quartz-research-note/quartz/processors/parse.ts
Jacky Zhao 86e41a82de format
2023-08-09 09:18:44 -07:00

162 lines
4.9 KiB
TypeScript

import esbuild from "esbuild"
import remarkParse from "remark-parse"
import remarkRehype from "remark-rehype"
import { Processor, unified } from "unified"
import { Root as MDRoot } from "remark-parse/lib"
import { Root as HTMLRoot } from "hast"
import { ProcessedContent } from "../plugins/vfile"
import { PerfTimer } from "../perf"
import { read } from "to-vfile"
import { FilePath, QUARTZ, slugifyFilePath } from "../path"
import path from "path"
import workerpool, { Promise as WorkerPromise } from "workerpool"
import { QuartzLogger } from "../log"
import { trace } from "../trace"
import { BuildCtx } from "../ctx"
export type QuartzProcessor = Processor<MDRoot, HTMLRoot, void>
export function createProcessor(ctx: BuildCtx): QuartzProcessor {
const transformers = ctx.cfg.plugins.transformers
// base Markdown -> MD AST
let processor = unified().use(remarkParse)
// MD AST -> MD AST transforms
for (const plugin of transformers.filter((p) => p.markdownPlugins)) {
processor = processor.use(plugin.markdownPlugins!(ctx))
}
// MD AST -> HTML AST
processor = processor.use(remarkRehype, { allowDangerousHtml: true })
// HTML AST -> HTML AST transforms
for (const plugin of transformers.filter((p) => p.htmlPlugins)) {
processor = processor.use(plugin.htmlPlugins!(ctx))
}
return processor
}
function* chunks<T>(arr: T[], n: number) {
for (let i = 0; i < arr.length; i += n) {
yield arr.slice(i, i + n)
}
}
async function transpileWorkerScript() {
// transpile worker script
const cacheFile = "./.quartz-cache/transpiled-worker.mjs"
const fp = "./quartz/worker.ts"
return esbuild.build({
entryPoints: [fp],
outfile: path.join(QUARTZ, cacheFile),
bundle: true,
keepNames: true,
platform: "node",
format: "esm",
packages: "external",
sourcemap: true,
sourcesContent: false,
plugins: [
{
name: "css-and-scripts-as-text",
setup(build) {
build.onLoad({ filter: /\.scss$/ }, (_) => ({
contents: "",
loader: "text",
}))
build.onLoad({ filter: /\.inline\.(ts|js)$/ }, (_) => ({
contents: "",
loader: "text",
}))
},
},
],
})
}
export function createFileParser(ctx: BuildCtx, fps: FilePath[]) {
const { argv, cfg } = ctx
return async (processor: QuartzProcessor) => {
const res: ProcessedContent[] = []
for (const fp of fps) {
try {
const perf = new PerfTimer()
const file = await read(fp)
// strip leading and trailing whitespace
file.value = file.value.toString().trim()
// Text -> Text transforms
for (const plugin of cfg.plugins.transformers.filter((p) => p.textTransform)) {
file.value = plugin.textTransform!(ctx, file.value)
}
// base data properties that plugins may use
file.data.slug = slugifyFilePath(path.posix.relative(argv.directory, file.path) as FilePath)
file.data.filePath = fp
const ast = processor.parse(file)
const newAst = await processor.run(ast, file)
res.push([newAst, file])
if (argv.verbose) {
console.log(`[process] ${fp} -> ${file.data.slug} (${perf.timeSince()})`)
}
} catch (err) {
trace(`\nFailed to process \`${fp}\``, err as Error)
}
}
return res
}
}
const clamp = (num: number, min: number, max: number) =>
Math.min(Math.max(Math.round(num), min), max)
export async function parseMarkdown(ctx: BuildCtx, fps: FilePath[]): Promise<ProcessedContent[]> {
const { argv } = ctx
const perf = new PerfTimer()
const log = new QuartzLogger(argv.verbose)
// rough heuristics: 128 gives enough time for v8 to JIT and optimize parsing code paths
const CHUNK_SIZE = 128
const concurrency = ctx.argv.concurrency ?? clamp(fps.length / CHUNK_SIZE, 1, 4)
let res: ProcessedContent[] = []
log.start(`Parsing input files using ${concurrency} threads`)
if (concurrency === 1) {
try {
const processor = createProcessor(ctx)
const parse = createFileParser(ctx, fps)
res = await parse(processor)
} catch (error) {
log.end()
throw error
}
} else {
await transpileWorkerScript()
const pool = workerpool.pool("./quartz/bootstrap-worker.mjs", {
minWorkers: "max",
maxWorkers: concurrency,
workerType: "thread",
})
const childPromises: WorkerPromise<ProcessedContent[]>[] = []
for (const chunk of chunks(fps, CHUNK_SIZE)) {
childPromises.push(pool.exec("parseFiles", [argv, chunk, ctx.allSlugs]))
}
const results: ProcessedContent[][] = await WorkerPromise.all(childPromises).catch((err) => {
const errString = err.toString().slice("Error:".length)
console.error(errString)
process.exit(1)
})
res = results.flat()
await pool.terminate()
}
log.end(`Parsed ${res.length} Markdown files in ${perf.timeSince()}`)
return res
}