在PostgreSQL / SQLAlchemy中加快JSONB全文搜索

Question

The performance of my JSONB full text search is awfully slow using PostgreSQL and SQLAlchemy. How can I speed it up?

Model

class Book(Base):
    __tablename__ = "book"
    id = Column(Integer, primary_key=True)
    jsondata = Column(JSONB)
    __table_args__ = (Index('index_jsondesc',
                      text("(jsondata->'description') jsonb_path_ops"),
                      postgresql_using="gin"),)

Full text search in the JSONB column

class BookSearch:
    def __init__(self):
        pass
    def search(keyword):
        self.query = self.query.filter(Book.jsondata['description'].cast(Unicode).match(keyword))

booksearch = BookSearch()
booksearch.search("Python")

Answer 1

Given selective enough queries, speeding up full text search queries means having the proper index in place. jsonb_path_ops do not benefit full text search:

The non-default GIN operator class jsonb_path_ops supports indexing the @> operator only.

Instead you need (for example) a functional index for explicit to_tsvector() :

class Book(Base):
    __tablename__ = "book"
    id = Column(Integer, primary_key=True)
    jsondata = Column(JSONB)
    __table_args__ = (
        Index('index_jsondesc',
              func.to_tsvector('english', jsondata['description'].astext),
              postgresql_using="gin"),
    )

Note that you must choose the configuration to use when defining the index. Your query must then match the configuration used in the index:

def search(keyword):
    tsvector = func.to_tsvector('english', Book.jsondata['description'].astext)
    self.query = self.query.filter(tsvector.match(keyword))