7. Tips and tricks

This section lists some tips and tricks that might be useful for using Literate.

Filesize of generated notebooks

When Literate executes a notebook the return value, i.e. the result of the last Julia expression in each cell is captured. By default Literate generates multiple renderings of the result in different output formats or MIMEs, just like IJulia.jl does. All of these renderings are embedded in the notebook and it is up to the notebook frontend viewer to select the most appropriate format to show to the user.

A common example is images, which can often be displayed in multiple formats, e.g. PNG (image/png), SVG (image/svg+xml) and HTML (text/html). As a result, the filesize of the generated notebook can become large.

In order to remedy this you can use the clever Julia package DisplayAs to limit the output capabilities of an object. For example, to "force" an image to be captures as image/png only, you can use

import DisplayAs
img = plot(...)
img = DisplayAs.PNG(img)

This can save some memory, since the image is never captured in e.g. SVG or HTML formats.

Note

It is best to always let the object be showable as text/plain. This can be achieved by nested calls to DisplayAs output types. For example, to limit an image img to be showable as just image/png and text/plain you can use

img = plot(...)
img = DisplayAs.Text(DisplayAs.PNG(img))

Adding admonitions using compound line filtering

Admonitions are a useful feature for drawing attention to particular elements of documentation. They are documented in Documenter.jl and an example of their use can be seen above in the blue 'note' box. Admonitions is a specific Julia markdown feature, and they are not recognized by either common mark or Jupyter notebooks. The md line filter can be used to make sure admonitions only show up in markdown output, for example:

#md # !!! note "Be aware!"
#md #     This a note style admonition!

It is important to note that both #md and the second # are required. Literate.jl interprets the first #md as a markdown exclusive line, and then strips it out. The second # tells Literate.jl that the line should be parsed as markdown and not a Julia code block. If you only include #md and not the second # then it will be parsed into Julia example block in the final documentation and not an actual admonition.

Custom parsing for markdown and notebook compatible admonitions

As mentioned above, admonitions are not compatible with Jupyter notebooks. (Though at time of writing this documentation, this is an open issue in Jupyter so may change in the future.) For now, we can write a custom preprocessor function so that admonitions are interpreted as quotes (with their own special formatting) in notebooks and proper admonitions in markdown. For the case of note admonitions, this can be written as follows:

function md_note(str)
    str = replace(str, r"^#note # (.*)$"m => s"""
    # !!! note
    #     \1""")
    return str
end

function nb_note(str)
    str = replace(str, r"^#note # (.*)$"m => s"""
    # > *Note*
    # > \1""")
    return str
end

using Literate

Literate.markdown("example.jl", "tmp/"; preprocess = md_note)

Literate.notebook("example.jl", "tmp/"; preprocess = nb_note)

This will allow us to turn the following source code in example.jl:

#note # This is a useful note.

into the correct admonition syntax in the markdown file generated:

!!! note
    This is a useful note.

and a quotation style formatting in the generated notebook cell:

> *Note*
> This is a useful note.

which, in an actual notebook cell, will look similar to:

Note
This is a useful note.

Debugging code execution

When Literate is executing code (i.e. when execute = true is set), it does so quietly. All the output gets captured and nothing gets printed into the terminal. This can make it tricky to know where things go wrong, e.g. when the execution stalls due to an infinite loop.

To help debug this, Literate has an @debug statement that prints out each code block that is being executed. In general, to enable the printing of Literate's @debug statements, you can set the JULIA_DEBUG environment variable to "Literate".

The easiest way to do that is to set the variable in the Julia session before running Literate by doing

ENV["JULIA_DEBUG"]="Literate"

Alternatively, you can also set the environment variable before starting the Julia session, e.g.

$ JULIA_DEBUG=Literate julia

or by wrapping the Literate calls in an withenv block

withenv("JULIA_DEBUG" => "Literate") do
    Literate.notebook("myscript.jl"; execute=true)
end