Improve DuckDB insert performance; support Postgres materialized views; MySQL connection metrics (#311)

* Revert "Revert "Bump secrecy version (#248)" (#252)" (#254)

This reverts commit 3c38758.

* Upgrade datafusion-federation (#258)

* Upgrade federation (#259)

* Add primary keys mismatch verification for DuckDB table creation (#260)

* Add indexes mismatch detection for DuckDB table creation (#261)

* Add indexes mismatch detection for DuckDB table creation

* Update src/duckdb.rs

Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>

* Update src/duckdb.rs

Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>

* Update src/duckdb.rs

---------

Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>

* Add test for DuckDB memory_limit option (#263)

* Use quote_identifier for table name when fetching existing primary keys (#265)

* Defer index/constraint creation until after the initial data load. (#280)

* Defer index/constraint creation until after the initial data load.

* revert: Apply primary keys only at table creation

---------

Co-authored-by: peasee <98815791+peasee@users.noreply.github.com>

* refactor: New TableManager interfaces for DuckDB (#287)

* wip

* refactor: New TableCreator interfaces for DuckDB

* wip: Add some stubbed tests

* wip: #[ignore] not #[skip]

* test: Add creator.rs tests

* test: Add overwrite and append tests

* fix: Add schema validation back, add more unit tests

* refactor: Rename TableCreator::list_internal_tables

* chore: Remove commented out code

* Switch from Appender to duckdb_arrow_scan (#288)

* Switch from Appender to duckdb_arrow_scan

* Fix duckdb-rs commit

* Update src/duckdb.rs

Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>

---------

Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>

* duckdb-rs commit from spiceai-1.1.3-backported

* refactor: Append reuses existing base table, more tests, ViewCreator

* fix: InsertOp::Replace needs Append

* Add temp directory parameter (#289)

* refactor: Make TableDefinition PartialEq all of the time

* review: Rename TableCreator to TableManager, use option for internal table state

* refactor: Update error messaging on indexes

* Apply suggestions from code review

Co-authored-by: Phillip LeBlanc <phillip@leblanc.tech>

* fix: Address internal tables that are subsets or other table names

* feat: Add TableDefinition function to determine if it exists

* fix: Allow TableDefinition::name() to be public

* clippy: must_use

---------

Co-authored-by: Phillip LeBlanc <phillip@leblanc.tech>
Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>

* fix: Make an initial table during `TableProviderFactory::create()` for DuckDB (#290)

* fix: Make initial table if none exist for DuckDB

* refactor: Defer indexes to after first append load

* test: Update tests for append

* fix: Index detection for overwrite mode

* clippy: Reference immediately dereferenced

* test: Fix table creator tests

* Fixes a warning about missing indexes that shouldn't show up for newly created tables. (#292)

* De-duplicate attachments in DuckDBAttachments (#294)

* Allow connection pool size configuration for duckdb connection pool (#275)

* Add connection_pool_size parameter for duckdb table provider

* fix tests

* remove user configured connection_pool param, add min_idle configuration to pool

* fix

* apply suggestions

* fix tests

* add DuckDBConnectionPoolBuilder method

* remove idle size calculation, introduce non-breaking get_or_init instance methods

* fix tests

* Add mode to DuckDbConnectionPoolBuilder, remove lifetime for DuckDBConnectionPoolBuilder

* move build pool functionality into the DuckDB ConnectionPool Builder

* Use builder when get_or_init instance

* only pass pool to the get_or_init_instance_with_builder

* remove unnecessary error

* Add DuckDB setting 'preserve_insertion_order` (#298)

* Update the PostgreSQL query to support inferring the schema from materialized views (#300)

* Update the PostgreSQL query to support inferring the schema from materialized views

* update postgres schema tests, to use complex table for views

* Update examples/postgres.rs

* Update examples/postgres.rs

* Upgrade mysql_async, expose metrics and remove unnecessary type parameters (#302)

* Upgrade mysql_async, expose metrics and remove unnecessary type parameters

* Fix lint issue

* Expose the connection pool metrics via the table factory (#304)

* Support the `pool_min` and `pool_max` MySQL connection parameters (#305)

* Support the `pool_min` and `pool_max` parameters

* Support the `pool_min` and `pool_max` MySQL connection parameters

* Exercise the pool_min/pool_max options

* Fixing merge issues + use spiceai_duckdb_fork

---------

Co-authored-by: Sergei Grebnov <sergei.grebnov@gmail.com>
Co-authored-by: Luke Kim <80174+lukekim@users.noreply.github.com>
Co-authored-by: peasee <98815791+peasee@users.noreply.github.com>
Co-authored-by: Qianqian <130200611+Sevenannn@users.noreply.github.com>
Co-authored-by: Evgenii Khramkov <evgenii@spice.ai>

Author: 6 people

Cargo.lock

@@ -48,4 +48,4 @@ datafusion-proto = { version = "46" }
 datafusion-physical-expr = { version = "46" }
 datafusion-physical-plan = { version = "46" }
 datafusion-table-providers = { path = "core" }
-duckdb = { version = "=1.2.1" }
+duckdb = { version = "=1.2.1", package = "spiceai_duckdb_fork" } # Forked to add support for duckdb_scan_arrow, pending: https://github.com/duckdb/duckdb-rs/pull/488

Cargo.toml

@@ -64,7 +64,18 @@ CREATE TABLE companies (
   name VARCHAR(100)
 );
 
-INSERT INTO companies (id, name) VALUES (1, 'Acme Corporation');
+INSERT INTO companies (id, name) VALUES
+    (1, 'Acme Corporation'),
+    (2, 'Widget Inc.'),
+    (3, 'Gizmo Corp.'),
+    (4, 'Tech Solutions'),
+    (5, 'Data Innovations');
+
+CREATE VIEW companies_view AS
+  SELECT id, name FROM companies;
+
+CREATE MATERIALIZED VIEW companies_materialized_view AS
+  SELECT id, name FROM companies;
 EOF
 ```
 

README.md

@@ -109,6 +109,7 @@ rstest = "0.25.0"
 test-log = { version = "0.2", features = ["trace"] }
 tokio-stream = { version = "0.1", features = ["net"] }
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }
+tempfile = "3.19.1"
 
 [features]
 duckdb = [

core/Cargo.toml

@@ -76,6 +76,21 @@ async fn main() {
     )
     .expect("failed to register table");
 
+    let companies_view = table_factory
+        .table_provider(TableReference::bare("companies_view"))
+        .await
+        .expect("to create table provider for view");
+
+    let companies_materialized_view = table_factory
+        .table_provider(TableReference::bare("companies_materialized_view"))
+        .await
+        .expect("to create table provider for materialized view");
+
+    ctx.register_table("companies_view", companies_view)
+        .expect("to register view");
+    ctx.register_table("companies_materialized_view", companies_materialized_view)
+        .expect("to register materialized view");
+
     // Query Example 1: Query the renamed table through default catalog
     let df = ctx
         .sql("SELECT * FROM datafusion.public.companies_v2")
@@ -89,4 +104,18 @@ async fn main() {
         .await
         .expect("select failed");
     df.show().await.expect("show failed");
+
+    let df = ctx
+        .sql("SELECT * FROM companies_view")
+        .await
+        .expect("select from view failed");
+
+    df.show().await.expect("show failed");
+
+    let df = ctx
+        .sql("SELECT * FROM companies_materialized_view")
+        .await
+        .expect("select from materialized view failed");
+
+    df.show().await.expect("show failed");
 }