以下内容来自Oracle FAQ writen By Kevin,关于ANYDATA类型在项目中的应用。

My newest project needed to create a record keeping component that would keep track of balancing results over time. Sounded good, meant my customers could look back over time to see how things went, and it provided persistent evidence of good results. I figured on adding two routines (save/get) to a logging package already in existence to save balancing data via an autonomous transaction and retrieve it when needed. But in writing these routines it dawned on me that they would destroy the reusable nature of the logging package. Finally, a real life use for ANYDATA.

ANYDATA是一种oracle数据类型(实际上是SYS用户下的对象)。

ANYDATA is an Oracle data type (an object actually), that is as they say “self describing”. Your guess is as good as mine as to what that really means. The party line goes something like: an ANYDATA data item contains data along with a data type descriptor (including data type name) of what the data looks like. Well… I sort of understand. But as usual, looking at some working code seems like the best way to really understand, so here is a quick and dirty introduction to ANYDATA along with how I used it in a real life scenario.

Here is a simple table that has a column defined as ANYDATA. 在列定义中使用ANYDATA

create table temp1
(
a number not null
,b sys.anydata
)
/

OK, not too intimidating. The idea here is that we can store (with a few exceptions), any data we want in this ANYDATA column. It is actually pretty simple to put data into an ANYDATA data item. You just use one of the handy methods of the ANYDATA object. Here we can see the storage of number, date, and varchar2 values. Notice the use of the “convert-XXX” functions. Turns out there is a convert function for almost every type of data Oracle supports.

insert into temp1 values (1,sys.anydata.convertnumber(1))
/
insert into temp1 values (2,sys.anydata.convertdate(sysdate))
/
insert into temp1 values (3,sys.anydata.convertvarchar2('a'))
/ commit
/ SQL> select count(*) from temp1
2 / COUNT(*)
----------
3 1 row selected.

Yep, three rows went in. Wonder what it looks like?

SQL> col b format a20 trunc
SQL> select * from temp1
2 / A B()
--- --------------------
1 ANYDATA()
2 ANYDATA()
3 ANYDATA() 3 rows selected.

Hmm.. not much help there. What else can this thing tell me.

SQL> col typename format a20 trunc
SQL> select temp1.*,sys.anydata.gettypename(temp1.b) typename from temp1
2 / A B() TYPENAME
--- -------------------- --------------------
1 ANYDATA() SYS.NUMBER
2 ANYDATA() SYS.DATE
3 ANYDATA() SYS.VARCHAR2 3 rows selected.

We used one of the many methods of the ANYDATA object type to find out the name of the data type of each specific column value. Yes, kind of neat, but how do I see the data? Well, that gets a little more involved. Because the data stored is actually of many different types, you can’t just select it. You will have to write some PL/SQL functions to help out.

Each data type convert method like the three we used in the inserts above, of the ANYDATA object type, has a corresponding get method that will give what you stored, back to you. Problem is, these get methods are not simple functions they are procedures so we need to write our own wrapper functions to effectively use these methods. Something like this package will do fine.

Each specific function in our package accepts an ANYDATA input and returns a specific output for its type. In the package body you will see the corresponding get methods that are invoked.

create or replace
package pkg_temp1
as function getnumber (anydata_p in sys.anydata) return number;
function getdate (anydata_p in sys.anydata) return date;
function getvarchar2 (anydata_p in sys.anydata) return varchar2; end;
/
show errors create or replace
package body pkg_temp1
as function getnumber (anydata_p in sys.anydata) return number is
x number;
thenumber_v number;
begin
x := anydata_p.getnumber(thenumber_v);
return (thenumber_v);
end; function getdate (anydata_p in sys.anydata) return date is
x number;
thedate_v date;
begin
x := anydata_p.getdate(thedate_v);
return (thedate_v);
end; function getvarchar2 (anydata_p in sys.anydata) return varchar2 is
x number;
thevarchar2_v varchar2(4000);
begin
x := anydata_p.getvarchar2(thevarchar2_v);
return (thevarchar2_v);
end; end;
/
show errors

With this package in place we can now see our data.

col thevalue format a20 trunc
select temp1.*,sys.anydata.gettypename(temp1.b) typename
,case
when sys.anydata.gettypename(temp1.b) = 'SYS.NUMBER' then
to_char(pkg_temp1.getnumber(temp1.b))
when sys.anydata.gettypename(temp1.b) = 'SYS.DATE' then
to_char(pkg_temp1.getdate(temp1.b),'dd-mon-rrrr hh24:mi:ss')
when sys.anydata.gettypename(temp1.b) = 'SYS.VARCHAR2' then
pkg_temp1.getvarchar2(temp1.b)
end thevalue
from temp1
/ A B() TYPENAME THEVALUE
--- -------------------- -------------------- --------------------
1 ANYDATA() SYS.NUMBER 1
2 ANYDATA() SYS.DATE 25-may-2007 16:35:57
3 ANYDATA() SYS.VARCHAR2 a 3 rows selected.

Turns out we can even store user defined datatypes (objects and collections) in an ANYDATA column. Pay particular attention to the fact that we used the OID option of the create type commands below. This is one gotcha of ANYDATA.

If we store an object or collection in an ANYDATA column and later we drop the object or collection, then any ANYDATA column value dependent upon the dropped object will become inaccessible. So much for "self describing" eh. Even if we recreate the object and/or collection later, exactly as it was before, we still won't be able to access the data.

The fact is, that object types and collections have at least two identifications, a name, and an OID. Depending upon which one you use, you may or may not be able to find your collection data later.

Oracle always creates object types with an OID value. You don't normally supply one, so Oracle defaults a value for you. It uses a GUID construct for this (read up on sys_guid() if you want to know), which in theory is a string that is unique around the world (yep). When you recreate an object or collection (same name, same attributes, same everything it had before), Oracle still creates it with a different OID. Some internal Oracle functions reference the OID not the object name, and ANYDATA is one of them it seems. So even if after dropping an object or collection that an ANYDATA column value was dependent upon, you recreate the object or collection with the same definition, the previously inaccessible ANYDATA column value still won't be accessible because the stored data is remembering the old OID value.

But, if you create your object using an OID value, then you can always recreate it with the same OID value and the ANYDATA column value will recognize it and you are safe. There are presumably two reasons for needing OID values as regards types: a) to ensure an ANYDATA value will always work, and b) to be able to share types across database instances.

create or replace type o_temp1 oid '3150D5BF61DE33EDE0440003BA62E91A' is object (a number,b number,c number)
/ insert into temp1 values (4,sys.anydata.convertobject(o_temp1(1,2,3)))
/ create or replace type c_temp1 oid '3150D5BF61DF33EDE0440003BA62E91A' is table of o_temp1
/ set serveroutput on
declare
c_temp1_v c_temp1;
begin
select cast(multiset(select * from (
select 1 c1,2 c2,3 c3 from dual union all
select 4 c1,5 c2,6 c3 from dual union all
select 7 c1,8 c2,9 c3 from dual
)
) as c_temp1
)
into c_temp1_v
from dual;
dbms_output.put_line(c_temp1_v.count);
insert into temp1 values (5,sys.anydata.convertcollection(c_temp1_v));
end;
/

Of course we will need two more functions for these two additional data type definitions we just created.

   function get_o_temp1 (anydata_p in sys.anydata) return o_temp1;
function get_c_temp1 (anydata_p in sys.anydata) return c_temp1; function get_o_temp1 (anydata_p in sys.anydata) return o_temp1 is
x number;
o_temp1_v o_temp1;
begin
x := anydata_p.getobject(o_temp1_v);
return (o_temp1_v);
end; function get_c_temp1 (anydata_p in sys.anydata) return c_temp1 is
x number;
c_temp1_v c_temp1;
begin
x := anydata_p.getcollection(c_temp1_v);
return (c_temp1_v);
end;

After adding these functions to our helper package we can see the data.

col typename format a20 trunc
select temp1.*,sys.anydata.gettypename(temp1.b) typename from temp1
/ A B() TYPENAME
--- -------------------- -------------------
1 ANYDATA() SYS.NUMBER
2 ANYDATA() SYS.DATE
3 ANYDATA() SYS.VARCHAR2
4 ANYDATA() KMEADE.O_TEMP1
5 ANYDATA() KMEADE.C_TEMP1 5 rows selected. COL AC_TEMP1 FORMAT A62
select temp1.*
,case when sys.anydata.gettypename(temp1.b) = 'SYS.NUMBER' then pkg_temp1.getnumber(temp1.b) end anumber
,case when sys.anydata.gettypename(temp1.b) = 'SYS.DATE' then pkg_temp1.getdate(temp1.b) end adate
,case when sys.anydata.gettypename(temp1.b) = 'SYS.VARCHAR2' then pkg_temp1.getvarchar2(temp1.b) end avarchar2
,case when substr(sys.anydata.gettypename(temp1.b),instr(sys.anydata.gettypename(temp1.b),'.')+1) = 'O_TEMP1' then
pkg_temp1.get_o_temp1(temp1.b) end ao_temp1
,case when substr(sys.anydata.gettypename(temp1.b),instr(sys.anydata.gettypename(temp1.b),'.')+1) = 'C_TEMP1' then
pkg_temp1.get_c_temp1(temp1.b) end ac_temp1
from temp1
/ A B() ANUMBER ADATE AVARCHAR2 AO_TEMP1(A, B, C) AC_TEMP1(A, B, C)
--- -------------------- ---------- --------- -------------------- -------------------- ------------------------------
1 ANYDATA() 1
2 ANYDATA() 25-MAY-07
3 ANYDATA() a
4 ANYDATA() O_TEMP1(1, 2, 3)
5 ANYDATA() C_TEMP1(O_TEMP1(1, 2, 3), O_TE 5 rows selected.

(sorry, cut off the collection rows because of space on the page, they are there which you will see if you run the test cases)

So, this is nice and all, but why would anyone what to go to all this trouble just to save some data in a table. Well, most of the time you won’t. But there are times when you won’t know what data you are getting in advance, or more likely, you don’t want to know. The balancing data retention component I mentioned in opening is one such case.

Consider this situation: an application system wants to save the data it used to balance for posterity. It has this data as a set of rows. What do you do? Most people would create a table that maps to the data and the tell the application system “INSERT HERE”. Sounds OK, till then next application system comes along. They want to do the same thing except their data don’t look like the table you created for balancing for those other guys, so now what do you do? Well you got two choices:

1) create another table that maps to this new application’s data and tell them to “INSERT HERE”. This works but you can see it suffers from the fact that it means each application system that comes on board will need to create new objects to support balancing data retention, and this should after all be a common function done in a consistent way, but you have no common function any more.

2) Or, you decide to implement “A STANDARD” for balancing data. Well standards aren’t bad but you’ll have two problems with this idea: a) you have to build a standard which in light of the fact that the first application already did something and would have to change if you make the standard something different from what they did, means there will be pressure from several corners to adopt application A balancing as the standard, b) somebody will always be able to come up with a hard requirement that won’t fit your standard for balancing.

But, with a table something like this one:

create table hhi_acmr_job_run_bal
(
hhi_acmr_job_run_bal_id number not null
, hhi_acmr_job_run_id number not null
, descriptor varchar2(100) not null
, balancing_dataset sys.anydata
)
tablespace ACMRDIMS_TABLE
/

You can build a reusable store/retrieve mechanism for balancing. It provides a simple framework and makes clear what the responsibilities of the reusable code are, and what the responsibilities of each application system are. You expose one interface and create a guide that tells application developers what steps they need to go through to correctly save their balancing work. In general they will need to do the following:

1) Define the layout of their balancing objects. They might in fact have more than one set of balancing data.
2) Create an object type and collection that maps to each layout.
3) Write functions in their applications that deal with the specifics of their layout. These would include:
. a) a select function (or view) to gather balancing data into rows
. b) a collect function to convert these rows to a collection
. c) a save function that invokes the ANYDATA convertcollection call and passes the result to the balancing save code.
. d) a get function to do to opposite of (C) using the ANYDATA getcollection call
. e) maybe a view that hides everything

The advantage to using ANYDATA here is that the balancing retention component remains simple and easy to use because it doesn’t do a whole lot and doesn’t care about the specifics of data types of each application. Additionally, you still have a standard in place because you have created a clear division of work and a consistent process to get the work done. Best of all, you have not tied anyone’s hands as to what they build.

Well, that is about it. One final note. Not all data types are supported by ANYDATA, even though the documentation may suggest they are. Of particular note are XML data, CLOB/BLOB data. Oracle is working on it. DESC SYS.ANYDATA to see what is available and remember, CLOB/BLOB don't work event hough it looks like it will.

Kevin

An Introduction to ANYDATA的更多相关文章

  1. A chatroom for all! Part 1 - Introduction to Node.js(转发)

    项目组用到了 Node.js,发现下面这篇文章不错.转发一下.原文地址:<原文>. ------------------------------------------- A chatro ...

  2. Introduction to graph theory 图论/脑网络基础

    Source: Connected Brain Figure above: Bullmore E, Sporns O. Complex brain networks: graph theoretica ...

  3. INTRODUCTION TO BIOINFORMATICS

    INTRODUCTION TO BIOINFORMATICS      这套教程源自Youtube,算得上比较完整的生物信息学领域的视频教程,授课内容完整清晰,专题化的讲座形式,细节讲解比国内的京师大 ...

  4. mongoDB index introduction

    索引为mongoDB的查询提供了有效的解决方案,如果没有索引,mongodb必须的扫描文档集中所有记录来match查询条件的记录.然而这些扫描是没有必要,而且每一次操作mongod进程会处理大量的数据 ...

  5. (翻译)《Hands-on Node.js》—— Introduction

    今天开始会和大熊君{{bb}}一起着手翻译node的系列外文书籍,大熊负责翻译<Node.js IN ACTION>一书,而我暂时负责翻译这本<Hands-on Node.js> ...

  6. Introduction of OpenCascade Foundation Classes

    Introduction of OpenCascade Foundation Classes Open CASCADE基础类简介 eryar@163.com 一.简介 1. 基础类概述 Foundat ...

  7. 000.Introduction to ASP.NET Core--【Asp.net core 介绍】

    Introduction to ASP.NET Core Asp.net core 介绍 270 of 282 people found this helpful By Daniel Roth, Ri ...

  8. Introduction to Microsoft Dynamics 365 licensing

    Microsoft Dynamics 365 will be released on November 1. In preparation for that, Scott Guthrie hosted ...

  9. RabbitMQ消息队列(一): Detailed Introduction 详细介绍

     http://blog.csdn.net/anzhsoft/article/details/19563091 RabbitMQ消息队列(一): Detailed Introduction 详细介绍 ...

  10. Introduction - SNMP Tutorial

    30.1 Introduction In addition to protocols that provide network level services and application progr ...

随机推荐

  1. SpringMVC05——SSM整合

    整合SSM 需求:熟练掌握MySQL数据库,Spring,JavaWeb及MyBatis知识,简单的前端知识 CREATE DATABASE `ssmbuild`; USE `ssmbuild`; D ...

  2. [转帖]K8s Pod Command 与容器镜像 Cmd 启动优先级详解

    https://cloud.tencent.com/developer/article/1638844 前言 创建 Pod 时,可以为其下的容器设置启动时要执行的命令及其参数.如果要设置命令,就填写在 ...

  3. 部分MySQL的SQL信息整理

    模块补丁信息查看 select su as 补丁模块, count(1) as 数量 from gsppatchlog where TO_DAYS( NOW( ) ) - TO_DAYS(deploy ...

  4. [转帖]TiDB的tikv节点的压缩算法

    简介:TiDB的tikv节点实用的RocksDB,RocksDB的默认压缩算法为:[no:no:lz4:lz4:lz4:zstd:zstd] RocksDB 每一层数据的压缩方式,可选的值为:no,s ...

  5. [转帖]分析redis 大key

    http://www.lishuai.fun/2023/05/05/redis-bigkey/#/%E5%AE%89%E8%A3%85 redis-rdb-tools 是一个 python 的解析 r ...

  6. [转帖]Kafka-Kraft 模式架构部署

    news文章来源: Kafka-Kraft 模式架构部署 Kafka网址:https://kafka.apache.org/ PS:因环境原因此文档内端口都有修改! 1.去官网下载二进制包 PS:3. ...

  7. [转帖]nginx性能和软中断

    https://plantegg.github.io/2022/11/04/nginx%E6%80%A7%E8%83%BD%E5%92%8C%E8%BD%AF%E4%B8%AD%E6%96%AD/ n ...

  8. Docker 运行 MongoDB的简单办法

    Docker 运行 MongoDB的简单办法 第一步拉取镜像 docker pull mongo 第二步创建自己的目录 地址 10.24.22.240 创建目录 mkdir /mongodb 第三步 ...

  9. ELK运维文档

    Logstash 目录 Logstash Monitoring API Node Info API Plugins Info API Node Stats API Hot Threads API lo ...

  10. 《SAIS Supervising and Augmenting Intermediate Steps for Document-Level Relation Extraction》论文阅读笔记

    代码   原文地址   预备知识: 1.什么是标记索引(token indices)? 标记索引是一种用于表示文本中的单词或符号的数字编码.它们可以帮助计算机理解和处理自然语言.例如,假如有一个字典{ ...