Parses Markdown files using Remark.
npm install --save gatsby-transformer-remark
// In your gatsby-config.js
plugins: [
{
resolve: `gatsby-transformer-remark`,
options: {
// CommonMark mode (default: true)
commonmark: true,
// Footnotes mode (default: true)
footnotes: true,
// Pedantic mode (default: true)
pedantic: true,
// GitHub Flavored Markdown mode (default: true)
gfm: true,
// Plugins configs
plugins: [],
},
},
],
The following parts of options
are passed down to Remark as options:
options.commonmark
options.footnotes
options.pedantic
options.gfm
The details of the Remark options above could be found in remark-parse
's documentation
A full explanation of how to use markdown in Gatsby can be found here: Creating a Blog with Gatsby
There are many Gatsby Remark plugins which you can install to customize how Markdown is processed. Many of them are demoed at https://using-remark.gatsbyjs.org/. See also the source code for using-remark.
It recognizes files with the following extensions as Markdown:
- md
- markdown
Each Markdown file is parsed into a node of type MarkdownRemark
.
All frontmatter fields are converted into GraphQL fields. TODO link to docs on auto-inferring types/fields.
This plugin adds additional fields to the MarkdownRemark
GraphQL type
including html
, excerpt
, headings
, etc. Other Gatsby plugins can also add
additional fields.
A sample GraphQL query to get MarkdownRemark nodes:
{
allMarkdownRemark {
edges {
node {
html
headings {
depth
value
}
frontmatter {
# Assumes you're using title in your frontmatter.
title
}
}
}
}
}
Using the following GraphQL query you'll be able to get the table of contents
{
allMarkdownRemark {
edges {
node {
html
tableOfContents
}
}
}
}
By default the tableOfContents is using the field slug
to generate URLs. You can however provide another field using the pathToSlugField parameter. Note that providing a non existing field will cause the result to be null. To alter the default values for tableOfContents generation, include values for heading
(string) and/or maxDepth
(number 1 to 6) in graphQL query. If a value for heading
is given, the first heading that matches will be omitted and the toc is generated from the next heading of the same depth onwards. Value for maxDepth
sets the maximum depth of the toc (i.e. if a maxDepth of 3 is set, only h1 to h3 headings will appear in the toc).
{
allMarkdownRemark {
edges {
node {
html
tableOfContents(
pathToSlugField: "frontmatter.path"
heading: "only show toc from this heading onwards"
maxDepth: 2
)
frontmatter {
# Assumes you're using path in your frontmatter.
path
}
}
}
}
}
To pass default options to the plugin generating the tableOfContents, configure it in gatsby-config.js as shown below. The options shown below are the defaults used by the plugin.
// In your gatsby-config.js
plugins: [
{
resolve: `gatsby-transformer-remark`,
options: {
tableOfContents: {
heading: null,
maxDepth: 6,
},
},
},
]
By default, excerpts have a maximum length of 140 characters. You can change the default using the pruneLength
argument. For example, if you need 500 characters, you can specify:
{
allMarkdownRemark {
edges {
node {
html
excerpt(pruneLength: 500)
}
}
}
}
By default, Gatsby will return excerpts as plain text. This might be useful for populating opengraph HTML tags for SEO reasons. You can also explicitly specify a PLAIN
format like so:
{
allMarkdownRemark {
edges {
node {
excerpt(format: PLAIN)
}
}
}
}
It's also possible to ask Gatsby to return excerpts formatted as HTML. You might use this if you have a blog post whose an excerpt contains markdown content--e.g. header, link, etc.--and you want these links to render as HTML.
{
allMarkdownRemark {
edges {
node {
excerpt(format: HTML)
}
}
}
}
You can also get excerpts in Markdown format.
{
allMarkdownRemark {
edges {
node {
excerpt(format: MARKDOWN)
}
}
}
}
gatsby-transformer-remark
uses gray-matter to parse markdown frontmatter, so you can specify any of the options mentioned here in the gatsby-config.js
file.
If you don't want to use pruneLength
for excerpts but a custom separator, you can specify an excerpt_separator
in the gatsby-config.js
file:
{
"resolve": `gatsby-transformer-remark`,
"options": {
"excerpt_separator": `<!-- end -->`
}
}
Any file that does not have the given excerpt_separator
will fall back to the default pruning method.
By default, excerpt
uses underscore.string/prune
which doesn't handle non-latin characters (https://github.com/epeli/underscore.string/issues/418).
If that is the case, you can set truncate
option on excerpt
field, like:
{
markdownRemark {
excerpt(truncate: true)
}
}
If your Markdown file contains HTML except
will not return a value.
In that case, you can set an excerpt_separator
in the gatsby-config.js
file:
{
"resolve": `gatsby-transformer-remark`,
"options": {
"excerpt_separator": `<!-- endexcerpt -->`
}
}
Edit your Markdown files to include that HTML tag after the text you'd like to appear in the excerpt:
---
title: "my little pony"
date: "2017-09-18T23:19:51.246Z"
---
<p>Where oh where is that pony?</p>
<!-- endexcerpt -->
<p>Is he in the stable or down by the stream?</p>
Then specify MARKDOWN
as the format in your graphql query:
{
markdownRemark {
excerpt(format: MARKDOWN)
}
}