Do Carve's readers agree?

With the spec, with each other, and with the formats around them. Each claim below is a case someone can open, run again, and trace to the commit it was measured at.

One ownership case

Loading…

Carve source


        

What the readers produced


          
    Inspect this case

    Proofs

    Does the spec hold together, and do the readers follow it?

    Checked theorems
    –
    Ownership cases
    –
    Reader disagreements
    –

    Open the proofs

    Carve next to Markdown and Djot

    Two measured questions. Does a small edit stay small, compared with Djot and CommonMark on the same edits? And how much of a CommonMark or Djot document keeps its structure when imported into Carve?

    Each row is a fixed set of cases, not a ranking of the languages. Speed is measured separately in carve-bench.

    Edits that leave the rest of the document unchanged

    EditCarveDjotCommonMark

    Imported examples that keep their structure

    SourceKeptDiffers by designDiffersOut of scope

    Measured at these commits

    Current ownership and adapter results use the commits below. Behavior and scaling datasets retain their own pins and measurement dates. Where the two lanes measured different commits of the same reader, the row is marked.

    ReaderCurrent ownershipCompatibility

    What this does not show