Is there a plan to add file caching to getGitTimestamp to reduce build time

[ATQQ]

Is your feature request related to a problem? Please describe.

When enabling either lastupdate or sitemap, my site's build time increased from 30s to 110s (on Mac Intel).

Describe the solution you'd like

add a cache file record timestamp

Describe alternatives you've considered

1. patch pkg
2. add lastUpdated cache plugin(but can`t override sitemap )

Additional context

No response

Validations

• Follow our Code of Conduct
• Read the docs.
• Read the Contributing Guidelines.
• Check that there isn't already an issue that asks for the same feature to avoid creating a duplicate.

[brc-dd]

Can be added. What are you using as the cache key?

[ATQQ]

now use md content-md5+filename

md5_filename.timestamp

[brc-dd]

Can you check with PR #4958 once? It doesn't implement the persistent FS cache yet, but it's taking almost half the time to build vitepress docs now (on m2 pro; on CI there isn't much difference). The difference will likely also depend on how many pages you have.

pnpm add https://pkg.pr.new/vitepress@4958

[ATQQ]

Great work, the performance is excellent! By the way, for my custom theme to reuse the generated cache, is there any chance we could expose these two methods: cacheAllGitTimestamps and getGitTimestamp? 😄

[brc-dd]

Yeah exported. Try once with your FS cache implementation too, if there is noticeable improvement we can add that too.

[ATQQ]

Okay. Based on the latest implementation, I'll give it a try. My previous file cache was built on the legacy logic and only took effect during subsequent builds. Currently, it seems the performance is quite similar to the new approach of fetching everything with a single git command.

[ATQQ]

There's a bug where the cache isn't fetching the correct number of results in one go.

I have over 500 documents, but the cache only captured a little over 400.

vitepress/src/node/utils/getGitTimestamp.ts

Lines 81 to 84 in bce0b53

	for (const [file, ts] of out) {
	const abs = path.resolve(gitRoot, file)
	if (fs.existsSync(abs)) cache.set(slash(abs), ts)
	}

[ATQQ]

I'm wondering if the current implementation has some limitations. I checked, and the output of git ls-files is as expected—it lists all tracked files correctly.

git ls-files '*.md' | wc -l

[ATQQ]

I've noticed the debug log code, and it is indeed being triggered. 😁

vitepress/src/node/utils/getGitTimestamp.ts

Lines 98 to 99 in bce0b53

	// most likely will never happen except for recently added files in dev
	debug(`[cache miss] ${file}`)

[brc-dd]

Interesting. I’ll check with your repo once.

Edit: Hmm, I tried running locally:

// ...
      out.clear()
      console.log(`cached git timestamps for ${cache.size} files`)
      resolve()

import { cacheAllGitTimestamps } from './src/node/utils/getGitTimestamp'

console.time('time')
await cacheAllGitTimestamps('/Users/divyansh/sugar-blog/packages/blogpress')
console.timeEnd('time')

gives:

cached git timestamps for 552 files
time: 183.676ms

Parsing implementation seems to be fine then? Can you try logging the root argument of cacheAllGitTimestamps in your patch? Is it pointing to the correct path? Also, there is a line like this:

if (fs.existsSync(abs)) cache.set(slash(abs), ts)

Can you try removing the existsSync condition and build with DEBUG="vitepress:git" to see if there are any cache misses?

On your repo with pnpm add vitepress@https://pkg.pr.new/vitepress@bce0b53

It seems to log twice for some reason, but the file count is correct?

Can you try cloning the repo to a clean directory and then testing? Also, can you share the output of git --version and the file generated by this:

cd packages/blogpress/
git log --pretty=format:%x1e%at%x00 --name-only -z -- '*.md' >out.log

[ATQQ]

Git Version

git --version
git version 2.39.3 (Apple Git-146)

Here are the current error logs from my local machine:
out.log

---

Looks like my local repo is just being weird then. I'm gonna try with a fresh clone.
 

[ATQQ]

The problematic repository.

1. remove exist

ached git timestamps for 442 files

2. no remove

cached git timestamps for 404 files

The files in the diff look like ones that got deleted from the repo's history.

Renamed README.md to index.md during the migration from VuePress.