[EthanVieira]

Allows Databricks TABLE queries to reference multipart names such as main.raw.source.

Reference: Names

[codecov-commenter]

Codecov Report

❌ Patch coverage is 94.11765% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 82.55%. Comparing base (9f66d76) to head (7465099).

Files with missing lines	Patch %	Lines
src/ast/query.rs	50.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2594   +/-   ##
=======================================
  Coverage   82.54%   82.55%           
=======================================
  Files          42       42           
  Lines       35556    35558    +2     
  Branches    35556    35558    +2     
=======================================
+ Hits        29349    29354    +5     
  Misses       2672     2672           
+ Partials     3535     3532    -3     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[LucaCappelletti94]

@EthanVieira please do add an entry to the fuzzer seed corpus with this syntax if there isn't one already

[LucaCappelletti94]

I believe that we need to make the Table struct hold an ObjectName, as TableFactor::Table does, so a qualified name of any length keeps its parts.

@@ -300,6 +300,4 @@ impl fmt::Display for SetQuantifier {
 pub struct Table {
-    /// Optional table name (absent for e.g. `TABLE` command without argument).
-    pub table_name: Option<Ident>,
-    /// Optional schema/catalog name qualifying the table.
-    pub schema_name: Option<Ident>,
+    /// Optional, possibly qualified, table name (absent for e.g. `TABLE` command without argument).
+    pub name: Option<ObjectName>,
 }

This of course will call for several changes both in the repo and downstream, but it seems like the cleanest solution to me. @alamb opinions?

[LucaCappelletti94]

You should keep every name part as its own Ident instead of joining the qualifiers into one. As written, main.raw.source and `main.raw`.source both give a schema_name whose value is "main.raw", and `a b`.raw.source gives the value "`a b`.raw" with the backticks inside it, so consumers cannot recover the parts. This relies on the Table change in the global note.

Suggested change

	let max_parts = if self.dialect.supports_multipart_table_query_name() {
	usize::MAX
	} else {
	2
	};
	let mut parts = vec![self.parse_identifier()?];
	while parts.len() < max_parts && self.consume_token(&Token::Period) {
	parts.push(self.parse_identifier()?);
	}
	Ok(Table {
	name: Some(ObjectName::from(parts)),
	})

[LucaCappelletti94]

You should assert the parsed parts, because verified_stmt only checks the round trip and the joined Ident passes it.

Suggested change

	databricks().verified_stmt("CREATE VIEW copy AS TABLE main.raw.source");
	let Statement::CreateView(view) =
	databricks().verified_stmt("CREATE VIEW copy AS TABLE main.raw.source")
	else {
	unreachable!()
	};
	assert_eq!(
	*view.query.body,
	SetExpr::Table(Box::new(Table {
	name: Some(ObjectName::from(vec![
	Ident::new("main"),
	Ident::new("raw"),
	Ident::new("source"),
	])),
	}))
	);

[EthanVieira]

I believe that we need to make the Table struct hold an ObjectName, as TableFactor::Table does, so a qualified name of any length keeps its parts.

@@ -300,6 +300,4 @@ impl fmt::Display for SetQuantifier {
 pub struct Table {
-    /// Optional table name (absent for e.g. `TABLE` command without argument).
-    pub table_name: Option<Ident>,
-    /// Optional schema/catalog name qualifying the table.
-    pub schema_name: Option<Ident>,
+    /// Optional, possibly qualified, table name (absent for e.g. `TABLE` command without argument).
+    pub name: Option<ObjectName>,
 }

This of course will call for several changes both in the repo and downstream, but it seems like the cleanest solution to me. @alamb opinions?

Haven't heard from @alamb, but I went ahead and did this.