chore: remove deny with sqlx

[driftluo]

What problem does this PR solve?

remove deny with sqlx

By the way, I think that SQL creates tables and indexes that can use IF NOT EXISTS to avoid judging whether it has been initialized. Although it is a non-standard statement, but both SQLite and Postgresql support it.

ckb/util/rich-indexer/resources/create_postgres_table.sql

Line 1 in 21f49f2

CREATE TABLE block(

ckb/util/rich-indexer/src/store.rs

Line 68 in 21f49f2

let require_init = self.is_postgres_require_init(db_config).await?;

ckb/util/rich-indexer/src/store.rs

Lines 77 to 79 in 21f49f2

	if require_init {
	self.create_tables_for_postgres().await?;
	}

ckb/util/rich-indexer/src/store.rs

Lines 251 to 266 in 21f49f2

	fn is_sqlite_require_init(db_config: &RichIndexerConfig) -> bool {
	// for test
	if db_config.store == Into::<PathBuf>::into(MEMORY_DB) {
	return true;
	}
	if !db_config.store.exists() {
	if let Some(parent) = db_config.store.parent() {
	std::fs::create_dir_all(parent).expect("Create db directory");
	}
	OpenOptions::new()
	.write(true)
	.create(true)
	.open(&db_config.store)
	.expect("Create db file");
	return true;

ckb/util/rich-indexer/src/store.rs

Lines 204 to 221 in 21f49f2

	pub async fn is_postgres_require_init(
	&mut self,
	db_config: &RichIndexerConfig,
	) -> Result<bool> {
	// Connect to the "postgres" database first
	let mut temp_config = db_config.clone();
	temp_config.db_name = "postgres".to_string();
	let uri = build_url_for_postgres(&temp_config);
	let mut connection_options = AnyConnectOptions::from_str(&uri)?;
	connection_options.log_statements(LevelFilter::Trace);
	let tmp_pool_options = AnyPoolOptions::new();
	let pool = tmp_pool_options.connect_with(connection_options).await?;
	// Check if database exists
	let query =
	SQLXPool::new_query(r#"SELECT EXISTS (SELECT FROM pg_database WHERE datname = $1)"#)
	.bind(db_config.db_name.as_str());
	let row = query.fetch_one(&pool).await?;

Check List

Tests

• Unit test
• Integration test

Release note

None: Exclude this PR from the release note.

[eval-exec]

Perhaps we can check and return early with Ok(row_lock.get::<i64, _>(0) == 1) after immediately fetching row_lock after L220 ?

[eval-exec]

How about move row_lock.get::<i64, _>(0) == 1 checking in this line?

[15168316096]

@driftluo
I understand that this pr may be two things to avoid: one is to pre-judge the secondary creation of tables and indexes, and to use sqlx and pgpool to optimize asynchronous queries.
The function points of verification should be,

1. The function of rich-indexer returns to normal
2. Query optimization of slow queries in pgpool asynchronous mode
   cc @gpBlockchain