Editing Manual/Importing XML dumps (section)

==== On Python ==== <!--T:161-->
</translate>
<syntaxhighlight lang="python">
from pathlib import Path
from lxml import etree
import re
from pywikibot import Link, Site

""" The exported XML preserves local namespace names. To import into a project in another language,
 you need to convert them to canonical names. Otherwise the pages will be imported as originals
 without recognizing their namespace.
XML contains namespace numbers, but mapping them can be problematic. Because they can differ between projects.
 For example, in ru.wikisource NS Page is number 104, while in uk.wikisource it is 252.
To solve these, the Link() class from pywikibot is used.

The Link() required the language code and family of project. XML does not contain these as separate
 values. Instead, they are in another format, like `enwikisource` or 'commonswiki', which needs to be parsed.    
"""

xml_folder = 'PATH_TO_XML_DUMP'
for xml_file in Path(xml_folder).glob('*.xml'):
    root = etree.parse(xml_file).getroot()
    ns = {'': root.nsmap[None]}  # xml namespace

    # Parsing language code and family of the project
    dbname = root.find('.//siteinfo/dbname', ns).text
    m = re.search(r'^(\w+?)(wiki|wikisource|wikiquote|wikinews|wikibooks)$', dbname)
    if m:
        site = Site(code=m.group(1), fam=m.group(2))
    elif 'commons' in dbname:
        site = Site('commons')
    elif 'mediawikiwiki' in dbname:
        site = Site('mediawiki')
    else:
        print("Site didn't recognized.")
        break

    wiki = []
    for page in root.findall('.//page', ns):
        title_source = page.find('title', ns).text
        title = Link(title_source, site).ns_title().replace(' ', '_')
        revision_id = page.find('.//revision/id', ns).text
        print(f'{title}, revision id: {revision_id}')
        text = page.find('.//revision/text', ns).text
        wiki.append("\n'''%s'''\n%s\n" % (title, text))
    Path(f'out_{xml_file.stem}.wiki').write_text('\n'.join(wiki))
</syntaxhighlight>

<translate>