html-to-text

Advanced converter that parses HTML and returns beautiful text.

Features

• Inline and block-level tags.
• Tables with colspans and rowspans.
• Links with both text and href.
• Word wrapping.
• Unicode support.
• Plenty of customization options.

Changelog

Available here: CHANGELOG.md

Version 6 contains a ton of changes, so it worth to take a look at the full changelog.

Version 7 contains an important change for custom formatters.

Version 8 brings the selectors support to greatly increase the flexibility but that also changes some things introduced in version 6. Base element(s) selection also got important changes.

Version 9 drops a lot of previously deprecated options, introduces some new formatters and new capabilities for custom formatters. Now a dual-mode package (cjs and esm). CLI is moved to a separate package.

Installation

npm install html-to-text

Usage

Convert a single document:

const { convert } = require('html-to-text');
// There is also an alias to `convert` called `htmlToText`.

const options = {
  wordwrap: 130,
  // ...
};
const html = '<div>Hello World</div>';
const text = convert(html, options);
console.log(text); // Hello World

Configure html-to-text once to convert many documents with the same options (recommended for good performance when processing big batches of documents):

const { compile } = require('html-to-text');

const options = {
  wordwrap: 130,
  // ...
};
const compiledConvert = compile(options); // options passed here

const htmls = [
  '<div>Hello World!</div>',
  '<div>こんにちは世界！</div>',
  '<div>Привіт Світ!</div>'
];
const texts = htmls.map(compiledConvert);
console.log(texts.join('\n'));
// Hello World!
// こんにちは世界！
// Привіт Світ!

Both convert and compiledConvert can take one more optional argument - metadata object that can be used by formatters.

Options

General options

Option | Default | Description ----------------------- | ------------ | ----------- baseElements | | Describes which parts of the input document have to be converted and present in the output text, and in what order. baseElements.selectors | ['body'] | Elements matching any of provided selectors will be processed and included in the output text, with all inner content.Refer to Supported selectors section below. baseElements.orderBy | 'selectors' | 'selectors' - arrange base elements in the same order as baseElements.selectors array;'occurrence' - arrange base elements in the order they are found in the input document. baseElements.returnDomByDefault | true | Convert the entire document if none of provided selectors match. decodeEntities | true | Decode HTML entities found in the input HTML if true. Otherwise preserve in output text. encodeCharacters | {} | A dictionary with characters that should be replaced in the output text and corresponding escape sequences. formatters | {} | An object with custom formatting functions for specific elements (see Override formatting section below). limits | | Describes how to limit the output text in case of large HTML documents. limits.ellipsis | '...' | A string to insert in place of skipped content. limits.maxBaseElements | undefined | Stop looking for more base elements after reaching this amount. Unlimited if undefined. limits.maxChildNodes | undefined | Maximum number of child nodes of a single node to be added to the output. Unlimited if undefined. limits.maxDepth | undefined | Stop looking for nodes to add to the output below this depth in the DOM tree. Unlimited if undefined. limits.maxInputLength | 16_777_216 | If the input string is longer than this value - it will be truncated and a message will be sent to stderr. Ellipsis is not used in this case. Unlimited if undefined. longWordSplit | | Describes how to wrap long words. longWordSplit.wrapCharacters | [] | An array containing the characters that may be wrapped on. Checked in order, search stops once line length requirement can be met. longWordSplit.forceWrapOnLimit | false | Break long words at the line length limit in case no better wrap opportunities found. preserveNewlines | false | By default, any newlines \n from the input HTML are collapsed into space as any other HTML whitespace characters. If true, these newlines will be preserved in the output. This is only useful when input HTML carries some plain text formatting instead of proper tags. selectors | [] | Describes how different HTML elements should be formatted. See Selectors section below. whitespaceCharacters | ' \t\r\n\f\u200b' | A string of characters that are recognized as HTML whitespace. Default value uses the set of characters defined in HTML4 standard. (It includes Zero-width space compared to living standard.) wordwrap | 80 | After how many chars a line break should follow.Set to null or false to disable word-wrapping.

Deprecated or removed options

Old option | Depr. | Rem. | Instead use -------------------------- | --- | ----- | ----------------- baseElement | 8.0 | | baseElements: { selectors: [ 'body' ] } decodeOptions | | 9.0 | Entity decoding is now handled by htmlparser2 itself and entities internally. No user-configurable parts compared to he besides boolean decodeEntities. format | | 6.0 | The way formatters are written has changed completely. New formatters have to be added to the formatters option, old ones can not be reused without rewrite. See new instructions below. hideLinkHrefIfSameAsText | 6.0 | 9.0 | selectors: [ { selector: 'a', options: { hideLinkHrefIfSameAsText: true } } ] ignoreHref | 6.0 | 9.0 | selectors: [ { selector: 'a', options: { ignoreHref: true } } ] ignoreImage | 6.0 | 9.0 | selectors: [ { selector: 'img', format: 'skip' } ] linkHrefBaseUrl | 6.0 | 9.0 | selectors: [{ selector: 'a', options: { baseUrl: 'https://example.com' } },{ selector: 'img', options: { baseUrl: 'https://example.com' } }] noAnchorUrl | 6.0 | 9.0 | selectors: [ { selector: 'a', options: { noAnchorUrl: true } } ] noLinkBrackets | 6.0 | 9.0 | selectors: [ { selector: 'a', options: { linkBrackets: false } } ] returnDomByDefault | 8.0 | | baseElements: { returnDomByDefault: true } singleNewLineParagraphs | 6.0 | 9.0 | selectors: [{ selector: 'p', options: { leadingLineBreaks: 1, trailingLineBreaks: 1 } },{ selector: 'pre', options: { leadingLineBreaks: 1, trailingLineBreaks: 1 } }] tables | 8.0 | | selectors: [ { selector: 'table.class#id', format: 'dataTable' } ] tags | 8.0 | | See Selectors section below. unorderedListItemPrefix | 6.0 | 9.0 | selectors: [ { selector: 'ul', options: { itemPrefix: ' * ' } } ] uppercaseHeadings | 6.0 | 9.0 | selectors: [{ selector: 'h1', options: { uppercase: false } },...{ selector: 'table', options: { uppercaseHeaderCells: false } }]

Other things removed:

• fromString method - use convert or htmlToText instead;
• positional arguments in BlockTextBuilder methods - pass option objects instead.

Selectors

Some example:

const { convert } = require('html-to-text');

const html = '<a href="/page.html">Page</a><a href="!#" class="button">Action</a>';
const text = convert(html, {
  selectors: [
    { selector: 'a', options: { baseUrl: 'https://example.com' } },
    { selector: 'a.button', format: 'skip' }
  ]
});
console.log(text); // Page [https://example.com/page.html]

Selectors array is our loose approximation of a stylesheet.

• highest specificity selector is used when there are multiple matches;
• the last selector is used when there are multiple matches of equal specificity;
• all entries with the same selector value are merged (recursively) at the compile stage, in such way so the last defined properties are kept and the relative order of unique selectors is kept;
• user-defined entries are appended after predefined entries;
• Every unique selector must have format value specified (at least once);
• unlike in CSS, values from different matched selectors are NOT merged at the convert stage. Single best match is used instead (that is the last one of those with highest specificity).

To achieve the best performance when checking each DOM element against provided selectors, they are compiled into a decision tree. But it is also important how you choose selectors. For example, div#id is much better than #id - the former will only check divs for the id while the latter has to check every element in the DOM.

Supported selectors

html-to-text relies on parseley and selderee packages for selectors support.

Following selectors can be used in any combinations:

• * - universal selector;
• div - tag name;
• .foo - class name;
• #bar - id;
• [baz] - attribute presence;
• [baz=buzz] - attribute value (with any operators and also quotes and case sensitivity modifiers - syntax);
• + and > combinators (other combinators are not supported).

You can match <p style="...; display:INLINE; ...">...</p> with p[style*="display:inline"i] for example.

Predefined formatters

Following selectors have a formatter specified as a part of the default configuration. Everything can be overridden, but you don't have to repeat the format or options that you don't want to override. (But keep in mind this is only true for the same selector. There is no connection between different selectors.)

Selector | Default format | Notes ------------- | ------------------- | ----- * | inline | Universal selector. a | anchor | article | block | aside | block | blockquote | blockquote | br | lineBreak | div | block | footer | block | form | block | h1 | heading | h2 | heading | h3 | heading | h4 | heading | h5 | heading | h6 | heading | header | block | hr | horizontalLine | img | image | main | block | nav | block | ol | orderedList | p | paragraph | pre | pre | table | table | Equivalent to block. Use dataTable instead for tabular data. ul | unorderedList | wbr | wbr |

More formatters also available for use:

Format | Description ---------------- | ----------- dataTable | For visually-accurate tables. Note that this might be not search-friendly (output text will look like gibberish to a machine when there is any wrapped cell contents) and also better to be avoided for tables used as a page layout tool. skip | Skips the given tag with it's contents without printing anything. blockString | Insert a block with the given string literal (formatOptions.string) instead of the tag. blockTag | Render an element as HTML block tag, convert it's contents to text. blockHtml | Render an element with all it's children as HTML block. inlineString | Insert the given string literal (formatOptions.string) inline instead of the tag. inlineSurround | Render inline element wrapped with given strings (formatOptions.prefix and formatOptions.suffix). inlineTag | Render an element as inline HTML tag, convert it's contents to text. inlineHtml | Render an element with all it's children as inline HTML.

Format options

Following options are available for built-in formatters.

Option | Default | Applies to | Description ------------------- | ----------- | ------------------ | ----------- leadingLineBreaks | 1, 2 or 3 | all block-level formatters | Number of line breaks to separate previous block from this one.Note that N+1 line breaks are needed to make N empty lines. trailingLineBreaks | 1 or 2 | all block-level formatters | Number of line breaks to separate this block from the next one.Note that N+1 line breaks are needed to make N empty lines. baseUrl | null | anchor, image | Server host for link href attributes and image src attributes relative to the root (the ones that start with /).For example, with baseUrl = 'http://asdf.com' and <a href='/dir/subdir'>...</a> the link in the text will be http://asdf.com/dir/subdir. linkBrackets | ['[', ']'] | anchor, image | Surround links with these brackets.Set to false or ['', ''] to disable. pathRewrite | undefined | anchor, image | A function to rewrite link href attributes and image src attributes. Optional second argument is the metadata object.Applied before baseUrl. hideLinkHrefIfSameAsText | false | anchor | By default links are translated in the following way:<a href='link'>text</a> => becomes => text [link].If this option is set to true and link and text are the same, [link] will be omitted and only text will be present. ignoreHref | false | anchor | Ignore all links. Only process internal text of anchor tags. noAnchorUrl | true | anchor | Ignore anchor links (where href='#...'). itemPrefix | ' * ' | unorderedList | String prefix for each list item. uppercase | true | heading | By default, headings (<h1>, <h2>, etc) are uppercased.Set this to false to leave headings as they are. length | undefined | horizontalLine | Length of the line. If undefined then wordwrap value is used. Falls back to 40 if that's also disabled. trimEmptyLines | true | blockquote | Trim empty lines from blockquote.While empty lines should be preserved in HTML, space-saving behavior is chosen as default for convenience. uppercaseHeaderCells | true | dataTable | By default, heading cells (<th>) are uppercased.Set this to false to leave heading cells as they are. maxColumnWidth | 60 | dataTable | Data table cell content will be wrapped to fit this width instead of global wordwrap limit.Set this to undefined in order to fall back to wordwrap limit. colSpacing | 3 | dataTable | Number of spaces between data table columns. rowSpacing | 0 | dataTable | Number of empty lines between data table rows. string | '' | blockString, inlineString | A string to be inserted in place of a tag. prefix | '' | inlineSurround | String prefix to be inserted before inline tag contents. suffix | '' | inlineSurround | String suffix to be inserted after inline tag contents.

Deprecated format options

Old option | Applies to | Depr. | Rem. | Instead use ------------------- | ------------------ | ----- | ---- | --------------------- noLinkBrackets | anchor | 8.1 | | linkBrackets: false

Override formatting

formatters option is an object that holds formatting functions. They can be assigned to format different elements in the selectors array.

Each formatter is a function of four arguments that returns nothing. Arguments are:

• elem - the HTML element to be processed by this formatter;
• walk - recursive function to process the children of this element. Called as walk(elem.children, builder);
• builder - BlockTextBuilder object. Manipulate this object state to build the output text;
• formatOptions - options that are specified for a tag, along with this formatter (Note: if you need general html-to-text options - they are accessible via builder.options).

Custom formatter example:

const { convert } = require('html-to-text');

const html = '<foo>Hello World</foo>';
const text = convert(html, {
  formatters: {
    // Create a formatter.
    'fooBlockFormatter': function (elem, walk, builder, formatOptions) {
      builder.openBlock({ leadingLineBreaks: formatOptions.leadingLineBreaks || 1 });
      walk(elem.children, builder);
      builder.addInline('!');
      builder.closeBlock({ trailingLineBreaks: formatOptions.trailingLineBreaks || 1 });
    }
  },
  selectors: [
    // Assign it to `foo` tags.
    {
      selector: 'foo',
      format: 'fooBlockFormatter',
      options: { leadingLineBreaks: 1, trailingLineBreaks: 1 }
    }
  ]
});
console.log(text); // Hello World!

New in version 9: metadata object can be provided as the last optional argument of the convert function (or the function returned by compile function). It can be accessed by formatters as builder.metadata.

Refer to generic formatters of the base package and text formatters of this package for more examples. The easiest way to write your own is to pick an existing one and customize.

Refer to BlockTextBuilder for available functions and arguments.

Custom metadata

If you need to supply extra information about your HTML documents to use in custom formatters - it can be done with the help of metadata object.

It is supplied as an extra argument to the convert function:

import { compile, convert } from 'html-to-text';

// for batch use:
const compiledConvert = compile(options);
let text = compiledConvert(html, metadata);

// for single use:
let text = convert(html, options, metadata);

And it can be accessed within formatter functions as builder.metadata.

Call other formatters from a custom formatter

Most of the times this is not what you actually need. Most practical problems can be solved with selectors.

If you really need to inspect the node internals, not just attributes, then you can do it like this:

const options = {
  // ...
  formatters: {
    filterBlockFormatter: function (elem, walk, builder, formatOptions) {
      // all built-in and custom formatters available by name
      const blockFormatter = builder.options.formatters['block'];
      if (blockFormatter && elem.children.some(/* predicate */)) {
        blockFormatter(elem, walk, builder, formatOptions);
      }
    }
  },
  selectors: [
    {
      selector: 'div.questionable',
      format: 'filterBlockFormatter',
      options: { leadingLineBreaks: 1, trailingLineBreaks: 1 }
    }
  ],
  // ...
}

Example

• Input text: test.html
• Output text: test.txt

Contributors

• @mlegenhausen - creator;
• @KillyMXI - maintainer since 2020;
• Everyone else who added something to the tool or helped us shaping it via issues and PRs.

License

MIT License