alpha
Login
or
Join now
boltless.me
/
zoekt
Star
0
Fork
0
Atom
Configure Feed
Issues
Pull Requests
Commits
Tags
Feed URL
Select the types of activity you want to include in your feed.
fork of https://github.com/sourcegraph/zoekt
Star
0
Fork
0
Atom
Configure Feed
Issues
Pull Requests
Commits
Tags
Feed URL
Select the types of activity you want to include in your feed.
Overview
Issues
Pulls
Pipelines
zoekt
/
internal
/
e2e
/
at
456196a74a627db053099c7cb7795380bf1be645
2 folders
5 files
Stefan Hengl
ranking: add tiebreakers to BM25 (#914)
1y ago
914a27d7
examples
Move root-level index code to index package (#902) In the repo root, we have a bunch of low level logic around index building and searching. So we end up exposing internal logic through the main public `zoekt` package, for example `zoekt.Merge(...)`. This PR moves it into the `build` package, so all code related to index building lives together. It then renames `build` to `index` to reflect the broader focus on indexing and searching the index.
1 year ago
testdata
scoring: use repo freshness as tiebreaker (#832) We ignore priority and instead use the latest commit date as repo rank. This has a big impact for Sourcegraph because it means we switch from star count to repo freshness as tiebreaker. As a minor tweak, we also separate query based scores from tiebreakers. To achieve this we reserve the last 7 digits of a score for tiebreakers: - 5 digits (maxUint16) for repo rank - 2 digits ([0,10]) for file order (2 digits). Example: Before: score: 8775.35 <- atom(2):200, fragment:8550.00, repo-rank: 19, doc-order:6.35 After: score: 8750_00019_06.35 <- atom(2):200, fragment:8550.00, repo-rank: 19, doc-order:6.35
2 years ago
doc.go
ranking: add tiebreakers to BM25 (#914) This adds repo freshness and file order as tiebreakers to the final bm25 score, just like we have for Zoekt's default scoring. During testing I found that it is a lot less likely for the tiebreakers to have an effect with BM25 because the score depends on qualites of the document, such as the relative length and number of matches, which usually differ even with the quality of the match is similar. Test plan: - Score tests still pass - manual testing: see screenshots
1 year ago
e2e_index_test.go
Move root-level index code to index package (#902) In the repo root, we have a bunch of low level logic around index building and searching. So we end up exposing internal logic through the main public `zoekt` package, for example `zoekt.Merge(...)`. This PR moves it into the `build` package, so all code related to index building lives together. It then renames `build` to `index` to reflect the broader focus on indexing and searching the index.
1 year ago
e2e_rank_test.go
Move root-level index code to index package (#902) In the repo root, we have a bunch of low level logic around index building and searching. So we end up exposing internal logic through the main public `zoekt` package, for example `zoekt.Merge(...)`. This PR moves it into the `build` package, so all code related to index building lives together. It then renames `build` to `index` to reflect the broader focus on indexing and searching the index.
1 year ago
e2e_test.go
zoekt-archive-index: split out ranking tests and archive indexing (#712) We had ranking e2e tests living in the zoekt-archive-index cmd for convenience since that contained useful functions for indexing a remote tarball from the GitHub API. This commit splits the archive functionality into a new internal/archive package and the ranking tests into a new internal/e2e package. The zoekt-archive-index code is now quite minimal. This is similiar to how zoekt-git-index mostly just calls out to the gitindex package. What is different is that archive package is marked internal, unlike gitindex. gitindex should also be internal, but the code predates go's support for internal. I suspect more of our e2e tests will end up in this package. Test Plan: go test ./...
2 years ago
scoring_test.go
ranking: add tiebreakers to BM25 (#914) This adds repo freshness and file order as tiebreakers to the final bm25 score, just like we have for Zoekt's default scoring. During testing I found that it is a lot less likely for the tiebreakers to have an effect with BM25 because the score depends on qualites of the document, such as the relative length and number of matches, which usually differ even with the quality of the match is similar. Test plan: - Score tests still pass - manual testing: see screenshots
1 year ago