I noticed that the current approach to Markdown generation is that we first iterate through the Markdown and collect all the information from it, and then, in a new cycle, we iterate through the HTML and collect a small portion of the information from it.
Because of this, we run the risk of different behavior between the generation in GitHub and our own. I've definitely pinpointed the risk we have when generating the ToC, the order of headings, and the IDs in content-repeating headings.
Source github / npmx
With the current solution, we can handle this for HTML headings, but it will be after all other elements and incorrect.
It's important to loop through all elements in a single loop
One more example of partially missed logic - eslint for platinum sponsor link
I noticed that the current approach to Markdown generation is that we first iterate through the Markdown and collect all the information from it, and then, in a new cycle, we iterate through the HTML and collect a small portion of the information from it.
Because of this, we run the risk of different behavior between the generation in GitHub and our own. I've definitely pinpointed the risk we have when generating the ToC, the order of headings, and the IDs in content-repeating headings.
Source github / npmx
With the current solution, we can handle this for HTML headings, but it will be after all other elements and incorrect.
It's important to loop through all elements in a single loop
One more example of partially missed logic - eslint for platinum sponsor link