From cea6834fef54da59fc1692d1db0221b93793238f Mon Sep 17 00:00:00 2001
From: Jacky Zhao <j.zhao2k19@gmail.com>
Date: Wed, 09 Aug 2023 07:26:33 +0000
Subject: [PATCH] profiling, better concurrency heuristics
---
quartz/processors/parse.ts | 68 +++++++++++++++------------------
1 files changed, 31 insertions(+), 37 deletions(-)
diff --git a/quartz/processors/parse.ts b/quartz/processors/parse.ts
index 55783dc..299d59a 100644
--- a/quartz/processors/parse.ts
+++ b/quartz/processors/parse.ts
@@ -7,22 +7,23 @@
import { ProcessedContent } from "../plugins/vfile"
import { PerfTimer } from "../perf"
import { read } from "to-vfile"
-import { FilePath, QUARTZ, ServerSlug, slugifyFilePath } from "../path"
+import { FilePath, QUARTZ, slugifyFilePath } from "../path"
import path from "path"
-import os from "os"
import workerpool, { Promise as WorkerPromise } from "workerpool"
-import { QuartzTransformerPluginInstance } from "../plugins/types"
import { QuartzLogger } from "../log"
import { trace } from "../trace"
+import { BuildCtx } from "../ctx"
export type QuartzProcessor = Processor<MDRoot, HTMLRoot, void>
-export function createProcessor(transformers: QuartzTransformerPluginInstance[]): QuartzProcessor {
+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!())
+ processor = processor.use(plugin.markdownPlugins!(ctx))
}
// MD AST -> HTML AST
@@ -30,7 +31,7 @@
// HTML AST -> HTML AST transforms
for (const plugin of transformers.filter((p) => p.htmlPlugins)) {
- processor = processor.use(plugin.htmlPlugins!())
+ processor = processor.use(plugin.htmlPlugins!(ctx))
}
return processor
@@ -54,6 +55,8 @@
platform: "node",
format: "esm",
packages: "external",
+ sourcemap: true,
+ sourcesContent: false,
plugins: [
{
name: "css-and-scripts-as-text",
@@ -72,42 +75,36 @@
})
}
-export function createFileParser(
- transformers: QuartzTransformerPluginInstance[],
- baseDir: string,
- fps: FilePath[],
- allSlugs: ServerSlug[],
- verbose: boolean,
-) {
+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 transformers.filter((p) => p.textTransform)) {
- file.value = plugin.textTransform!(file.value)
+ 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.relative(baseDir, file.path) as FilePath)
- file.data.allSlugs = allSlugs
+ 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 (verbose) {
- console.log(`[process] ${fp} -> ${file.data.slug}`)
+ if (argv.verbose) {
+ console.log(`[process] ${fp} -> ${file.data.slug} (${perf.timeSince()})`)
}
} catch (err) {
trace(`\nFailed to process \`${fp}\``, err as Error)
- throw err
}
}
@@ -115,29 +112,22 @@
}
}
-export async function parseMarkdown(
- transformers: QuartzTransformerPluginInstance[],
- baseDir: string,
- fps: FilePath[],
- verbose: boolean,
-): Promise<ProcessedContent[]> {
+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(verbose)
+ 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
- let concurrency = fps.length < CHUNK_SIZE ? 1 : os.availableParallelism()
-
- // get all slugs ahead of time as each thread needs a copy
- const allSlugs = fps.map((fp) =>
- slugifyFilePath(path.relative(baseDir, path.resolve(fp)) as FilePath),
- )
+ 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(transformers)
- const parse = createFileParser(transformers, baseDir, fps, allSlugs, verbose)
+ const processor = createProcessor(ctx)
+ const parse = createFileParser(ctx, fps)
res = await parse(processor)
} catch (error) {
log.end()
@@ -153,10 +143,14 @@
const childPromises: WorkerPromise<ProcessedContent[]>[] = []
for (const chunk of chunks(fps, CHUNK_SIZE)) {
- childPromises.push(pool.exec("parseFiles", [baseDir, chunk, allSlugs, verbose]))
+ childPromises.push(pool.exec("parseFiles", [argv, chunk, ctx.allSlugs]))
}
- const results: ProcessedContent[][] = await WorkerPromise.all(childPromises)
+ 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()
}
--
Gitblit v1.10.0