Extensions: Tables & Strikethrough

CommonMark is deliberately small. It standardises the constructs everyone already agreed on and stops there — no tables, no strikethrough, no task lists. Everything beyond that is an extension, and knowing which layer a feature lives in is the difference between portable Markdown and Markdown that only renders on one site.

Three layers, not one

A parser that claims "GFM support" owes you exactly those five things. Anything in the third row is a promise made by a website, not by a specification — which is why the same document looks different on a wiki, in a chat client, and in a static site generator.

Strikethrough

~~deleted~~ becomes <del>deleted</del>. The GFM spec accepts a matching pair of one or two tildes, so ~deleted~ also works on GitHub; two is the portable form because plenty of CommonMark-only parsers implement neither.

Tables

| Name  | Age |
|:------|----:|
| Alice |  30 |

The header row is ordinary text until the delimiter row under it proves otherwise. That row is the disambiguator: without it, | a | b | is just a paragraph that happens to contain pipes. Colons in the delimiter cells set the alignment — leading colon left, trailing colon right, both center, neither default.

The trap that costs an afternoon

Inline transforms rewrite the string you are still scanning. Each pass can walk straight back into the HTML the previous pass produced:

Look at what the first line did: the emphasis pass matched an asterisk that was left over from the strong pass and one that came after a closing tag, so it wrapped </strong> inside an <em>. The tags are now interleaved and the HTML is invalid. Stashing each finished fragment behind a placeholder makes later passes structurally unable to see it.

Your exercise

Apply inline code, strong, em and strikethrough, in that order.

The mistake the grader catches is exactly the one above. A hidden test feeds ~~outer **inner *deepest*** end~~ and expects <del>outer <strong>inner *deepest</strong>* end</del> — with a literal asterisk surviving in the middle. Substituting in place produces interleaved <em> and </strong> tags and fails.

The second trap is running strikethrough before inline code: a test contains `~~code~~` next to real strikethrough on the same line, and the tildes inside the backticks must stay literal. Stash the code span first and the rest of the passes cannot reach it.