Skip to content

Commit

Permalink
refactor: improve content transformer maintenability
Browse files Browse the repository at this point in the history
  • Loading branch information
@pi0
pi0 committed Mar 18, 2024
1 parent a4e2121 commit e7244af
Showing 1 changed file with 146 additions and 99 deletions.
245 changes: 146 additions & 99 deletions app/server/plugins/content.ts
View file
Original file line number Diff line number Diff line change
@@ -1,124 +1,102 @@
// @ts-ignore
export default defineNitroPlugin((nitroApp) => {
nitroApp.hooks.hook('content:file:afterParse' as any, (file) => {
nitroApp.hooks.hook('content:file:afterParse', (file: ContentFile) => {
// Filter out non-markdown files
if (!file._id.endsWith('.md')) {
if (!file._id?.endsWith('.md')) {
return
}

// Set the icon for the file if it is not already set
if (!file.icon) {
file.icon = resolveIcon(file._path)
}

// Remove first h1 from markdown files as it is added to front-matter as title
if (file.body?.children?.[0]?.tag === 'h1') {
const text = getTextContents(file.body.children[0].children)
if (file.title === text) {
file.body.children.shift()
}
}
transformFile(file)
resolveFileIcon(file)

// Only use the first blockquote as the description
const firstChild = file.body.children?.[0]
const firstChildText = getTextContents(firstChild?.children)
if (firstChild?.tag === 'blockquote' && firstChildText && !firstChildText.startsWith('!')) {
file.description = firstChildText
file.body.children.shift()
} else {
file.description = '' // Avoid duplication
}

// Handle GitHub flavoured markdown blockquotes
// https://github.com/orgs/community/discussions/16925
for (const [idx, node] of (file.body?.children || []).entries()) {
if (
node.tag === 'blockquote' && // blockquote > p x 2 > span > text
['!NOTE', '!TIP', '!IMPORTANT', '!WARNING', '!CAUTION'].includes(
node.children?.[0]?.children?.[0]?.children?.[0]?.value,
)
) {
node.type = 'element'
node.tag = node.children?.[0]?.children?.[0]?.children?.[0]?.value.slice(1).toLowerCase()
node.children[0].children.shift()
}

// CONVERT OL->LI to Steps
// TODO: Find a way to opt out of this transformation if needed within markdown.
if (node.tag === 'ol' && node.children.length > 0 && node.children[0].tag === 'li') {
const stepsChildren = node.children.map((li) => {
const label = li.children?.[0]?.value ?? undefined
// Exclude br tags from children to avoid spacing
const children = (label ? li.children.slice(1) : []).filter((child) => !['br'].includes(child.tag))

return {
type: 'element',
tag: 'div',
props: {
label,
},
children,
}
})

// For now we only check if there is at least (1) content to generate the steps..
const stepsHaveContent = stepsChildren.some((step) => step.children.length > 0)
if (stepsHaveContent) {
node.type = 'element'
node.tag = 'Steps'
node.props = {}
node.children = stepsChildren
}
}

// Generate Code Groups
generateCodeGroup(idx, file.body.children)
transformGithubAlert(node)
transformStepsList(node)
transformCodeGroups(idx, file.body?.children)
}
})
})

function isNamedCodeBlock(children: any): boolean {
return children?.tag === 'pre' && children?.children?.[0]?.tag === 'code' && children?.props?.filename
// --- transform github alerts ---

// Handle GitHub flavoured markdown blockquotes
// https://github.com/orgs/community/discussions/16925
function transformGithubAlert(node: ContentNode) {
const firstChildValue = node.children?.[0]?.children?.[0]?.children?.[0]?.value || ''
if (
node.tag === 'blockquote' && // blockquote > p x 2 > span > text
['!NOTE', '!TIP', '!IMPORTANT', '!WARNING', '!CAUTION'].includes(firstChildValue)
) {
node.type = 'element'
node.tag = firstChildValue.slice(1).toLowerCase()
node.children?.[0].children?.shift()
}
}

function generateCodeGroup(currChildIdx: number, children: any[]) {
if (!isNamedCodeBlock(children[currChildIdx])) {
return
// --- transform steps list ---

function transformStepsList(node: ContentNode) {
// CONVERT OL->LI to Steps
// TODO: Find a way to opt out of this transformation if needed within markdown.
if (node.tag === 'ol' && (node.children?.length || 0) > 0 && node.children?.[0].tag === 'li') {
const stepsChildren = node.children.map((li) => {
const label = li.children?.[0]?.value ?? undefined
// Exclude br tags from children to avoid spacing
const children = ((label && li.children?.slice(1)) || []).filter((child) => !['br'].includes(child.tag || ''))

return {
type: 'element',
tag: 'div',
props: {
label,
},
children,
}
})

// For now we only check if there is at least (1) content to generate the steps..
const stepsHaveContent = stepsChildren.some((step) => step.children.length > 0)
if (stepsHaveContent) {
node.type = 'element'
node.tag = 'Steps'
node.props = {}
node.children = stepsChildren
}
}
}

const group: any[] = []
// --- transform first h1 and blockquote ---

for (let i = currChildIdx; i < children.length; i++) {
const nextNode = children[i]
if (!isNamedCodeBlock(nextNode)) {
break
function transformFile(file: ContentFile) {
// Remove first h1 from markdown files as it is added to front-matter as title
if (file.body?.children?.[0]?.tag === 'h1') {
const text = _getTextContents(file.body.children[0].children)
if (file.title === text) {
file.body.children.shift()
}
group.push(nextNode)
children[i] = { type: 'text', value: '' }
}

// Replace current children with the new code group if it has two or more code blocks
if (group.length > 1) {
children[currChildIdx] = {
type: 'element',
tag: 'CodeGroup',
children: [...group],
}
// Only use the first blockquote as the description
const firstChild = file.body?.children?.[0]
const firstChildText = _getTextContents(firstChild?.children)
if (firstChild?.tag === 'blockquote' && firstChildText && !firstChildText.startsWith('!')) {
file.description = firstChildText
file.body?.children?.shift()
} else {
file.description = '' // Avoid duplication
}
}

function getTextContents(children: any[]): string {
return (children || [])
.map((child) => {
if (child.type === 'element') {
return getTextContents(child.children)
}
return child.value
})
.join('')
// --- resolve icon ---

function resolveFileIcon(file: ContentFile) {
if (file.icon) {
return
}
file.icon = _resolveIcon(file._path)
}

// A set of common icons
const commonIcons = [
const _commonIcons = [
{
pattern: 'guide',
icon: 'ph:book-open-duotone',
Expand All @@ -145,16 +123,85 @@ const commonIcons = [
},
]

function resolveIcon(path: string) {
function _resolveIcon(path: string = '') {
// Split the path into parts and reverse it
const paths = path.slice(1).split('/').reverse()

// Search for icons in reverse order
for (const p of paths) {
for (const icon of commonIcons) {
for (const icon of _commonIcons) {
if (p.includes(icon.pattern)) {
return icon.icon
}
}
}
}

// --- transform code groups ---

function transformCodeGroups(currChildIdx: number, children: ContentNode[] = []) {
if (!children?.length || !_isNamedCodeBlock(children[currChildIdx])) {
return
}

const group: ContentNode[] = []

for (let i = currChildIdx; i < children.length; i++) {
const nextNode = children[i]
if (!_isNamedCodeBlock(nextNode)) {
break
}
group.push(nextNode)
children[i] = { type: 'text', value: '' }
}

// Replace current children with the new code group if it has two or more code blocks
if (group.length > 1) {
children[currChildIdx] = {
type: 'element',
tag: 'CodeGroup',
children: [...group],
}
}
}

function _isNamedCodeBlock(children: ContentNode): boolean {
return children?.tag === 'pre' && children?.children?.[0]?.tag === 'code' && children?.props?.filename
}

// --- internal utils ---

function _getTextContents(children: ContentNode[] = []): string {
return (children || [])
.map((child) => {
if (child.type === 'element') {
return _getTextContents(child.children)
}
return child.value
})
.join('')
}

// --- types ---

// TODO: @nuxt/content runtimes seems both not well typed and also crashes my TS server or might be doing it wrong.

interface ContentNode {
type?: string
tag?: string

children?: ContentNode[]
props?: Record<string, any>
value?: string
}

interface ContentFile {
_id?: string
_path?: string
icon?: string
description?: string
title?: string
body?: {
children?: ContentNode[]
}
}

0 comments on commit e7244af

Please sign in to comment.