nb

Render Jupyter Notebooks in pure Go π
This package is inspired by @yuin's goldmark and is designed to be as clear and extensible.
The implementation follows the official Jupyter Notebook format spec (nbformat) and produces an output similar to that of nbconvert (Jupyter's team own reference implementation) both structurally and visually.
Although the current release only supports v4.4 notebooks, support for other formats will be added soon (see the Roadmap).
The package comes with an HTML renderer out of the box and can be extended to convert notebooks to other formats, such as LaTeX or PDF.
π This package is being actively developed: its structure and APIs might change overtime.
If you find any bugs, please consider opening an issue or submitting a PR.
Installation
go get github.com/bevzzz/nb
Usage
nb's default, no-frills converter can render markdown, code, and raw cells out of the box:
b, err := os.ReadFile("notebook.ipynb")
if err != nil {
panic(err)
}
err := nb.Convert(os.Stdout, b)
To produce richer output nb relies on a flexible extension API and a collection of built-in adapters and standalone extensions that allow using other packages to render parts of the notebook:
import (
"github.com/bevzzz/nb"
synth "github.com/bevzzz/nb-synth"
"github.com/bevzzz/nb/extension"
"github.com/bevzzz/nb/extension/adapter"
jupyter "github.com/nb/extension/extra/goldmark-jupyter"
"github.com/robert-nix/ansihtml"
"github.com/yuin/goldmark"
highlighting "github.com/yuin/goldmark-highlighting/v2"
)
md := goldmark.New(
goldmark.WithExtensions(
jupyter.Attachments(),
highlighting.Highlighting,
),
)
c := nb.New(
nb.WithExtensions(
jupyter.Goldmark(md),
synth.Highlighting,
extension.NewStream(
adapter.AnsiHtml(ansihtml.ConvertToHTML),
),
),
)
err := nb.Convert(os.Stdout, b)
The snippet above uses these additional dependencies which I encourage you to check out:
Extending nb does not end here. Your project may already use a different Markdown renderer, or require custom handling of certain mime-/cell types, in which case I hope the existing extensions will serve as useful reference implementations.
Styling the notebook: batteries included π
nb comes with the Jupyter's classic light theme, which you can capture by passing a dedicated CSSWriter and adding it to the final HTML.
Mind that the default theme is ~1000 lines long and might not fit the existing style in a more complex project.
In that case you probably want to write your own CSS.
Click to expand
// Write both CSS and notebook's HTML to intermediate destinations
var body, css bytes.Buffer
// Configure your converter
c := nb.New(
nb.WithRenderOptions(
render.WithCellRenderers(
html.NewRenderer(
html.WithCSSWriter(&css),
),
),
),
)
err := c.Convert(&body, b)
if err != nil {
panic(err)
}
// Create the final output
f, _ := os.OpenFile("notebook.html", os.O_RDWR, 0644)
defer f.Close()
f.WriteString("<html><head><style>")
io.Copy(f, &css)
f.WriteString("</style></head>")
f.WriteString("<body>")
io.Copy(f, &body)
f.WriteString("</body></html>")
Roadmap πΊ
- v0.3.0:
- Support for older
nbformat versions:
- Custom CSS (class prefix / class names).
I really the way
chroma exposes its styling API and I'll try to do something similar.
- v0.4.0:
- Support for
v1.0 notebooks
- Built-in pretty-printing for JSON outputs
- Other:
- I am curious about how
nb's performance measures against other popular libraries like nbconvert (Python) and quarto (Javascript), so I want to do some benchmarking later.
- As of now, I am not planning on adding converters to other formats (LaTeX, PDF, reStructuredText), but I will gladly consider this if there's a need for those.
If you have any other ideas or requests, please feel welcome to add a proposal in a new issue.
Miscellaneous
Math
Since Jupyter notebooks are often used for scientific work, you may want to display mathematical notation in your output.
MathJax is a powerful tool for that and adding it to your HTML header is the simplest way to get started.
Notice, that we want to remove <pre> from the the list of skipped tags, as default HTML renderer will wrap raw and markdown cells in a <pre> tag.
<html>
<head>
<script>
MathJax = {
options: {
skipHtmlTags: [ // includes "pre" by default
"script",
"noscript",
"style",
"textarea",
"code",
"annotation",
"annotation-xml"
],
}
};
</script>
</head>
</html>
MathJax is very configurable and you can read more about that here.
You may also find the official MathJax config used in the Jupyter project useful.
License
This software is released under the MIT License.