pkra
commented
3 years ago Example with linkedom
const {DOMParser, parseHTML} = require('linkedom');
let recurseTheDom = {};
// for npx usage
try {
recurseTheDom = require(require.resolve('ams-xml-to-html')+'/lib/recurseTheDom');
} catch (e) {
recurseTheDom = require('./lib/recurseTheDom');
}
const setHead = (xmldoc, htmldoc) => {
// TODO not in xslt not for articles but added by ams-html; change after switch to JS
// add viewport meta tag
const viewportmeta = htmldoc.createElement('meta');
viewportmeta.setAttribute('name', 'viewport');
viewportmeta.setAttribute('content', 'width=device-width');
const isBook = xmldoc.firstElementChild.tagName === 'book'; // TODO extract into property or function?
if (isBook) htmldoc.head.insertAdjacentElement('afterbegin', viewportmeta);
// add charset meta tag
const charset = htmldoc.createElement('meta');
// TODO switch to modern charset and remove xslt matching code below
// charset.setAttribute('charset', 'utf-8');
// matches xslt
charset.setAttribute('http-equiv', 'Content-Type');
charset.setAttribute('content', 'text/html; charset=utf-8');
htmldoc.querySelector('head').insertAdjacentElement('afterbegin', charset);
// set title
const xmlTitle =
xmldoc.querySelector('front>article-meta>title-group>alt-title') ||
xmldoc.querySelector(
'book-meta>book-title-group>book-title, front>article-meta>title-group>article-title'
);
htmldoc.title = xmlTitle ? xmlTitle.textContent : 'AMS Publication';
const root = xmldoc.querySelector('article, book');
const lang = root.getAttribute('xml:lang') || 'en';
htmldoc.querySelector('html').setAttribute('lang', lang);
};
const xml2html = xmlstring => {
const xml = (new DOMParser).parseFromString(xmlstring, 'text/xml' ).defaultView;
const xmldoc = xml.window.document;
const html = (new DOMParser).parseFromString('<!doctype html> <html lang="en"><head><title></title></head><body></body> </html>', 'text/html').defaultView;
const htmldoc = html.window.document;
setHead(xmldoc, htmldoc);
const root = xmldoc.querySelector('book, article');
recurseTheDom(htmldoc)(htmldoc.body, root);
return html;
};
module.exports = xml2html;
if (require.main === module) {
const fs = require('fs');
const path = require('path');
const inputFileName = process.argv[2];
const outputFileName = process.argv[3];
const inputString = fs.readFileSync(path.resolve(inputFileName));
const dom = xml2html(inputString);
fs.writeFileSync(
path.resolve(outputFileName),
'<!DOCTYPE html>\n' + dom.document.toString()
);
}This passes all but 11 tests, most of which are about subtitle handling (the rest are false positives due to change in attribute order at serialization).
- no document.head => querySelector
This is really a "whole toolchain" issue as jsdom is used throughout the toolchain. But we have to start here anyway.
With the recent move to MathJax v3, we're experiencing memory issues - we generally have to bump node's available memory to 4GB but even then some papers (!) will run out of memory. jsdom is known to be memory intensive and especially bad at allowing for garbage collection (cf. https://github.com/AmerMathSoc/ams-bookhtml/issues/131).
Some alternative are: