Preview documents in the browser, without an office suite

A PDF, drawn from BDF

Converted ahead of time, as a server would. Turn its pages, select its text, search it, zoom.

742 KB as a PDF of 20 pages, 392 KB as BDF with its fonts. The pages are fetched by range requests as they are about to be shown.

And back out of BDF

The document above, demo.bdf, written back as a PDF file and as Markdown, and its text handed to a search index: each made inside your browser by the same Go code a server runs, built as WebAssembly. Nothing is uploaded.

bdf pdf -tagged demo.bdf demo-from-bdf.pdf

bdf markdown demo.bdf out/

bdf text demo.bdf demo-text.json

import MiniSearch from "minisearch";
import { normalizeQuery } from "@bdfkit/core"; // NFKC, case and kana folded, as the viewer's search

const index = new MiniSearch({
  fields: ["text", "title"], storeFields: ["page"],
  tokenize, // the words, and the pairs of neighboring characters of Japanese text (no dictionary needed)
  processTerm: (term) => normalizeQuery(term) || null,
});
// one record a page, from bdf text (Document.SearchText on a server)
index.addAll(text.views.flatMap((v) => v.pages.map((p) => ({ id: `${v.id}#${p.page}`, page: p.page, title: v.title, text: p.text }))));
const hits = index.search("canvas", { prefix: true, fuzzy: 0.2 }); // each hit knows its page

const CJK = /[\p{scx=Han}\p{scx=Hiragana}\p{scx=Katakana}\p{scx=Hangul}]+/gu;
function tokenize(text) {
  const out = [];
  for (const word of text.normalize("NFKC").split(/[\s\p{P}\p{S}]+/u)) {
    let at = 0;
    for (const m of word.matchAll(CJK)) {
      if (m.index > at) out.push(word.slice(at, m.index));
      const run = [...m[0]];
      if (run.length === 1) out.push(run[0]);
      for (let i = 1; i < run.length; i++) out.push(run[i - 1] + run[i]);
      at = m.index + m[0].length;
    }
    if (at < word.length) out.push(word.slice(at));
  }
  return out;
}

Each kind of document, in its own shape

A sheet stays a sheet that scrolls; a book turns its pages. All of them are drawn by the same renderer.

• 

  PDF

  Pages with their fonts, text to select and the structure of tagged PDF. Illustrator files too.

  Open in the viewer About the format

• 

  EPUB

  Books too, with pages that turn: reflowable ones in vertical Japanese, and fixed layouts.

  Open in the viewer About the format

• 

  PowerPoint

  Slides with their masters, charts, SmartArt and formulas.

  Open in the viewer About the format

• 

  Excel

  Sheets that scroll without page breaks, with frozen panes, headers and tabs. CSV, TSV and Parquet files are shown the same way.

  Open in the viewer About the format

• 

  Word

  Pages, or one long column to scroll. Vertical Japanese text too.

  Open in the viewer About the format

• 

  draw.io

  Pages as tabs, with today's AWS icons. Visio drawings too.

  Open in the viewer About the format

• 

  DXF

  Model space and layouts. Jw_cad, SXF, CGM and HP-GL/2 too.

  Open in the viewer About the format

• 

  KiCad

  Schematic sheets and boards. Gerber and Excellon files too.

  Open in the viewer About the format

• 

  Scores

  MML, MIDI and MusicXML, engraved as scores that the viewer plays.

  Open in the viewer About the format

• 

  Audio

  MP3, M4A, FLAC, Ogg, WAV and AIFF: the cover art, tags and lyrics on a card that plays, the line being sung marked.

  Open in the viewer About the format

29 formats in all, among them Visio, Illustrator, Photoshop, TIFF, HTML, Markdown and font files: all formats

Try it with your files

Files are converted inside your browser. They are not uploaded.

• Viewer

  Drop a file to view it: search, text selection, pages that turn, scores that play.

• Forms

  A Word, Excel or PowerPoint template with tags, a form made of them, and the filled form as a PDF file.

• Thumbnails

  Thumbnails at four sizes at once, drawn without a browser engine, as on a server.

• Search text

  The text and metadata of a document, page by page, for a search index.

Fit it into your system

Three sample projects, each a small Go server with its pages.

• Convert in the browser

  For documents made in large numbers but seldom opened, such as records kept for audits. The server only makes thumbnails; a file is converted in the browser of whoever opens it.

  Read more

• Convert on the server

  The server makes the preview and the thumbnail once. Browsers only draw.

  Read more

• With a search engine

  The text goes to a search engine page by page. A hit opens the document at its page.

  Read more

Why BDF

• No office suite to run

  No headless LibreOffice, no cgo, no external program: the converters are Go packages in one binary.

• In the shape of the content

  Fixed pages for slides and drawings, endless sheets for workbooks, pages or one column for text.

• Made for browsers

  Instructions map onto Canvas 2D. Fonts, images and compression are left to the browser.

• Many formats, one renderer

  Search, text selection and the text for screen readers work the same for every format.

• Only the converters you need

  Each format is a Go package of its own: link in the ones your service needs, or the browser modules for those kinds of files.

• Find in page and screen readers

  The text lies over the picture as real text, so the browser's own find (Ctrl+F) finds it, and screen readers read its headings, lists and tables.

More about why