[codex] Update cppjieba submodule#42
Conversation
|
这是来自QQ邮箱的假期自动回复邮件。您好,邮件已收到,我会尽快给您回复。
|
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b9d8809d7d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| @@ -1 +1 @@ | |||
| Subproject commit b3602bef7d1f67521a61788a74fb5801a0e62cd3 | |||
| Subproject commit 3503cb3bc44405d940420cc714b0d217d481ff53 | |||
There was a problem hiding this comment.
Preserve LocalVector's small-buffer behavior
When simhash segments any non-empty document, cppjieba's Trie::Find builds one Dag per rune and immediately appends to Dag::nexts; this submodule revision pulls in limonp 4065c5f, where LocalVector is now a plain std::vector with no inline capacity. That turns the first append for almost every character into a heap allocation instead of using the previous 16-element stack buffer, so common make()/extract() calls allocate O(text length) more and can significantly slow large-document processing. Please keep the old small-buffer behavior or reserve/replace this hot-path container before bumping cppjieba.
Useful? React with 👍 / 👎.
Summary
Update
submodules/cppjiebafromb3602befto3503cb3b.This brings in the current
limonpLocalVectorcompatibility wrapper, the explicit<cassert>dependency fix, Bazel smoke coverage, and the latest cppjieba CI/test updates.Validation
Run from an in-repository build directory because the existing tests use relative source paths:
cmake -S . -B build -DCMAKE_BUILD_TYPE=Releasecmake --build build --parallel 2ctest --test-dir build --output-on-failureResult: 2/2 tests passed.