Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix: Add schema name if dropping index in pgvector store #1277

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

fix: Add schema name if dropping index in pgvector store #1277

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5804c5f
fix: Add schema name if dropping index in pgvector store
mabartcz Jan 6, 2025
a5bc364
Merge branch 'deepset-ai:main' into main
mabartcz Jan 9, 2025
8614b9e
fix: Remove check for deletion in src
mabartcz Jan 9, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 15 additions & 1 deletion integrations/pgvector/src/haystack_integrations/document_stores/pgvector/document_store.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -389,9 +389,23 @@ def _handle_hnsw(self):
)
return

sql_drop_index = SQL("DROP INDEX IF EXISTS {index_name}").format(index_name=Identifier(self.hnsw_index_name))
sql_drop_index = SQL("DROP INDEX IF EXISTS {schema_name}.{index_name}").format(
schema_name=Identifier(self.schema_name), index_name=Identifier(self.hnsw_index_name)
)
self._execute_sql(sql_drop_index, error_msg="Could not drop HNSW index")

index_exists_after_deletion = bool(
self._execute_sql(
"SELECT 1 FROM pg_indexes WHERE schemaname = %s AND tablename = %s AND indexname = %s",
(self.schema_name, self.table_name, self.hnsw_index_name),
"Could not check if HNSW index exists",
).fetchone()
)

if index_exists and index_exists_after_deletion:
error_message = "The HNSW was not dropped"
raise DocumentStoreError(error_message)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would remove this check.
If the HNSW index is not dropped, the following invocation self._create_hnsw_index() will fail. Right?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You are right; I overthought that. The only reason it might fail during creation is if it wasn't deleted beforehand. However, you wouldn't know that was the cause. That said, this will be tested moving forward.


self._create_hnsw_index()

def _create_hnsw_index(self):
Expand Down
42 changes: 42 additions & 0 deletions integrations/pgvector/tests/test_document_store.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,13 @@
#
# SPDX-License-Identifier: Apache-2.0

import os
import random
import string
from unittest.mock import patch

import numpy as np
import psycopg
import pytest
from haystack.dataclasses.document import ByteStream, Document
from haystack.document_stores.errors import DuplicateDocumentError
Expand Down Expand Up @@ -259,3 +263,41 @@ def test_from_pg_to_haystack_documents():
assert haystack_docs[2].meta == {"meta_key": "meta_value"}
assert haystack_docs[2].embedding == [0.7, 0.8, 0.9]
assert haystack_docs[2].score is None


@pytest.mark.integration
def test_hnsw_index_recreation_in_new_schema():
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the test can be improved/changed.

I would probably first create a Document Store instance,
then create another one with hnsw_recreate_index_if_exists=True and verify in some way that the HNSW index is first dropped and then recreated.

I can help with the implementation, if needed.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Your help would be greatly appreciated, as I’ve been trying to find a way to check if the index was recreated, but I haven’t found a solution yet. From this "high-level" scope, I can’t verify if the deletion was successful. Because it happened inside the "process".

# Set your Postgres connection string (or set PG_CONN_STR in your environment directly).
os.environ["PG_CONN_STR"] = "postgresql://postgres:postgres@localhost:5432/postgres"

table_name = "test_table"
index_name = f"{table_name}_index"
schema_name = "".join(random.choices(string.ascii_letters, k=8)).lower() # noqa: S311
embedding_dimension = 1024

# Create the new schema if it doesn't exist.
with psycopg.connect(os.environ["PG_CONN_STR"]) as connection:
with connection.cursor() as cursor:
cursor.execute(f"CREATE SCHEMA IF NOT EXISTS {schema_name};")
connection.commit()

# Instantiate the document store in the new schema with HNSW indexing.
document_store = PgvectorDocumentStore(
embedding_dimension=embedding_dimension,
schema_name=schema_name,
vector_function="cosine_similarity",
recreate_table=False,
search_strategy="hnsw",
table_name=table_name,
hnsw_index_name=index_name,
hnsw_recreate_index_if_exists=True, # This ensures we drop/re-create the index if it exists
keyword_index_name=f"{table_name}_keyword_index",
)

# First write documents
docs1 = [Document(content="Test Content 1", embedding=[0.8] * embedding_dimension)]
document_store.write_documents(docs1)

# Second write documents
docs2 = [Document(content="Test Content 2", embedding=[0.7] * embedding_dimension)]
document_store.write_documents(docs2)
Loading