How to search for file and folder name inside repo?

[finzzz]

I have enabled repo indexer so that I can search code inside files of certain repo. But I also want to search for file/folder name, how do I do that?

I have tried **keyword, **keyword**, **.py, etc. But it didn't work.

[lunny]

It's not supported yet.

[guillep2k]

Sorry for the cross-post:
It should not be that difficult to add file names to the repo indexer as indexed words; file names in this case should probably skip the indexer glob filter?

[ili101]

Any progress on this?
Or workaround available?

[guillep2k]

@ili101 we gladly welcome PRs. 😁

[ili101]

@ili101 we gladly welcome PRs. 😁

That's an awesome thing! unforgivably Go is not one of my expertise (yet) 😁

[love1900905]

+1, for many time, filename/pathname is even more meaningful than file content.

[rengui]

First of all, thanks great gitea! I'm using gitea heavyly and it works lightweightly and smoothly for years!

Back to this topic, I need this feature as well recently. After kinds of struggling,

• gitea code search, kinds of config => no supported yet.
• Kooder (a gitee open source code search tool for gitlab/gitee/gitea), can search filename itself, but not pathname yet (https://gitee.com/koode/kooder/issues/I4MWJA?from=project-issue).

It was resolved today, by change 1 line of code (gitea branch 1.15.7, for bleve search only):

// change modules\indexer\code\elastic_search.go if you are using ES.
modules\indexer\code\bleve.go

func (b *BleveIndexer) addUpdate(batchWriter ...
	......
	return batch.Index(id, &RepoIndexerData{
	RepoID:   repo.ID,
		CommitID: commitSha,
		Content:   string(charset.ToUTF8DropErrors(fileContents)),    // <--- this line

==>

	// Content:   string("Pathname: ") + string(update.Filename) + string(" \n") + string(charset.ToUTF8DropErrors(fileContents)),
	// replace '/' with ' / ', so that bleve think  '/foo/bar.txt' as  '/ foo / bar.txt'
	//   (then can be searched by both 'foo' and 'bar'; otherwise bleve only like '/foo/bar.txt' ???)
	Content:   string("Pathname: ") + strings.ReplaceAll(string(update.Filename), "/", " / ") + string(" \n") + string(charset.ToUTF8DropErrors(fileContents)),

then build your own gitea.exe and enjoy it :)

After replace this gitea.exe, you may need to delete all the code search index files (app.ini: REPO_INDEXER_PATH = ), restart gitea to make a full indexing (with pathname search enabled)

[lunny]

First of all, thanks great gitea! I'm using gitea heavyly and it works lightweightly and smoothly for years!

Back to this topic, I need this feature as well recently. After kinds of struggling,

• gitea code search, kinds of config => no supported yet.
• Kooder (a gitee open source code search tool for gitlab/gitee/gitea), can search filename itself, but not pathname yet (https://gitee.com/koode/kooder/issues/I4MWJA?from=project-issue).

It was resolved today, by change 1 line of code (gitea branch 1.15.7, for bleve search only):

// change modules\indexer\code\elastic_search.go if you are using ES. modules\indexer\code\bleve.go

func (b *BleveIndexer) addUpdate(batchWriter ...
	......
	return batch.Index(id, &RepoIndexerData{
	RepoID:   repo.ID,
		CommitID: commitSha,
		Content:   string(charset.ToUTF8DropErrors(fileContents)),    // <--- this line

==>

	// Content:   string("Pathname: ") + string(update.Filename) + string(" \n") + string(charset.ToUTF8DropErrors(fileContents)),
	// replace '/' with ' / ', so that bleve think  '/foo/bar.txt' as  '/ foo / bar.txt'
	//   (then can be searched by both 'foo' and 'bar'; otherwise bleve only like '/foo/bar.txt' ???)
	Content:   string("Pathname: ") + strings.ReplaceAll(string(update.Filename), "/", " / ") + string(" \n") + string(charset.ToUTF8DropErrors(fileContents)),

then build your own gitea.exe and enjoy it :)

After replace this gitea.exe, you may need to delete all the code search index files (app.ini: REPO_INDEXER_PATH = ), restart gitea to make a full indexing (with pathname search enabled)

A general method is to add a new field Filename in RepoIndexerData.

[rengui]

A general method is to add a new field Filename in RepoIndexerData.

Yes you are right, from gitea deve pov, it shall go in a general way, ensure code quality and extensibility.
What I mentioned was just from a gitea user pov, try to hack and get a working WA asap. :)

[delanym]

@lunny why cant your workaround be a PR?

[wxiaoguang]

FYI, there is a new feature: Go to file

• Feature: Find files in repo #15028

It should be more convenient if you know you are searching for file names:

[delanym]

@wxiaoguang thanks this does the expected.
It's a little slow on large repos the each time I bring up the page to do a search. What configuration options are there?

[wxiaoguang]

It's a little slow on large repos the each time I bring up the page to do a search.

Yup, it's not optimized yet. How large is your repo? For linux kernel repo (4G, 80k files) is about ten seconds on my side.

What configuration options are there?

What's the configuration do you mean? This feature doesn't have a config option yet, it's in 1.17 release.

[delanym]

@wxiaoguang config to enable a cache. I can cope with 10sec the first time the repo is indexed, but not every time I search for a file.
Im testing on a repo/branch with 86000 files - it's more like 15sec.
The equivalent find takes 0.18s: find . -iname "*IdleState*"

Also it should treat the search query as a sequence of strings, not a sequence of chars. Currently a search for "IdleState" returns a file like
IssuingSystem/Modules/UserControls/Views/ShortcutBarPresenter.cs

And actually - if I know anything about grep - its probably slower because its searching those individual characters.

[oetiker]

In our setup, we would love to be able to search for a filename across all repos ...

[michaelfresco]

@delanym
I was looking at the code from Lunny but, this function looks a bit different at the moment. Does anyone know how to implement the fix right now?

gitea/modules/indexer/code/bleve.go

Line 183 in e81ccc4

func (b *BleveIndexer) addUpdate(ctx context.Context, batchWriter git.WriteCloserError, batchReader *bufio.Reader, commitSha string,

[delvh]

This has already been implemented for some time (I think since 1.18?):

with the following dialogue

[wxiaoguang]

Hmm, IIRC this issue is asking about "repo indexer" , the "goto file" could help in some cases but not 100% resolves the issue.

[delvh]

Ah, you mean a global search through all repos?
Yes, that isn't possible yet.
However, this issue read like Go to file to me as it does exactly what this issue is asking for (search for file and folder names inside repo)…