I have an idea to write a bot that can fix lint errors in all Wikimedia sites. Writing a bot from the beginning is very time-wasting. And need to be tested for a long time. Many other Wikimedia sites still have thousands of pages that need to fix the lint errors as fast as possible (High priority).
By chance, I notice that if we use visual editor parsoid API services, to convert the wikitext of a page to html, then using the same API to convert the html generated to wikitext again, it will fix most of lint errors inside the page. Since parsoid fix html errors when trying to convert to wikitextx using this API
The problem is that there is an option called "scrub_wikitext" that make some cosmetics, and normalizations to the wikitext generated. I tried to disable this option each time, but it keep making the normalizations.
Is there a way to convert from html to wikitext without making normalizations and in the same time fixing the html errors?. If this is possible, it can be very very helpful globally. Thanks.