Use case

1. Stack Exchange archives: https://archive.org/details/stackexchange
2. Wikipedia dumps.
3. OpenStreetMap dumps.

Describe the solution you'd like

Add support for XML as input format.

The user should specify one of the three flavours with a setting:

1. elements:

<table>
    <row>
        <column1_name>value</column1_name>
        <column2_name>value</column2_name>
        ...
    </row>
    ...
</table>

2. attributes

<table>
    <row column1_name="value" column2_name="value" ... />
    ...
</table>

3. cells

<table>
    <row>
        <cell name="column1_name">value</cell>
        <cell name="column2_name">value</cell>
        ...
    </row>
    ...
</table>

and the settings with the path to the table (like /table),
the name of row element,
the name of the cell element and the name of the attribute with column name (for cells variant).

The format should not use full-featured XML parser, but should support:

• decoding of XML entities;
• optional BOM at the beginning of the file, including UTF-8 BOM;
• UTF16 and UTF-32 BE/LE encodings;
• CDATA;
• attributes in single and double quotes;
• self-closing tags or separate closing tags;
• skipping XML header and DOCTYPE;
• processing of invalid (unescaped) charaters;

Additional context

Let these settings also control XML output format (that we already have).