feat(sql): add Doris/SelectDB completion and semantic diagnostics

apps/desktop/src/lib/sql/semantic/dialect.ts

@@ -48,6 +48,12 @@ function roleForGenericQualifier(parts: string[], context: "table" | "column" |
   return "schema";
 }
 
+function roleForMysqlLikeQualifier(parts: string[], context: "table" | "column" | "routine"): "catalog" | "schema" | "table" | "package" | "unknown" {
+  if (context === "column") return "table";
+  if (context === "routine") return parts.length >= 2 ? "package" : "schema";
+  return parts.length >= 1 ? "schema" : "unknown";
+}
+
 export const SQL_SEMANTIC_DIALECTS: Record<string, SqlSemanticDialectAdapter> = {
   generic: {
     id: "generic",
@@ -77,11 +83,19 @@ export const SQL_SEMANTIC_DIALECTS: Record<string, SqlSemanticDialectAdapter> =
     projectionAliasVisibility: { where: false, groupBy: true, having: true, orderBy: true },
     normalizeIdentifier: defaultNormalize,
     quoteIdentifier: (identifier) => quoteWith(identifier, "`"),
-    qualifierRole(parts, context) {
-      if (context === "column") return parts.length >= 2 ? "table" : "table";
-      if (context === "routine") return parts.length >= 2 ? "package" : "schema";
-      return parts.length >= 1 ? "schema" : "unknown";
-    },
+    qualifierRole: roleForMysqlLikeQualifier,
+  },
+  doris: {
+    id: "doris",
+    identifierQuotes: [
+      { open: "`", close: "`" },
+      { open: '"', close: '"' },
+    ],
+    supportsAsForTableAlias: true,
+    projectionAliasVisibility: { where: false, groupBy: true, having: true, orderBy: true },
+    normalizeIdentifier: defaultNormalize,
+    quoteIdentifier: (identifier) => quoteWith(identifier, "`"),
+    qualifierRole: roleForMysqlLikeQualifier,
   },
   clickhouse: {
     id: "clickhouse",
@@ -156,11 +170,17 @@ export const SQL_SEMANTIC_DIALECTS: Record<string, SqlSemanticDialectAdapter> =
 export function sqlReferenceAnalysisDialectFor(options: { databaseType?: DatabaseType; identifierQuote?: string; fallbackDialect: string }): string {
   if (options.databaseType === "kyuubi") return "spark";
   if (options.databaseType === "kingbase" && options.identifierQuote === "`") return "mysql";
+  if (options.databaseType === "doris" || options.databaseType === "starrocks") return "doris";
   return options.fallbackDialect;
 }
 
-export function sqlSemanticDialectFor(options: { databaseType?: DatabaseType; dialect?: "mysql" | "postgres" | "sqlserver" | "clickhouse" }): SqlSemanticDialectAdapter {
+export function sqlSemanticDialectFor(options: { databaseType?: DatabaseType; dialect?: "mysql" | "postgres" | "sqlserver" | "clickhouse" | "doris" }): SqlSemanticDialectAdapter {
   if (options.databaseType === "clickhouse") return SQL_SEMANTIC_DIALECTS.clickhouse;
+  // Doris/StarRocks connections ride the editor's MySQL fallback dialect (codeMirrorSqlDialect maps
+  // them to "mysql"), so the explicit-dialect branch below would otherwise mask the doris adapter
+  // (LATERAL VIEW modeling, etc.) on the real editor path. Like clickhouse, they win on
+  // databaseType regardless of the passed dialect.
+  if (options.databaseType === "doris" || options.databaseType === "starrocks") return SQL_SEMANTIC_DIALECTS.doris;
   if (options.dialect && SQL_SEMANTIC_DIALECTS[options.dialect]) return SQL_SEMANTIC_DIALECTS[options.dialect];
   switch (options.databaseType) {
     case "postgres":
@@ -171,8 +191,6 @@ export function sqlSemanticDialectFor(options: { databaseType?: DatabaseType; di
     case "uxdb":
       return SQL_SEMANTIC_DIALECTS.postgres;
     case "mysql":
-    case "doris":
-    case "starrocks":
       return SQL_SEMANTIC_DIALECTS.mysql;
     case "sqlserver":
       return SQL_SEMANTIC_DIALECTS.sqlserver;
@@ -204,6 +222,6 @@ export function sqlSemanticDialectFor(options: { databaseType?: DatabaseType; di
  * "generic" default), matching tokenizeSqlSemantic's default and preserving the behavior callers
  * had before dialect-aware scanning existed.
  */
-export function resolveSqlDialectId(options: { databaseType?: DatabaseType; dialect?: "mysql" | "postgres" | "sqlserver" | "clickhouse" }): string {
+export function resolveSqlDialectId(options: { databaseType?: DatabaseType; dialect?: "mysql" | "postgres" | "sqlserver" | "clickhouse" | "doris" }): string {
   return options.databaseType || options.dialect ? sqlSemanticDialectFor(options).id : "mysql";
 }

apps/desktop/src/lib/sql/semantic/model.ts

@@ -464,6 +464,46 @@ function parseRowSourceList(state: ParseState, target: number, introducer: strin
   return parsed ? { sources: [parsed.source], nextIndex: parsed.nextIndex } : null;
 }
 
+function parseDorisLateralView(state: ParseState, index: number, sourceIndex: number): { source: SqlSemanticRowSource; nextIndex: number } | null {
+  if (state.dialect.id !== "doris" || state.tokens[index]?.normalized !== "lateral" || state.tokens[index + 1]?.normalized !== "view") return null;
+  // Doris's `LATERAL VIEW [OUTER] fn(...) alias AS col` -- skip the optional OUTER marker so the
+  // OUTER form models the same function columns as the plain form.
+  let functionIndex = index + 2;
+  if (state.tokens[functionIndex]?.normalized === "outer") functionIndex += 1;
+  const functionName = readQualifiedName(state.tokens, functionIndex, state.dialect);
+  if (!functionName || state.tokens[functionName.nextIndex]?.text !== "(") return null;
+  const close = findMatchingParenToken(state.tokens, functionName.nextIndex);
+  if (close < 0) return null;
+  let aliasIndex = close + 1;
+  if (state.tokens[aliasIndex]?.normalized === "as") aliasIndex += 1;
+  const alias = state.tokens[aliasIndex];
+  if (!alias || alias.kind !== "word") return null;
+  let columnIndex = aliasIndex + 1;
+  if (state.tokens[columnIndex]?.normalized === "as") columnIndex += 1;
+  const columns: string[] = [];
+  while (state.tokens[columnIndex]?.kind === "word") {
+    columns.push(state.tokens[columnIndex].text);
+    columnIndex += 1;
+    if (state.tokens[columnIndex]?.text !== ",") break;
+    columnIndex += 1;
+  }
+  const endToken = state.tokens[Math.max(aliasIndex, columnIndex - 1)] ?? alias;
+  const name = alias.text;
+  return {
+    source: {
+      id: `table-function:${sourceIndex}:${name}`,
+      kind: "table_function",
+      name,
+      alias: name,
+      qualifierParts: [name],
+      qualifiedName: functionName.name,
+      sourceSpan: { start: state.tokens[index].span.start, end: endToken.span.end },
+      columns: columns.length ? columns : undefined,
+    },
+    nextIndex: columnIndex,
+  };
+}
+
 function parseRowSourcesAtDepth(state: ParseState, sourceDepth: number, sourceIndexOffset = 0): SqlSemanticRowSource[] {
   const sources: SqlSemanticRowSource[] = [];
   let inSelectFromClause = false;
@@ -497,9 +537,20 @@ function parseRowSourcesAtDepth(state: ParseState, sourceDepth: number, sourceIn
       sources.push(...parsed.sources);
       index = parsed.nextIndex - 1;
 
-      const separator = state.tokens[parsed.nextIndex];
+      // Track the position past any LATERAL VIEW clauses so the FROM-list separator check sees
+      // the real comma after them (parsed.nextIndex points at the first "lateral" token itself).
+      let afterSources = parsed.nextIndex;
+      let lateral = parseDorisLateralView(state, afterSources, sourceIndexOffset + sources.length);
+      while (lateral) {
+        sources.push(lateral.source);
+        index = lateral.nextIndex - 1;
+        afterSources = lateral.nextIndex;
+        lateral = parseDorisLateralView(state, afterSources, sourceIndexOffset + sources.length);
+      }
+
+      const separator = state.tokens[afterSources];
       if (normalized !== "from" || separator?.text !== "," || separator.depth !== sourceDepth) break;
-      target = parsed.nextIndex + 1;
+      target = afterSources + 1;
     }
   }
   return dedupeSources(sources);

apps/desktop/src/lib/sql/semantic/tokens.ts

@@ -128,7 +128,7 @@ export function tokenizeSqlSemantic(input: string, dialectId = "mysql", options?
       continue;
     }
 
-    if (ch === "#" && dialectId === "mysql") {
+    if (ch === "#" && (dialectId === "mysql" || dialectId === "doris")) {
       index += 1;
       while (index < input.length && input[index] !== "\n" && input[index] !== "\r") index += 1;
       tokens.push(token("comment", input.slice(start, index), start, index, depth));

apps/desktop/src/lib/sql/semantic/types.ts

@@ -129,7 +129,7 @@ export interface SqlSemanticModel {
 
 export interface SqlSemanticBuildOptions {
   databaseType?: DatabaseType;
-  dialect?: "mysql" | "postgres" | "sqlserver";
+  dialect?: "mysql" | "postgres" | "sqlserver" | "clickhouse" | "doris";
   /**
    * Live CodeMirror state for the document being edited, when the caller has one (i.e. this is a
    * real editor completion request, not a pure-string test/utility call). When present and its