@markset-lang/render-downgrade

Renders a Markset tree as plain CommonMark: what a document becomes when it is published somewhere that has never heard of Markset.

npm i @markset-lang/render-downgrade
# or
pnpm add @markset-lang/render-downgrade

import { downgrade } from "@markset-lang/render-downgrade";

const { markdown, diagnostics } = downgrade(source);

renderDowngrade(tree) does the same from a tree you already parsed, and downgradeTree(tree) returns the transformed mdast without serializing it, for a pipeline that wants to keep going.

What each construct becomes

Every construct has a normative downgrade in spec §4, and the conformance suite pins each one:

| Construct | Downgrade | |---|---| | callout | A blockquote opening with **Note:** and the title | | card | Its title as a heading one level below the current section, then its body | | grid, steps | The list, unchanged | | columns | The columns in source order, one after another | | tabs | The headings and their content, unchanged | | metrics | The table, unchanged | | figure | The content, then the caption as an italic paragraph | | span | Its text, attributes dropped |

Attribute lines and specifiers are dropped. Frontmatter is kept. A directive that failed validation contributes its content without the fence, so nothing an author wrote is lost.

Why it exists

Markset's third invariant is that every construct degrades: to a defined CommonMark output, and to a readable naive rendering when raw source is pasted into a stock parser. This renderer is the first half of that contract, and it is deliberately simple — the reference implementation built it before the HTML renderer, because a construct that does not downgrade cleanly should not exist.

Serialization is fixed so the suite can pin output byte for byte: - bullets, * for emphasis and strong, fenced code, GFM tables padded to column width, and column alignment preserved.

Specification: https://markset.org/spec/. Source: https://github.com/markset-lang/markset.

MIT.