df7cb / pg_filedump

pg_filedump provides facilities for low-level examination of PostgreSQL tables and indexes
43 stars 18 forks source link

vector data #17

Closed FlynnHuang8 closed 4 months ago

FlynnHuang8 commented 1 year ago

Can you support parsing of -D vector, i.e. vector data?

df7cb commented 1 year ago

Hi, I guess you mean https://github.com/pgvector/pgvector ?

What's the use case for pg_filedump here? Non-core datatypes would best be extracted by getting PostgreSQL to run on the data directory, and just SELECTing the data.

FlynnHuang8 commented 1 year ago

When I build a vector index using pgvector, I want to use pg_dump to export the data and see the index structure, but it doesn't parse the HNSW index correctly, it would be great if it could support exporting!

df7cb commented 1 year ago

pg_dump won't look inside indexes.

If you want to look into page structures, try the pageinspect extension.

df7cb commented 1 year ago

Fwiw if someone submits a PR for hnsw index support, I wouldn't be opposed to merging it, but that's probably better handled first with in-PostgreSQL extensions.