Raw vectors data layer in HNSW + move to base class #523
+108
−89
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…estruction to base
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #523 +/- ##
==========================================
+ Coverage 96.93% 97.01% +0.07%
==========================================
Files 100 100
Lines 5287 5291 +4
==========================================
+ Hits 5125 5133 +8
+ Misses 162 158 -4 ☔ View full report in Codecov by Sentry. |
GuyAv46
reviewed
Nov 12, 2024
| @@ -1673,8 +1660,7 @@ void HNSWIndex<DataType, DistType>::removeAndSwap(idType internalId) { | |||
|
|
|||
| // Get the last element's metadata and data. | |||
| // If we are deleting the last element, we already destroyed it's metadata. | |||
| DataBlock &last_vector_block = vectorBlocks.back(); | |||
| auto last_element_data = last_vector_block.removeAndFetchLastElement(); | |||
| auto *last_element_data = this->vectors->getElement(curElementCount); | |||
There was a problem hiding this comment.
Consider using the API here
Suggested change
| auto *last_element_data = this->vectors->getElement(curElementCount); | |
| auto *last_element_data = getDataByInternalId(curElementCount); |
| void DataBlocksContainer::saveBlocks(std::ostream &output) const { | ||
| // Save number of blocks | ||
| unsigned int num_blocks = this->numBlocks(); | ||
| Serializer::writeBinaryPOD(output, num_blocks); |
There was a problem hiding this comment.
We should consider only saving the vectors without the metadata about the number of blocks and their sizes, so we can load them into other containers (or to different block sizes)
There was a problem hiding this comment.
This also means we don't need to add serialization to the container class, keeping it on the algorithm level
| assert(VecSimType_sizeof(vecType)); | ||
| this->vectors = new (this->allocator) DataBlocksContainer( | ||
| this->blockSize, this->dataSize, this->allocator, this->preprocessors->getAlignment()); |
There was a problem hiding this comment.
Suggested change
| this->blockSize, this->dataSize, this->allocator, this->preprocessors->getAlignment()); | |
| this->blockSize, this->dataSize, this->allocator, this->getAlignment()); |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Describe the changes in the pull request
Use the new
RawDataContainerinterface in HNSW, currently with an explicitDataBlocksContainerimplementation, and move the abstractvectorsmember to the base class.This includes:
DataBlocksContainerresponsibility, as we should not access the blocks directly anymore (should be applied for the graph data blocks later on as well).Mark if applicable