2. File Format
The source file format for Literate is a regular, commented, julia (.jl
) scripts. The idea is that the scripts also serve as documentation on their own and it is also simple to include them in the test-suite, with e.g. include
, to make sure the examples stay up do date with other changes in your package.
2.1. Syntax
The basic syntax is simple:
- lines starting with
#
are treated as markdown, - all other lines are treated as julia code.
Leading whitespace is allowed before #
, but it will be removed when generating the output. Since #
-lines is treated as markdown we can not use that for regular julia comments, for this you can instead use ##
, which will render as #
in the output.
Lets look at a simple example:
# # Rational numbers
#
# In julia rational numbers can be constructed with the `//` operator.
# Lets define two rational numbers, `x` and `y`:
## Define variable x and y
x = 1//3
y = 2//5
# When adding `x` and `y` together we obtain a new rational number:
z = x + y
In the lines starting with #
we can use regular markdown syntax, for example the #
used for the heading and the backticks for formatting code. The other lines are regular julia code. We note a couple of things:
- The script is valid julia, which means that we can
include
it and the example will run (for example in thetest/runtests.jl
script, to include the example in the test suite). - The script is "self-explanatory", i.e. the markdown lines works as comments and thus serve as good documentation on its own.
For simple use this is all you need to know. The following additional special syntax can also be used:
#md
,#nb
,#jl
,#src
: tags to filter lines, see Filtering Lines,#-
(#+
): tag to manually control chunk-splits, see Custom control over chunk splits.
There is also some default convenience replacements that will always be performed, see Default Replacements.
Literate version 2.7 adds support for Julia multiline comments for markdown input. All multiline comments in the input are rewritten to regular comments as part of the preprocessing step, before any other processing is performed. For Literate to recognize multiline comments it is required that the start token (#=
) and end token (=#
) are placed on their own lines. Note also that it is allowed to have more than one =
in the tokens, for example
#=
This multiline comment
is treated as markdown.
=#
#=====================
This is also markdown.
=====================#
is rewritten to
# This multiline comment
# is treated as markdown.
# This is also markdown.
2.2. Filtering Lines
It is often useful to filter out lines in the source depending on the output format. For this purpose there are a number of "tokens" that can be used to mark the purpose of certain lines:
#md
: line exclusive to markdown output,#nb
: line exclusive to notebook output,#jl
: line exclusive to script output,#src
: line exclusive to the source code and thus filtered out unconditionally.
Lines starting or ending with one of these tokens are filtered out in the preprocessing step. In addition, for markdown output, lines ending with #hide
are filtered out similar to Documenter.jl.
#src
and #hide
are quite similar. The only difference is that #src
lines are filtered out before execution (if execute=true
) and #hide
lines are filtered out after execution.
The #hide
token requires at least Literate version 2.6.
Filter tokens at the end of the line requires at least Literate version 2.3.
The tokens can also be negated, for example a line starting with #!nb
would be included in markdown and script output, but filtered out for notebook output.
Suppose, for example, that we want to include a docstring within a @docs
block using Documenter. Obviously we don't want to include this in the notebook, since @docs
is Documenter syntax that the notebook will not understand. This is a case where we can prepend #md
to those lines:
#md # ```@docs
#md # Literate.markdown
#md # Literate.notebook
#md # Literate.markdown
#md # ```
The lines in the example above would be filtered out in the preprocessing step, unless we are generating a markdown file. When generating a markdown file we would simple remove the leading #md
from the lines. Beware that the space after the tag is also removed.
The #src
token can also be placed at the end of a line. This is to make it possible to have code lines exclusive to the source code, and not just comment lines. For example, if the source file is included in the test suite we might want to add a @test
at the end without this showing up in the outputs:
using Test #src
@test result == expected_result #src
2.3. Default Replacements
The following convenience "macros"/source placeholders are always expanded:
@__NAME__
:expands to the
name
keyword argument toLiterate.markdown
,Literate.notebook
andLiterate.script
(defaults to the filename of the input file).@__REPO_ROOT_URL__
:Can be used to link to files in the repository. For example
@__REPO_ROOT_URL__/src/Literate.jl
would link to the source of the Literate module. This variable is automatically determined on Travis CI, GitHub Actions and GitLab CI, but can be configured, see Configuration.@__NBVIEWER_ROOT_URL__
:Can be used if you want a link that opens the generated notebook in http://nbviewer.jupyter.org/. This variable is automatically determined on Travis CI, GitHub Actions and GitLab CI, but can be configured, see Configuration.
@__BINDER_ROOT_URL__
:Can be used if you want a link that opens the generated notebook in https://mybinder.org/. For example, to add a binder-badge in e.g. the HTML output you can use:
[![Binder](https://mybinder.org/badge_logo.svg)](@__BINDER_ROOT_URL__/path/to/notebook.inpynb)
This variable is automatically determined on Travis CI, GitHub Actions and GitLab CI, but can be configured, see Configuration.
GitHub Actions support for the macros above requires at least Literate version 2.1.
GitLab CI support for the macros above requires at least Literate version 2.2.