Matplotlib is configured to use a high DPI, and communicates the metadata about “CSS pixels” via Jupyter cell metadata.

{
  "data": {
    "image/png": "…",
    "text/plain": ["<Figure size 480x480 with 1 Axes>"]
  },
  "metadata": {
    "image/png": {
      "height": 398,
      "width": 390
    }
  },
  "output_type": "display_data"
}

It looks like myst-nb adds that metadata to the docutils image node, but somehow it doesn’t end up in the <img> tag.

Possible causes might be

that myst-nb has its own cell.metadata.myst metadata key and might read that one only, not the plain cell.metadata.width/height keys.
that the metadata is unitless and is probably read from the notebook as an int.

Maybe myst-nb adding px to unitless width or height would work? docutils says it wants a “length” or “percentage”, which maybe has to be a string with “px” or “%” at the end

images in tutorials are rendered too large if not constrained by page width #192

Description

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions