Formatador XML

// Browsers ship a parser and a serialiser but no pretty printer. This walker
// indents only elements whose children are all elements: touching anything
// else would rewrite mixed content.
function indentXml(source, unit = '  ') {
  const doc = new DOMParser().parseFromString(source, 'application/xml');
  if (doc.querySelector('parsererror')) {
    throw new Error(doc.querySelector('parsererror').textContent.trim());
  }

  const walk = (el, depth) => {
    const kids = [...el.childNodes];
    const elementOnly =
      kids.some((n) => n.nodeType === 1) &&
      kids.every((n) => n.nodeType !== 3 || !n.nodeValue.trim());
    if (!elementOnly) return; // mixed or text-only: leave the subtree alone

    for (const n of kids) if (n.nodeType === 3) el.removeChild(n);
    for (const child of [...el.children]) {
      el.insertBefore(doc.createTextNode('\n' + unit.repeat(depth + 1)), child);
      walk(child, depth + 1);
    }
    el.appendChild(doc.createTextNode('\n' + unit.repeat(depth)));
  };

  walk(doc.documentElement, 0);
  return new XMLSerializer().serializeToString(doc);
}

// Browsers never resolve external entities, so XXE is not reachable here.
// Internal entity expansion is, so cap the input size before parsing.

# ElementTree.indent (3.9+) only adds whitespace where an element has no
# non-whitespace text, so mixed content survives. It does drop comments,
# because the default parser never builds them.
import xml.etree.ElementTree as ET
from defusedxml.ElementTree import fromstring

root = fromstring(source)          # safe: no entity expansion, no network
ET.indent(root, space='    ')      # four spaces
print(ET.tostring(root, encoding='unicode'))

# lxml keeps comments and processing instructions, and etree.indent applies
# the same mixed-content rule:
#
#   from lxml import etree
#   parser = etree.XMLParser(resolve_entities=False, no_network=True,
#                            load_dtd=False, huge_tree=False)
#   tree = etree.fromstring(source.encode(), parser)
#   etree.indent(tree, space='    ')
#   print(etree.tostring(tree, encoding='unicode'))
#
# Do not add remove_blank_text=True unless the document has a DTD. Without
# one, lxml guesses which blank text nodes are ignorable.

import javax.xml.XMLConstants;
import javax.xml.parsers.DocumentBuilderFactory;
import javax.xml.transform.*;
import javax.xml.transform.dom.DOMSource;
import javax.xml.transform.stream.StreamResult;
import javax.xml.xpath.*;
import org.w3c.dom.*;

DocumentBuilderFactory dbf = DocumentBuilderFactory.newInstance();
dbf.setFeature(XMLConstants.FEATURE_SECURE_PROCESSING, true);
dbf.setFeature("http://apache.org/xml/features/disallow-doctype-decl", true);
dbf.setXIncludeAware(false);
dbf.setExpandEntityReferences(false);
Document doc = dbf.newDocumentBuilder().parse(new java.io.File("in.xml"));

// The serialiser adds indentation on top of the whitespace already in the
// tree, so an already-indented file gets deeper on every run. Remove the
// blank text nodes first. The second predicate keeps the blanks that sit
// inside mixed content, where a sibling text node carries real characters.
XPath xpath = XPathFactory.newInstance().newXPath();
NodeList blanks = (NodeList) xpath.evaluate(
    "//text()[not(normalize-space())][not(../text()[normalize-space()])]",
    doc, XPathConstants.NODESET);
for (int i = 0; i < blanks.getLength(); i++) {
    Node n = blanks.item(i);
    n.getParentNode().removeChild(n);
}

TransformerFactory tf = TransformerFactory.newInstance();
tf.setFeature(XMLConstants.FEATURE_SECURE_PROCESSING, true);
Transformer t = tf.newTransformer();
t.setOutputProperty(OutputKeys.INDENT, "yes");
t.setOutputProperty("{http://xml.apache.org/xslt}indent-amount", "2");
t.transform(new DOMSource(doc), new StreamResult(System.out));

using System.Text;
using System.Xml;
using System.Xml.Linq;

// XDocument.Parse discards whitespace-only text nodes by default, which is
// what you want for element-only content. On mixed content it also removes
// the space in <p>a <b>x</b> <i>y</i></p>, so pass
// LoadOptions.PreserveWhitespace when the document carries prose.
var doc = XDocument.Parse(source);

// DtdProcessing is Prohibit by default for the reader XDocument builds, so
// external entities are never fetched. Say it out loud when you construct
// the reader yourself.
var settings = new XmlWriterSettings
{
    Indent = true,
    IndentChars = "    ",
    OmitXmlDeclaration = false,
};

var output = new StringBuilder();
using (var writer = XmlWriter.Create(output, settings))
{
    doc.Save(writer);
}
Console.WriteLine(output.ToString());

// XmlWriter stops indenting an element once character data has been written
// into it, so it will not reflow mixed content it is given.

<?php
$doc = new DOMDocument();

// Both flags must be set before loading. preserveWhiteSpace = false makes
// libxml2 drop blank text nodes; without a DTD it applies a heuristic, and
// that heuristic keeps blanks whose siblings carry real text, which is what
// protects mixed content. Run it on a copy and diff the first time.
$doc->preserveWhiteSpace = false;
$doc->formatOutput = true;

libxml_use_internal_errors(true);
if (!$doc->loadXML($source, LIBXML_NONET)) {
    foreach (libxml_get_errors() as $e) {
        fprintf(STDERR, "XML error at line %d, column %d: %s\n",
            $e->line, $e->column, trim($e->message));
    }
    libxml_clear_errors();
    exit(1);
}

echo $doc->saveXML();

# xmllint is part of libxml2 and is almost certainly already installed.
# --nonet stops it fetching a DTD the document references.
xmllint --format --nonet document.xml

# The indent unit comes from an environment variable, not a flag:
XMLLINT_INDENT='    ' xmllint --format --nonet document.xml

# Rewrite in place:
xmllint --format --nonet --output document.xml document.xml

# xmllint refuses to format a document that is not well-formed: it prints the
# first error and exits non-zero, leaving the output file untouched.
# libxml2 will not indent an element that has a text child, which is the same
# mixed-content rule this page applies.