Improve Markdown Paste-Import (Headings, Lists, Hierarchy, and Long Documents)

[kundeng]

Idea: Improve Markdown Paste-Import (Headings, Lists, Hierarchy, and Long Documents)

Summary

AFFiNE’s current behavior when pasting Markdown causes structured documents to collapse into incorrect or flat blocks. I tested this using a real Markdown document (multi-level headings, lists, horizontal rules, inline formatting, long multi-section content). The paste result did not preserve the structure: headings became plain text, lists became separate paragraphs, and section hierarchy was lost.

This makes it difficult to migrate notes from tools like Obsidian, Logseq, GitHub, or any Markdown-based knowledge workflow.

---

Reproducible Example

Here is a small excerpt from the Markdown I pasted:

# Introduction to NLP and Deep Learning (part 1/4)

## Course Overview
1. Course logistics and expectations.
2. Discussion on human language.
3. Introduction to the Word2Vec algorithm.

---

## Learning Goals
- Foundations of deep learning for NLP
- Linguistic understanding
- System building

Expected result:

• Heading block (H1)
• Heading block (H2)
• Ordered list block with three items
• Divider block
• Heading block (H2)
• Bullet list block with three items

Actual result:

• All content injected as plain paragraph blocks
• No list structure preserved
• No headings detected
• Horizontal rule ignored
• Deeply nested sections from the original document merged together

The full document I tested (a set of 4 lecture sections) contains many headings, lists, quote blocks, equations, and multi-paragraph sections—all of which flatten upon paste.

---

Why This Matters

Markdown is a primary format for:

• Technical notes
• Course materials
• Lecture summaries
• GitHub documentation
• PKM workflows (Obsidian, Logseq, Foam, Quartz)

Most users coming from Markdown-based systems expect paste-import to preserve structure. Without this, multi-thousand-word documents become unusable inside AFFiNE.

A reliable Markdown import path would dramatically improve adoption and migration.

---

Proposed Improvements

1. Auto-detect Markdown on paste

If clipboard text contains Markdown patterns (#, -, 1., ---, etc.), route it through a Markdown parser rather than plain-text insertion.

2. Parse via a full Markdown AST

Use a standard library such as remark, micromark, or markdown-it to produce an AST with nodes like:

• heading
• list / listItem
• paragraph
• blockquote
• thematicBreak
• code / inlineCode

3. Map AST nodes into BlockSuite blocks

Examples:

• heading(depth=1..6) → HeadingBlock(level=…)
• list(ordered=true/false) → ListBlock
• listItem(children) → Nested list blocks
• thematicBreak → DividerBlock
• blockquote → QuoteBlock

4. Preserve hierarchy and nesting

The sample I tested includes deeply nested lists and multi-level headings. These should map cleanly to nested blocks in AFFiNE.

5. Add a “Paste as Markdown (Convert to Blocks)” option

A context-menu override ensures predictable behavior.

---

Questions for the Maintainers

1. Is improved Markdown paste-import currently on the roadmap?
2. Would a proposal/PR implementing AST → block conversion be welcome?
3. Is there a preferred place in the existing codebase where Markdown detection and parsing should hook into the paste pipeline?
4. Should this target the MarkdownAdapter in BlockSuite, or should this be built as a separate adapter?

[cly2625]

+1