How jajir/hestiastore works
Project type and purpose
• HestiaStore is an embeddable, Java-based key-value storage engine library (no UI or network layer).
• It provides a lightweight API (Index and Directory abstractions) and implements on-disk/in-memory storage, optional write-ahead logging, caches, bloom filters, segments, and pluggable data file formats.
File structure
• Top-level: pom.xml, docs/, src/ (main, test, integration-test).
• src/main/java/org/hestiastore/index: Core API (Index, IndexConfiguration, Directory).
• Subpackages implement storage modules:
– directory: abstraction for MemDirectory, FsDirectory, ZipDirectory, etc.
– datatype: key/value serialization (TypeDescriptor, Convertor, VarLengthReader/Writer).
– chunkstore, chunkpairfile, datablockfile, sorteddatafile, unsorteddatafile: various on-disk file formats.
– segment: segment builder, compaction, splitting, searchers, consistency checks.
– bloomfilter: optional probabilistic membership check.
– cache: LRU and unique caches for segments or keys.
– log: write-ahead logging (LogWriter, LogFilesManager).
– properties: persisting index properties and configuration.
– scarceindex: alternative minimal index strategy.
– sst: high-level “Sorted String Table” index interface and state machine.
README for additional insights
• Describes features: in-memory or file-backed indexes, optional WAL, user-defined types, thread-safety toggle.
• Links to architecture docs, usage, configuration, logging guides.
Elements
• Labels: name each module/package clearly (e.g., “segment”, “chunkstore”).
– API (blue)
– Engine/core (green)
– Storage/IO (orange)