From c2432611257fbbbcbbe79b2ce8dd543087784a7a Mon Sep 17 00:00:00 2001
From: Luis Araya <35828647+Totopo27@users.noreply.github.com>
Date: Thu, 8 Oct 2026 10:00:11 -0600
Subject: [PATCH] fix(oracle): format byte and code-point function variants as
 functions

Closes #11251
---
 apps/desktop/src/lib/sql/sqlFormatter.ts | 24 +++++++++++++++++++++++-
 packages/app-tests/sqlFormatter.test.ts  | 18 ++++++++++++++++++
 2 files changed, 41 insertions(+), 1 deletion(-)

diff --git a/apps/desktop/src/lib/sql/sqlFormatter.ts b/apps/desktop/src/lib/sql/sqlFormatter.ts
index edf3a20979..33a40927c4 100644
--- a/apps/desktop/src/lib/sql/sqlFormatter.ts
+++ b/apps/desktop/src/lib/sql/sqlFormatter.ts
@@ -28,6 +28,28 @@ function resolveClickHouseIdentifierSafeDialect(sqlFormatter: SqlFormatterModule
   return clickHouseIdentifierSafeDialect;
 }
 
+// sql-formatter's plsql dialect only registers a minimal subset of character
+// functions (e.g. SUBSTR, INSTR, LENGTH) and omits Oracle's byte/multibyte/code-point
+// variants. Register them as reserved function names so functionCase normalizes them.
+const ORACLE_EXTENDED_FUNCTION_NAMES = ["SUBSTRB", "SUBSTR2", "SUBSTR4", "SUBSTRC", "INSTRB", "INSTR2", "INSTR4", "INSTRC", "LENGTHB", "LENGTH2", "LENGTH4", "LENGTHC"];
+let oracleExtendedDialect: SqlFormatterModule["plsql"] | null = null;
+
+function resolveOracleExtendedDialect(sqlFormatter: SqlFormatterModule): SqlFormatterModule["plsql"] {
+  if (oracleExtendedDialect) return oracleExtendedDialect;
+
+  const existingFunctions = new Set(sqlFormatter.plsql.tokenizerOptions.reservedFunctionNames);
+  const extraFunctions = ORACLE_EXTENDED_FUNCTION_NAMES.filter((name) => !existingFunctions.has(name));
+
+  oracleExtendedDialect = {
+    ...sqlFormatter.plsql,
+    tokenizerOptions: {
+      ...sqlFormatter.plsql.tokenizerOptions,
+      reservedFunctionNames: [...sqlFormatter.plsql.tokenizerOptions.reservedFunctionNames, ...extraFunctions],
+    },
+  };
+  return oracleExtendedDialect;
+}
+
 export function canFormatSqlForDatabaseType(dbType: string | null | undefined): boolean {
   return dbType !== "redis" && dbType !== "victoriametrics";
 }
@@ -367,7 +389,7 @@ export async function formatSqlText(sql: string, dialect: SqlFormatDialect = "ge
           },
         }
       : options;
-  const resolvedDialect = dialect === "clickhouse" ? resolveClickHouseIdentifierSafeDialect(sqlFormatter) : undefined;
+  const resolvedDialect = dialect === "clickhouse" ? resolveClickHouseIdentifierSafeDialect(sqlFormatter) : dialect === "oracle" ? resolveOracleExtendedDialect(sqlFormatter) : undefined;
   const formatWithFallback = (input: string): string => {
     try {
       if (resolvedDialect) {
diff --git a/packages/app-tests/sqlFormatter.test.ts b/packages/app-tests/sqlFormatter.test.ts
index 18da440376..d319fafb2f 100644
--- a/packages/app-tests/sqlFormatter.test.ts
+++ b/packages/app-tests/sqlFormatter.test.ts
@@ -301,3 +301,21 @@ test("compressSqlText preserves unterminated block comments", () => {
   const sql = "DELETE FROM users /* unfinished";
   assert.equal(compressSqlText(sql), sql);
 });
+
+test("Oracle: formats byte and code-point function variants with functionCase", async () => {
+  const formattedLower = await formatSqlText("SELECT SUBSTR(a, 1), SUBSTRB(b, 1), INSTRB(c, 'x'), LENGTHB(d) FROM dual", "oracle", {
+    functionCase: "lower",
+  });
+  assert.match(formattedLower, /\bsubstr\(/);
+  assert.match(formattedLower, /\bsubstrb\(/);
+  assert.match(formattedLower, /\binstrb\(/);
+  assert.match(formattedLower, /\blengthb\(/);
+
+  const formattedUpper = await formatSqlText("select substr(a, 1), substrb(b, 1), instrb(c, 'x'), lengthb(d) from dual", "oracle", {
+    functionCase: "upper",
+  });
+  assert.match(formattedUpper, /\bSUBSTR\(/);
+  assert.match(formattedUpper, /\bSUBSTRB\(/);
+  assert.match(formattedUpper, /\bINSTRB\(/);
+  assert.match(formattedUpper, /\bLENGTHB\(/);
+});