fix(formatter): use postgres canolical order to sort function options (#683)

Author: psteinroe

crates/pgls_pretty_print/src/nodes/alter_function_stmt.rs

@@ -36,9 +36,11 @@ pub(super) fn emit_alter_function_stmt(e: &mut EventEmitter, n: &AlterFunctionSt
     };
 
     // Emit actions (function options like IMMUTABLE, SECURITY DEFINER, etc.)
+    // Sort according to Postgres's canonical order
     if !n.actions.is_empty() {
         e.line(LineType::SoftOrSpace);
-        emit_comma_separated_list(e, &n.actions, |node, e| {
+        let sorted_actions = super::create_function_stmt::sort_function_options(&n.actions);
+        emit_comma_separated_list(e, &sorted_actions, |node, e| {
             let def_elem = assert_node_variant!(DefElem, node);
             super::create_function_stmt::format_function_option(e, def_elem, dollar_hint);
         });

crates/pgls_pretty_print/src/nodes/create_function_stmt.rs

@@ -90,8 +90,9 @@ pub(super) fn emit_create_function_stmt(e: &mut EventEmitter, n: &CreateFunction
         super::DollarQuoteHint::Function
     };
 
-    // Options
-    for option in &n.options {
+    // Options - sort according to Postgres's canonical order
+    let sorted_options = sort_function_options(&n.options);
+    for option in sorted_options {
         if let Some(pgls_query::NodeEnum::DefElem(def_elem)) = &option.node {
             e.line(LineType::Hard);
             format_function_option(e, def_elem, dollar_hint);
@@ -183,6 +184,53 @@ fn emit_function_parameter_list(e: &mut EventEmitter, params: &[&FunctionParamet
     }
 }
 
+/// Returns the canonical order for a function option.
+/// Postgres's canonical order (as seen in pg_dump output) is:
+/// 1. LANGUAGE
+/// 2. WINDOW
+/// 3. IMMUTABLE / STABLE / VOLATILE (volatility)
+/// 4. LEAKPROOF / NOT LEAKPROOF
+/// 5. STRICT / CALLED ON NULL INPUT (strict)
+/// 6. SECURITY DEFINER / SECURITY INVOKER (security)
+/// 7. PARALLEL (parallel)
+/// 8. COST (cost)
+/// 9. ROWS (rows)
+/// 10. SUPPORT (support)
+/// 11. SET options (set)
+/// 12. AS (function body)
+fn option_order(defname: &str) -> usize {
+    match defname.to_lowercase().as_str() {
+        "language" => 0,
+        "window" => 1,
+        "volatility" => 2,
+        "leakproof" => 3,
+        "strict" => 4,
+        "security" => 5,
+        "parallel" => 6,
+        "cost" => 7,
+        "rows" => 8,
+        "support" => 9,
+        "set" => 10,
+        "as" => 11,
+        _ => 100, // Unknown options go last
+    }
+}
+
+/// Sort function options according to Postgres's canonical order.
+pub(super) fn sort_function_options(
+    options: &[pgls_query::protobuf::Node],
+) -> Vec<pgls_query::protobuf::Node> {
+    let mut sorted: Vec<pgls_query::protobuf::Node> = options.to_vec();
+    sorted.sort_by_key(|node| {
+        if let Some(pgls_query::NodeEnum::DefElem(def_elem)) = &node.node {
+            option_order(&def_elem.defname)
+        } else {
+            100 // Non-DefElem nodes go last
+        }
+    });
+    sorted
+}
+
 pub(super) fn format_function_option(
     e: &mut EventEmitter,
     d: &pgls_query::protobuf::DefElem,

crates/pgls_pretty_print/src/normalize.rs

@@ -1205,14 +1205,63 @@ fn sql_value_to_func_call(
     })
 }
 
+/// Returns the canonical order for a function option.
+/// Postgres's canonical order (as seen in pg_dump output) is:
+/// 1. LANGUAGE
+/// 2. WINDOW
+/// 3. IMMUTABLE / STABLE / VOLATILE (volatility)
+/// 4. LEAKPROOF / NOT LEAKPROOF
+/// 5. STRICT / CALLED ON NULL INPUT (strict)
+/// 6. SECURITY DEFINER / SECURITY INVOKER (security)
+/// 7. PARALLEL (parallel)
+/// 8. COST (cost)
+/// 9. ROWS (rows)
+/// 10. SUPPORT (support)
+/// 11. SET options (set)
+/// 12. AS (function body)
+fn option_order(defname: &str) -> usize {
+    match defname.to_lowercase().as_str() {
+        "language" => 0,
+        "window" => 1,
+        "volatility" => 2,
+        "leakproof" => 3,
+        "strict" => 4,
+        "security" => 5,
+        "parallel" => 6,
+        "cost" => 7,
+        "rows" => 8,
+        "support" => 9,
+        "set" => 10,
+        "as" => 11,
+        _ => 100, // Unknown options go last
+    }
+}
+
+/// Sort function options according to Postgres's canonical order.
+fn sort_function_options(options: &mut [pgls_query::protobuf::Node]) {
+    options.sort_by_key(|node| {
+        if let Some(NodeEnum::DefElem(def_elem)) = &node.node {
+            option_order(&def_elem.defname)
+        } else {
+            100 // Non-DefElem nodes go last
+        }
+    });
+}
+
 /// Normalize function body strings by trimming whitespace.
 ///
 /// The formatter emits function bodies on separate lines from the dollar-quote delimiters,
 /// which adds leading/trailing newlines to the body. This normalization trims these
 /// so that semantically equivalent bodies compare equal.
+///
+/// Also sorts function options to canonical order so that semantically equivalent
+/// option orderings compare equal.
 fn normalize_function_body(node: &mut NodeEnum) {
     match node {
         NodeEnum::CreateFunctionStmt(stmt) => {
+            // Sort options to canonical order
+            sort_function_options(&mut stmt.options);
+
             for opt in &mut stmt.options {
                 if let Some(NodeEnum::DefElem(def)) = opt.node.as_mut()
                     && def.defname.eq_ignore_ascii_case("as")
@@ -1242,6 +1291,10 @@ fn normalize_function_body(node: &mut NodeEnum) {
         NodeEnum::InlineCodeBlock(block) => {
             block.source_text = block.source_text.trim().to_string();
         }
+        NodeEnum::AlterFunctionStmt(stmt) => {
+            // Sort actions to canonical order
+            sort_function_options(&mut stmt.actions);
+        }
         _ => {}
     }
 }

crates/pgls_pretty_print/tests/snapshots/multi/tests__aggregates_100.snap

@@ -1791,6 +1791,7 @@ create type avg_state as (total bigint, count bigint);
 
 create or replace function avg_transfn(state avg_state, n int)
 returns avg_state
+language plpgsql
 as $function$
 declare new_state avg_state;
 begin
@@ -1810,11 +1811,11 @@ begin
 
 	return null;
 end
-$function$
-language plpgsql;
+$function$;
 
 create function avg_finalfn(state avg_state)
 returns int
+language plpgsql
 as $function$
 begin
 	if state is null then
@@ -1823,11 +1824,11 @@ begin
 		return state.total / state.count;
 	end if;
 end
-$function$
-language plpgsql;
+$function$;
 
 create function sum_finalfn(state avg_state)
 returns int
+language plpgsql
 as $function$
 begin
 	if state is null then
@@ -1836,8 +1837,7 @@ begin
 		return state.total;
 	end if;
 end
-$function$
-language plpgsql;
+$function$;
 
 create aggregate my_avg (int) (stype = avg_state, sfunc = avg_transfn, finalfunc = avg_finalfn);
 
@@ -1924,6 +1924,7 @@ begin;
 
 create or replace function sum_transfn(state int, n int)
 returns int
+language plpgsql
 as $function$
 declare new_state int4;
 begin
@@ -1941,11 +1942,11 @@ begin
 
 	return null;
 end
-$function$
-language plpgsql;
+$function$;
 
 create function halfsum_finalfn(state int)
 returns int
+language plpgsql
 as $function$
 begin
 	if state is null then
@@ -1954,8 +1955,7 @@ begin
 		return state / 2;
 	end if;
 end
-$function$
-language plpgsql;
+$function$;
 
 create aggregate my_sum (int) (stype = int, sfunc = sum_transfn);
 
@@ -1969,8 +1969,8 @@ begin;
 
 create function balkifnull(bigint, int)
 returns bigint
-strict
 language plpgsql
+strict
 as $function$
 BEGIN
     IF $1 IS NULL THEN
@@ -2136,9 +2136,9 @@ begin;
 
 create function balkifnull(bigint, bigint)
 returns bigint
-parallel SAFE
-strict
 language plpgsql
+strict
+parallel SAFE
 as $function$
 BEGIN
     IF $1 IS NULL THEN
@@ -2186,8 +2186,8 @@ $function$;
 create function rwagg_finalfunc(x anyarray)
 returns anyarray
 language plpgsql
-strict
 immutable
+strict
 as $function$
 DECLARE
     res x%TYPE;
@@ -2209,8 +2209,8 @@ create aggregate rwagg (
 create function eatarray(x real[])
 returns real[]
 language plpgsql
-strict
 immutable
+strict
 as $function$
 BEGIN
     x[1] := x[1] + 1;

crates/pgls_pretty_print/tests/snapshots/multi/tests__aggregates_80.snap

@@ -2151,6 +2151,7 @@ create type avg_state as (total bigint, count bigint);
 
 create or replace function avg_transfn(state avg_state, n int)
 returns avg_state
+language plpgsql
 as $function$
 declare new_state avg_state;
 begin
@@ -2170,11 +2171,11 @@ begin
 
 	return null;
 end
-$function$
-language plpgsql;
+$function$;
 
 create function avg_finalfn(state avg_state)
 returns int
+language plpgsql
 as $function$
 begin
 	if state is null then
@@ -2183,11 +2184,11 @@ begin
 		return state.total / state.count;
 	end if;
 end
-$function$
-language plpgsql;
+$function$;
 
 create function sum_finalfn(state avg_state)
 returns int
+language plpgsql
 as $function$
 begin
 	if state is null then
@@ -2196,8 +2197,7 @@ begin
 		return state.total;
 	end if;
 end
-$function$
-language plpgsql;
+$function$;
 
 create aggregate my_avg (
   int
@@ -2316,6 +2316,7 @@ begin;
 
 create or replace function sum_transfn(state int, n int)
 returns int
+language plpgsql
 as $function$
 declare new_state int4;
 begin
@@ -2333,11 +2334,11 @@ begin
 
 	return null;
 end
-$function$
-language plpgsql;
+$function$;
 
 create function halfsum_finalfn(state int)
 returns int
+language plpgsql
 as $function$
 begin
 	if state is null then
@@ -2346,8 +2347,7 @@ begin
 		return state / 2;
 	end if;
 end
-$function$
-language plpgsql;
+$function$;
 
 create aggregate my_sum (int) (stype = int, sfunc = sum_transfn);
 
@@ -2371,8 +2371,8 @@ begin;
 
 create function balkifnull(bigint, int)
 returns bigint
-strict
 language plpgsql
+strict
 as $function$
 BEGIN
     IF $1 IS NULL THEN
@@ -2558,9 +2558,9 @@ begin;
 
 create function balkifnull(bigint, bigint)
 returns bigint
-parallel SAFE
-strict
 language plpgsql
+strict
+parallel SAFE
 as $function$
 BEGIN
     IF $1 IS NULL THEN
@@ -2608,8 +2608,8 @@ $function$;
 create function rwagg_finalfunc(x anyarray)
 returns anyarray
 language plpgsql
-strict
 immutable
+strict
 as $function$
 DECLARE
     res x%TYPE;
@@ -2631,8 +2631,8 @@ create aggregate rwagg (
 create function eatarray(x real[])
 returns real[]
 language plpgsql
-strict
 immutable
+strict
 as $function$
 BEGIN
     x[1] := x[1] + 1;