Fixes #3610: Add advanced SQL analytics to streaming cypher rows before returning results to the client

[vga91]

Fixes #3610

Overview

Fixes https://github.com/neo4j-contrib/neo4j-apoc-procedures/issues/3610

Created an apoc.jdbc.analytics(<cypherQuery>, <jdbcUrl>, <sqlQueryOverTemporaryTable>, <paramsList>, $config) procedure.
Imho is better than use an aggregation function, because it's more customizable, since I can choose what results to get and how to get them, otherwise to do something like below, with movies_count I should put e.g. a parameter aggKeys and increase the complexity.

            MATCH (p:Person)-[r:ACTED_IN]->(m:Movie)
            RETURN 
                p.name AS actor, 
                m.genre AS genre, 
                r.roles AS roles, 
                COUNT(m) AS movies_count

The procedure executes 3 steps:

1. creates a temporary table
2. populates the table starting from the <cypherQuery> param
3. queries the table with the <sqlQuery> param (which can be an complex analytics query) and returns data

Changes implemented

• Created test with apoc.load.jdbc* for DuckDB in extended/src/test/java/apoc/load/DuckDBJdbcTest.java
  • If we put the same JDBC url for all tests, flakily a random test remains pending for no apparent reason
  • also, created a test like this one for DuckDB
• Fixed errors related to DuckDB with Time and Date SQL Type:
  • https://github.com/vga91/neo4j-apoc-procedures/pull/468/files#diff-f8ec24426cb3f7100946579df5aed24135ebe4bb00d59c9b91be48c3545bfd92R235
  • https://github.com/vga91/neo4j-apoc-procedures/pull/468/files#diff-f8ec24426cb3f7100946579df5aed24135ebe4bb00d59c9b91be48c3545bfd92R263
  • with DuckDB we can create an in-memory database, where no data is persisted to disk , via jdbc:duckdb: URL. Therefore added fallback here to allow it.
• Improved executeUpdate and executeQuery, in order to pass also the Connection object and get that one if is not null, otherwise the temporary table will be destroyed if we create a new Connection.
• Created apoc.load.jdbc.analytics tests for:
  • DuckDb
  • MySQL
  • PostgreSQL
• Added docs
• config getOrDefault("temp_table", "neo4j_tmp_table")

EDIT: Changes after review

• added DuckDB, Postgres and MySql mapping to handle different types of numbers, dates and duration
• improved tests with the missing data types
• added escape chars to create table, based on Provider conf, to handle reserved keyword like time in MySQL
• refactoring: moved all jdbc stuff to apoc.load.jdbc package

To evaluate

• (maybe in a different PR/Issue): Improve handling: some SQL databases like Microsoft SQL Server create temp table in a different way

[RobertoSannino]

defined by the JDBC URL

[RobertoSannino]

Lets add a full example using an in-memory DuckDb to perform analytics

[RobertoSannino]

what if the Number is a Float?