Error Message and Stack Trace (if applicable)

No error raised. The child element is silently dropped.

Description

_root_to_dict in langchain_core/output_parsers/xml.py checks if root.text and returns early with {root.tag: root.text}. When an element has both text and children (mixed content), this early return discards all child elements.

The if root.text branch should only fire when there are zero children. When both text and children exist, text should be included alongside the children rather than replacing them.

System Info

TL;DR

The issue can be fixed by modifying the _root_to_dict function in langchain_core/output_parsers/xml.py to handle elements with both text and children.

Guidance

• The current implementation of _root_to_dict returns early when root.text is present, discarding child elements. This logic should be adjusted to only return early when there are no children.
• To verify the fix, run the provided reproduction steps and check that the output includes both the text and child elements.
• The modified function should check for the presence of children before deciding whether to return the text or a dictionary with text and children.
• Consider adding a test case to cover this scenario and prevent similar issues in the future.

Example

def _root_to_dict(root):
    if root.text and len(root) == 0:  # Check for text and no children
        return {root.tag: root.text}
    elif len(root) > 0:  # If there are children, include text alongside them
        result = {}
        if root.text:
            result['_text'] = root.text
        for child in root:
            if child.tag not in result:
                result[child.tag] = []
            result[child.tag].append(_root_to_dict(child))
        return {root.tag: result}
    else:
        return {root.tag: {}}

Notes

The provided example code snippet assumes that the desired output format is a dictionary with the text as a value under the _text key and the child elements as values under their respective tags.

Recommendation

Apply workaround: The issue can be resolved by modifying the _root_to_dict function as described above, allowing elements with both text and children to be parsed correctly.