Hive SQL 编译过程

Driver#compile 是编译的入口

private void compile(String command, boolean resetTaskIds, boolean deferClose) throws CommandProcessorResponse 

1. 变量替换
2. 语法分析
3. 语义分析

hookRunner.runBeforeParseHook(command);

ASTNode tree;
try {
  tree = ParseUtils.parse(command, ctx);
} catch (ParseException e) {
  parseError = true;
  throw e;
} finally {
  hookRunner.runAfterParseHook(command, parseError);
}

 public static ASTNode parse(String command, Context ctx) throws ParseException {
    return parse(command, ctx, null);
}

public static ASTNode parse(
      String command, Context ctx, String viewFullyQualifiedName) throws ParseException {
    ParseDriver pd = new ParseDriver();
    ASTNode tree = pd.parse(command, ctx, viewFullyQualifiedName);
    tree = findRootNonNullToken(tree);
    handleSetColRefs(tree);
    return tree;
  }

public ASTNode parse(String command, Context ctx, String viewFullyQualifiedName)
      throws ParseException {
      
    HiveLexerX lexer = new HiveLexerX(new ANTLRNoCaseStringStream(command));
    TokenRewriteStream tokens = new TokenRewriteStream(lexer);
    if (ctx != null) {
      if (viewFullyQualifiedName == null) {
        // Top level query
        ctx.setTokenRewriteStream(tokens);
      } else {
        // It is a view
        ctx.addViewTokenRewriteStream(viewFullyQualifiedName, tokens);
      }
      lexer.setHiveConf(ctx.getConf());
    }
    HiveParser parser = new HiveParser(tokens);
    if (ctx != null) {
      parser.setHiveConf(ctx.getConf());
    }
    parser.setTreeAdaptor(adaptor);
    HiveParser.statement_return r = null;
    try {
      r = parser.statement();
    } catch (RecognitionException e) {
      e.printStackTrace();
      throw new ParseException(parser.errors);
    }

    if (lexer.getErrors().size() == 0 && parser.errors.size() == 0) {
      LOG.debug("Parse Completed");
    } else if (lexer.getErrors().size() != 0) {
      throw new ParseException(lexer.getErrors());
    } else {
      throw new ParseException(parser.errors);
    }

    ASTNode tree = (ASTNode) r.getTree();
    tree.setUnknownTokenBoundaries();
    return tree;
  }

// PreAnalyzeHook

baseSemanticAnalyzer sem = SemanticAnalyzerFactory.get(queryState, tree);

sem.analyze(tree, ctx);

// AfterAnalyzeHook

public static baseSemanticAnalyzer get(QueryState queryState, ASTNode tree) throws SemanticException {
    baseSemanticAnalyzer sem = getInternal(queryState, tree);
    if(queryState.getHiveOperation() == null) {
      String query = queryState.getQueryString();
      if(query != null && query.length() > 30) {
        query = query.substring(0, 30);
      }
      String msg = "Unknown HiveOperation for query='" + query + "' queryId=" + queryState.getQueryId();
      //throw new IllegalStateException(msg);
      LOG.debug(msg);
    }
    return sem;
  }

如果启动基于代价的优化，则使用 CalcitePlanner，否则使用 SemanticAnalyzer。

 private static baseSemanticAnalyzer getInternal(QueryState queryState, ASTNode tree)
      throws SemanticException {
    if (tree.getToken() == null) {
      throw new RuntimeException("Empty Syntax Tree");
    } else {
      HiveOperation opType = commandType.get(tree.getType());
      queryState.setCommandType(opType);
      switch (tree.getType()) {
     
      default: {  // Query
        SemanticAnalyzer semAnalyzer = HiveConf
            .getBoolVar(queryState.getConf(), HiveConf.ConfVars.HIVE_CBO_ENABLED) ?
                new CalcitePlanner(queryState) : new SemanticAnalyzer(queryState);
        return semAnalyzer;
      }
      }
    }
  }