Rendering Jupyter notebook output

[darsnack]

In FastAI.jl, our examples/tutorials require training on a GPU. Running this code locally in a Jupyter notebook, then committing the notebook w/ output to the repo seems like the best solution. Currently, Publish.jl doesn't render the output of cells:

What would be required to render the output?

cc @lorenzoh

[MichaelHatherly]

Running this code locally in a Jupyter notebook, then committing the notebook w/ output to the repo seems like the best solution.

Yeah, that is probably the best approach for that use case.

I think it was probably just an oversight not rendering the output cells. Would probably need some changes to

Publish.jl/src/load.jl

Lines 124 to 146 in ad46c6c

	function loadfile(::Extension{:ipynb}, env, path)
	dict = open(JSON.parse, path)
	io = IOBuffer()
	if haskey(dict, "cells")
	for cell in dict["cells"]
	if haskey(cell, "cell_type")
	type = cell["cell_type"]
	source = get(cell, "source", "")
	if type == "markdown"
	join(io, source)
	println(io)
	elseif type == "code"
	println(io, CODE_FENCE, "julia")
	join(io, source)
	println(io)
	println(io, CODE_FENCE)
	end
	end
	end
	end
	parser = init_markdown_parser(env["publish"])
	return load_markdown(io, parser)
	end

to capture the cell output and then embed a suitable value within the resulting ast prior to the rendering stage.