Hive编译SQL语句

• org.apache.hadoop.hive.ql.Driver类

• public int compile(String command, boolean resetTaskIds, boolean deferClose)方法

1. 语法分析。

      ParseDriver pd = new ParseDriver();
      ASTNode tree = pd.parse(command, ctx);
      tree = ParseUtils.findRootNonNullToken(tree);

2. 语义分析和生成计划。

// Do semantic analysis and plan generation
 BaseSemanticAnalyzer sem = SemanticAnalyzerFactory.get(queryState, tree);
 sem.analyze(tree, ctx);

 // validate the plan
 sem.validate();
 acidInQuery = sem.hasAcidInQuery();

  // get the output schema
      schema = getSchema(sem, conf);
      plan = new QueryPlan(queryStr, sem, perfLogger.getStartTime(PerfLogger.DRIVER_RUN), queryId,
        queryState.getHiveOperation(), schema);

3. 初始化FetchTask

java
// initialize FetchTask right here
if (plan.getFetchTask() != null) {
plan.getFetchTask().initialize(queryState, plan, null, ctx.getOpContext());
}