
  rowValidator: (row: string[], getColumnIndex) => {
    const powerIndex = getColumnIndex('hp');
    return Number(row[powerIndex]) < 250;
  },
});

car;year
BMW X3;2017
Skoda Octavia;2012

Advanced example:

Uses custom temporary column for filtering. It uses an extra column with custom position to be filled on a fly.

import { tableParser } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: 'table',
  allowedColNames: {
    'Manufacture Year': 'year',
    'Horse Powers': 'hp',
    'Car Name': 'car',
  },
  temporaryColNames: ['Horse Powers'],
  extraCols: [
    {
      colName: 'favorite',
      data: '',
      position: 0,
    },
  ],
  rowValidator: (row: string[], getColumnIndex) => {
    const horsePowerIndex = getColumnIndex('hp');
    return Number(row[horsePowerIndex]) > 150;
  },
  rowTransform: (row: string[], getColumnIndex) => {
    const nameIndex = getColumnIndex('car');
    const favoriteIndex = getColumnIndex('favorite');

    if (row[nameIndex].includes('Alfa Romeo')) {
      row[favoriteIndex] = 'YES';
    } else {
      row[favoriteIndex] = 'NO';
    }
  },
  asArray: false,
  rowValuesAsArray: false
});

favorite;year;car
NO;2015;Audi S5
YES;2020;Alfa Romeo Giulia
NO;2017;BMW X3

Optional columns

Sometimes you can be in a situation where some if your columns are desired, but they are not available in a table. You can easily add an exception for them via optionalColNames property.

import { tableParser } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: 'table',
  allowedColNames: {
    'Car Name': 'car',
    'Rating': 'rating',
  },
  optionalColNames: ['rating']
});

Grouping and Aggregating

import { tableParser } from 'puppeteer-table-parser'

await tableParser(page, {
  selector: '#my-table',
  allowedColNames: {
    'Employee Name': 'name',
    'Age': 'age',
  },
  groupBy: {
    cols: ['name'],
    handler: (rows: string[][], getColumnIndex) => {
      const ageIndex = getColumnIndex('age');

      // select one with the minimal age
      return rows.reduce((previous, current) =>
        previous[ageIndex] < current[ageIndex] ? previous : current,
      );
    },
  }
});

For more, look at the test folder! 🙈