refactor: For MutableCSR, Let degree/cap list directly open in container and avoid copy in dump

[zhanglei1949]

Fix #265

[greptile-apps]

Your free trial has ended. If you'd like to continue receiving code reviews, you can add a payment method here.

[Copilot]

Pull request overview

This PR addresses #265 by refactoring MutableCsr to open and dump the degree (.deg) and capacity (.cap) lists directly via OpenContainer, reducing redundant O(V) element-wise copies during open/dump and aligning the approach with how the neighbor list is handled.

Changes:

• Refactor MutableCsr::open_internal() and dump() to use OpenContainer()/IDataContainer::Dump() directly for .deg and .cap.
• Replace the previous std::atomic<int>-backed degree buffer usage with an int buffer and atomic builtins in selected write paths.
• Minor formatting tidy in a test and ignore .aone_copilot/ in .gitignore.

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/storages/csr/mutable_csr.cc	Open/dump degree & capacity lists via containers; adjust degree access patterns across CSR ops.
include/neug/storages/csr/mutable_csr.h	Update member containers and degree access in put_edge()/views to use the new degree/cap containers.
tests/execution/test_value.cc	Formatting-only change to an EXPECT_DEATH call.
.gitignore	Add .aone_copilot/ ignore entry.

Comments suppressed due to low confidence (2)

src/storages/csr/mutable_csr.cc:448

• delete_edge bounds-checks offset against sz_arr[src] using a plain read from the degree array. Since degrees are incremented with atomic RMWs in put_edge/batch_put_edges, this non-atomic read can race with concurrent inserts/deletes and is undefined behavior in C++. Use an atomic load (e.g., std::atomic_ref<int>(sz_arr[src]).load(std::memory_order_relaxed) or __atomic_load_n) or ensure higher-level synchronization guarantees no concurrent degree modifications when calling this API.

    THROW_INVALID_ARGUMENT_EXCEPTION("src out of bound or offset out of bound");
  }
  nbr_t* nbrs = reinterpret_cast<nbr_t**>(adj_list_buffer_->GetData())[src];
  if (nbrs == nullptr) {

src/storages/csr/mutable_csr.cc:558

• In batch_put_edges, degrees are read/written via plain int accesses in earlier loops, but later updated with __atomic_fetch_add(&sz_arr[src], ...). Mixing atomic and non-atomic accesses to the same locations is undefined behavior and can lead to inconsistent degrees under concurrent readers/writers. Consider switching all degree accesses in this method to std::atomic_ref<int> (or __atomic_* builtins) for consistency, or clearly document/enforce that batch_put_edges runs single-threaded with no concurrent readers.

    nbr.data = data;
    nbr.timestamp.store(ts);
    added_edge_num++;
  }

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.