XML Formatter

XML is a text format for structured data, made of tags that nest inside each other. The same document can be laid out in many ways — but a single-line response from an API or a minified config from a build tool is genuinely hard to read. This tool turns a wall of text back into the indented, tag-per-line layout that XML is famous for, and checks along the way that the document is structurally well-formed — that its tags nest the way XML requires.

Formatting here means two things that are worth keeping separate. First, re-indentation: each element moves to its own line, nested deeper as its tags nest deeper. Second, validation: for markup to be reformatted at all it must first be structurally well-formed — every open tag closed, tags nested correctly, every attribute value in quotes, and exactly one root element. If any of that fails, the tool says so with a clear message rather than guessing at a layout that was never there. The validation is deliberately structural: it checks what indentation depends on, and everything else the layout keeps intact — text, entities, characters — passes through untouched.

It is also important to be precise about what formatting does not touch. XML text content — the actual words inside tags — is data, and it is never changed. The order of attributes on a tag is preserved as written, entity references like & stay the way they were typed, comments are kept, and CDATA sections (blocks of text with no markup at all) pass through byte-for-byte. The tool rewrites only the whitespace that sits between tags and is not part of the content. That distinction is the entire honesty of a formatter: it makes code readable, it does not invent anything.

The scanner walks the XML into tokens — open tags, close tags, self-closing tags, text, comments, CDATA, and processing instructions — and verifies structural well-formedness as it goes: every open tag is matched by a close tag of the same name in the correct nesting order, there is exactly one root element, attribute values are quoted, and no stray text appears outside the root. Then it re-emits the document, and two rules decide the layout.

Inside an element whose content is all markup — child elements, comments — each token starts on its own line at the chosen indentation, indented by its depth. Text that is only whitespace is formatting: it is replaced by that newline-and-indent, which is why the output derives from the indent you pick, not from whatever spacing the source had.

The moment an element contains real text — a word, a number, a CDATA block — that element's interior is emitted byte-for-byte. Its text is data, so nothing inside is re-spaced or moved onto new lines. CDATA stays inside its own brackets, entity references keep their exact form, and an element that mixes text with child elements keeps that exact mix. Attribute order is always preserved, attribute quotes are copied rather than normalized, and names such as xsi: are treated as plain names.