Prev Next

Database / LanceDB Interview questions

What is versioning in LanceDB?

Versioning means every write to a LanceDB table — adding rows, updating them, deleting them, or changing the schema — creates a new, immutable version of the table rather than overwriting existing data in place, giving every table a complete, git-like history of changes.

print(table.version)          # current version number
history = table.list_versions()
table.checkout(3)             # view an old version (read-only, time-travel mode)
table.checkout_latest()       # return to tracking the latest version

Because older versions aren't deleted immediately, a table can be "checked out" at a past version to inspect what the data looked like at that point in time, and a past version can be restored, which creates a new version whose data matches the old one rather than deleting the version history in between.

This matters a lot for reproducibility in ML workflows — being able to point at exactly the version of a dataset a model was trained or evaluated against — and it's also what makes destructive operations recoverable, since a delete doesn't erase data immediately but marks it as removed in a new version, with actual space reclaimed only later through an explicit cleanup operation.

What does every write to a LanceDB table create?
What does restoring an old version do to the version history?

More Related questions...

What is LanceDB? What is the purpose of LanceDB? What is the Lance columnar format? What are the key features of LanceDB? What is an embedded vector database? What is Apache Arrow, and how does LanceDB use it? Define a table in LanceDB? What is a vector embedding? What are the supported languages/SDKs for LanceDB? How do you create a table in LanceDB? What is ANN (Approximate Nearest Neighbor) search? What is the IVF_PQ index in LanceDB? What is full-text search in LanceDB? What is hybrid search in LanceDB? Define the embedding function registry in LanceDB? What is schema evolution in LanceDB? List the storage backends supported by LanceDB? What is a scalar index in LanceDB? What is versioning in LanceDB? How do you connect to a LanceDB database? What is the difference between LanceDB and Pinecone? What is the difference between LanceDB and Chroma? What is the difference between LanceDB OSS and LanceDB Cloud? Why is LanceDB well suited for multimodal AI data? How does the Lance format differ from Parquet? Explain how IVF_PQ indexing works internally? What is the difference between IVF_PQ and HNSW indexing in LanceDB? How does LanceDB implement hybrid search using reranking? What is Reciprocal Rank Fusion (RRF), and how is it used in LanceDB? Explain the lifecycle of a write operation in LanceDB (versioning)? How do you perform time travel queries in LanceDB? What is the difference between checkout and restore in LanceDB? What are tags in LanceDB versioning? What are branches in LanceDB, and how do they differ from tags? How does LanceDB handle deletes internally? Explain the internal working of LanceDB's zero-copy data access via Arrow? How do you integrate LanceDB with LangChain for RAG? What is the role of DataFusion in LanceDB's query execution? How do you implement custom embedding functions in LanceDB? When should you use a scalar index versus a vector index? How do you optimize a LanceDB table for query performance through compaction? What is prefiltering vs postfiltering in LanceDB queries? Explain the internal working of product quantization (PQ) in vector indexing? How does LanceDB support multi-process concurrent access? What are the trade-offs of running LanceDB embedded versus as a managed cloud service? Explain the internal working of the manifest and commit protocol in Lance? How do you implement multimodal search across text and images in LanceDB? What is the role of object storage (S3/GCS) in LanceDB's architecture? How do you monitor and troubleshoot slow vector search queries in LanceDB? Explain the execution flow of a RAG pipeline built with LanceDB as the retrieval layer?
Show more question and Answers...


Comments & Discussions