What Pasteful tests
Pasteful's formatter is checked against 2,954 deterministic records generated from 27 source documents. The corpus covers every source and destination currently catalogued by the shared formatting engine, then exercises the structures most likely to break during a copy: headings, nested lists, links, tables, code, block quotes, emoji, right-to-left text, invisible characters and malformed Markdown.
Every run produces ordered JSON Lines output. The checked-in output is hashed, so an intentional formatting change must update both the visible records and their SHA-256 baseline. A change that only appears on one JavaScript engine fails the parity contract instead of quietly reaching a user.
- 2,954 machine-readable records
- 27 stored and synthetic documents
- Every catalogued source and destination
- A generated document larger than 100 KB
- Byte-for-byte output with a reviewed SHA-256 baseline
Download the current baseline
The full JSONL file is published for developers, reviewers and researchers who want to inspect a result rather than rely on a marketing claim. A smaller JSON summary gives the record count, document count, format version and current checksum.
The current corpus checksum is c7d04baad993563969d6b714a2656073c912dae38f135e2532107c21edc857ee. The public build verifies that checksum before it publishes either download, so the summary and the records cannot drift apart unnoticed.
Download the complete JSONL corpus · Open the corpus summary
What the corpus does not prove
This corpus verifies Pasteful's deterministic parsing, cleanup and destination-output rules. It does not prove that every version of Word, Slack, WhatsApp or another receiving app will make the same clipboard choice. Desktop, web and mobile clients can accept different representations, and workspace settings can change the final paste.
A real compatibility result therefore needs a named destination client, operating system, version and test date. Pasteful's preview is useful evidence about the output it prepared; the receiving composer remains the final authority on what another person will see.
How to reproduce a destination test
Use one fixture that contains a heading, bold and italic text, a labelled link, nested bullets, a numbered list, a quote, inline code, a fenced code block and a small table. Paste it directly into the destination, record what survives, then run the same fixture through Pasteful and repeat the paste. Keep the source wording identical so the comparison measures formatting rather than editing.
Record the clipboard representations, destination client, operating system, app version and date. Check whether headings are semantic, links retain their destination, list depth survives, table values remain associated with their labels and code stays separate from prose. Redact personal material before sharing screenshots or results.
- Keep the fixture and wording identical
- Record client, operating system, version and date
- Compare direct paste with the Pasteful-prepared result
- Check meaning and structure, not only visual similarity
- Publish limitations beside successful results