mysql优化---订单查询优化:异步分页处理

Posted 2020-08-31 聚变归来

订单分页查询:

老的代码是顺序执行查询数据和计算总记录数,但是如果条件复杂的话(比如关联子表)查询的时间要超过20s种

 

 

    public static PagedList<Map<String, Object>> query(ITemplateService service, Identity tenantId, Identity userId, String entityName,
                                                       Map<String, Object> params, String columns, TCondition cond) {
        int page = WebHelper.getPageNo(params);
        int pageSize = WebHelper.getPageSize(params);
        String orderColumn = (String) params.get(JqgridConstant.ORDER_COLUMN); // 排序字段
        String orderSord = (String) params.get(JqgridConstant.ORDER_METHOD); // 排序方式，desc or asc

        handleOrderByColumn(cond, orderColumn, orderSord);
        int totalCount = service.getBaseService().query4Count(tenantId, entityName, cond);
        List<Map<String, Object>> list = service.query(columns, entityName, cond, tenantId, userId, pageSize, page);
        Translator.prepare(list, tenantId, service); // TODO
        return new PagedList<>(page, pageSize, totalCount, list);
    }

 

 

优化方法:

1.通过新启动一个线程来同时做之前需要顺序执行的两个Sql查询,最后等待全部计算完成,统一进行返回

2.对于一些特别复杂的条件的查询,如果内容的条数少于PageSize,那么计算总条数的sql就是不需要执行,可以用返回的list的szie当做总记录数

public static PagedList<Map<String, Object>> queryAsyn(ITemplateService service, Identity tenantId, Identity userId, String entityName,
                                                           Map<String, Object> params, String columns, TCondition cond) {
        int page = WebHelper.getPageNo(params);
        int pageSize = WebHelper.getPageSize(params);
        String orderColumn = (String) params.get(JqgridConstant.ORDER_COLUMN); // 排序字段
        String orderSord = (String) params.get(JqgridConstant.ORDER_METHOD); // 排序方式，desc or asc
        ExecutorService slaver = Executors.newSingleThreadExecutor();
        FutureTask<Integer> totalCountFuture = new FutureTask<>(new TotalCountJob(service, tenantId, entityName, cond));
        slaver.execute(totalCountFuture);
        handleOrderByColumn(cond, orderColumn, orderSord);
        slaver.shutdown();
        //主线程来取数据
        long time1 = System.nanoTime();
        List<Map<String, Object>> list = service.query(columns, entityName, cond, tenantId, userId, pageSize, page);
        long time2 = System.nanoTime();
        long diff = time2 - time1;
        logger.debug("查询方案统计-----查询分页list部分,用时:{}s,条件:{}", translateToSecond(diff), cond);
        Integer totalCount = null;
        int listSize = list.size();
        if (listSize < pageSize) {
            logger.info("本次查询不需要sql进行count操作");
            totalCount = listSize + (page - 1) * pageSize;
            slaver.shutdownNow();
        } else {
            try {
                //没做完就等着
                totalCount = totalCountFuture.get();
            } catch (Exception e) {
                totalCountFuture.cancel(true);
                logger.error("totalCount发生异常", e);
            }
        }

        Translator.prepare(list, tenantId, service);
        return new PagedList<>(page, pageSize, totalCount, list);
    }

    private static double translateToSecond(long diff) {
        return diff * 0.000000001;
    }


    static class TotalCountJob implements Callable<Integer> {

        private String tableName;
        private TCondition condition;
        private Identity tenantId;
        private ITemplateService service;

        public TotalCountJob(ITemplateService service, Identity tenantId, String tableName, TCondition condition) {
            this.service = service;
            this.tableName = tableName;
            this.condition = condition;
            this.tenantId = tenantId;
        }

        @Override
        public Integer call() throws Exception {
            long time1 = System.nanoTime();
            Integer totalCount = service.getBaseService().query4Count(tenantId, tableName, condition);
            long time2 = System.nanoTime();
            long diff = time2 - time1;
            logger.debug("查询方案统计-----查询分页count部分,用时:{}s,条件:{}", translateToSecond(diff), condition);
            return totalCount;
        }
    }

这是第一次优化的文章,欢迎访问:

 http://www.cnblogs.com/victor2302/p/6073821.html