#202 - Fix XmlReaderTask

[njoubert-cleverage]

Description

Fixes #202.

XmlReaderTask (through CleverAge\ProcessBundle\Filesystem\XmlFile::read()) does not handle bad input files:

• Empty file: the task crashes with a low-level ValueError: SplFileObject::fread(): Argument #1 ($length) must be greater than 0.
• Invalid XML: the result of DOMDocument::loadXML() is ignored. Only a PHP warning is raised, and an empty \DOMDocument (documentElement = null, no child nodes) is passed to the next tasks. In debug mode the Symfony error handler turns the warning into an exception, so the process fails; without debug (production), the process goes on silently with an empty document.

This PR:

• an empty file throws an explicit \UnexpectedValueException('XML file "<path>" is empty');
• loadXML() runs with libxml_use_internal_errors(true); the collected libxml errors are cleared and the previous setting is restored;
• if the load fails, or libxml reports an error or a fatal error, an \UnexpectedValueException('Invalid XML in file "<path>": <libxml message> (line L, column C); ...') is thrown; libxml warnings are still tolerated;
• update the XmlReaderTask reference documentation (Notes), add unit tests for XmlFile.

The new regression tests fail on main and pass with this fix. PHPUnit, PHPStan, PHP-CS-Fixer and Rector pass.

Requirements

• Documentation updates
  • Reference
  • Cookbooks
  • Changelog
• Unit tests

Breaking changes

Bug fix: XmlReaderTask now throws an \UnexpectedValueException when the XML is invalid, instead of passing an empty or partial \DOMDocument to the next tasks with only a PHP warning. This includes XML that parses but has libxml errors, such as an undefined namespace prefix (<root><x:a/></root>), which used to give a partial document. An empty file now throws an \UnexpectedValueException instead of a \ValueError. Processes relying on the silent empty document (e.g. in production, without the Symfony debug error handler) will now fail at the reader task: use error_strategy to handle these files.

🤖 Generated with Claude Code