PostgreSQL 自定义自动类型转换(CAST)

转载自：https://yq.aliyun.com/articles/228271

背景

PostgreSQL是一个强类型数据库，因此你输入的变量、常量是什么类型，是强绑定的，例如

在调用操作符时，需要通过操作符边上的数据类型，选择对应的操作符。

在调用函数时，需要根据输入的类型，选择对应的函数。

如果类型不匹配，就会报操作符不存在，或者函数不存在的错误。

1. postgres=#select'1'+'1';
2. ERROR:operatorisnot unique: unknown + unknown
3. LINE 1:select'1'+'1';
4. ^
5. HINT:Couldnot choose a best candidate operator.You might need to add explicit type casts.

那么使用起来是不是很不方便呢？

PostgreSQL开放了类型转换的接口，同时也内置了很多的自动类型转换。来简化操作。

查看目前已有的类型转换：

1. postgres=# \dC+
2. List of casts
3. Source type |Target type |Function|Implicit?|Description
4. -----------------------------+-----------------------------+--------------------+---------------+-------------
5. "char"| character | bpchar |in assignment |
6. "char"| character varying | text |in assignment |
7. "char"| integer | int4 |no|
8. "char"| text | text | yes |
9. abstime | date | date |in assignment |
10. abstime | integer |(binary coercible)|no|
11. abstime | time without time zone | time |in assignment |
12. ................................
13. timestamp without time zone | timestamp with time zone | timestamptz | yes |
14. timestamp without time zone | timestamp without time zone | timestamp | yes |
15. xml | character |(binary coercible)|in assignment |
16. xml | character varying |(binary coercible)|in assignment |
17. xml | text |(binary coercible)|in assignment |
18. (246 rows)

如果你发现有些类型转换没有内置，怎么办呢？我们可以自定义转换。

当然你也可以使用这种语法，对类型进行强制转换：

1. CAST(x AS typename)
2. or
3. x::typename

如何自定义类型转换(CAST)

自定义CAST的语法如下：

1. CREATE CAST (source_type AS target_type)
2. WITH FUNCTION function_name [(argument_type [,...])]
3. [ AS ASSIGNMENT | AS IMPLICIT ]
4. CREATE CAST (source_type AS target_type)
5. WITHOUT FUNCTION
6. [ AS ASSIGNMENT | AS IMPLICIT ]
7. CREATE CAST (source_type AS target_type)
8. WITH INOUT
9. [ AS ASSIGNMENT | AS IMPLICIT ]

解释：

1、WITH FUNCTION，表示转换需要用到什么函数。

2、WITHOUT FUNCTION，表示被转换的两个类型，在数据库的存储中一致，即物理存储一致。例如text和varchar的物理存储一致。不需要转换函数。

1. Two types can be binary coercible,
2. which means that the conversion can be performed “for free” without invoking any function.
3. This requires that corresponding values use the same internal representation.
4. For instance, the types text and varchar are binary coercible both ways.
5. Binary coercibility isnot necessarily a symmetric relationship.
6. For example, the cast from xml to text can be performed for free in the present implementation,
7. but the reverse direction requires a function that performs at least a syntax check.
8. (Two types that are binary coercible both ways are also referred to as binary compatible.)

3、WITH INOUT，表示使用内置的IO函数进行转换。每一种类型，都有INPUT 和OUTPUT函数。使用这种方法，好处是不需要重新写转换函数。

除非有特殊需求，我们建议直接使用IO函数来进行转换。

1. List of functions
2. Schema|Name|Result data type |Argument data types |Type
3. ------------+-----------------+------------------+---------------------+--------
4. pg_catalog | textin | text | cstring | normal
5. pg_catalog | textout | cstring | text | normal
6. pg_catalog | date_in | date | cstring | normal
7. pg_catalog | date_out | cstring | date | normal

1. You can define a cast as an I/O conversion cast byusing the WITH INOUT syntax.
2. An I/O conversion cast is performed by invoking the output function of the source data type,
3. and passing the resulting string to the input function of the target data type.
4. In many common cases,this feature avoids the need to write a separate cast functionfor conversion.
5. An I/O conversion cast acts the same as a regular function-based cast; only the implementation is different.

4、AS ASSIGNMENT，表示在赋值时，自动对类型进行转换。例如字段类型为TEXT，输入的类型为INT，那么可以创建一个 cast(int as text) as ASSIGNMENT。

1. If the cast is marked AS ASSIGNMENT then it can be invoked implicitly when assigning a value to a column of the target data type.
2. For example, supposing that foo.f1 is a column of type text,then:
3. INSERT INTO foo (f1) VALUES (42);
4. will be allowed if the cast from type integer to type text is marked AS ASSIGNMENT,
5. otherwise not.
6. (We generally use the term assignment cast to describe this kind of cast.)

5、AS IMPLICIT，表示在表达式中，或者在赋值操作中，都对类型进行自动转换。（包含了AS ASSIGNMENT，它只对赋值进行转换）

1. If the cast is marked AS IMPLICIT then it can be invoked implicitly in any context,
2. whether assignment or internally in an expression.
3. (We generally use the term implicit cast to describe this kind of cast.)
4. For example, consider this query:
5. SELECT 2+4.0;
6. The parser initially marks the constants as being of type integer and numeric respectively.
7. Thereisno integer + numeric operatorin the system catalogs, but there is a numeric + numeric operator.
8. The query will therefore succeed if a cast from integer to numeric is available andis marked AS IMPLICIT —
9. which in fact it is.
10. The parser will apply the implicit cast and resolve the query asif it had been written
11. SELECT CAST (2 AS numeric )+4.0;

6、注意，AS IMPLICIT需要谨慎使用，为什么呢？因为操作符会涉及到多个算子，如果有多个转换，目前数据库并不知道应该选择哪个？

1. Now, the catalogs also provide a cast from numeric to integer.
2. If that cast were marked AS IMPLICIT —(which it isnot—)
3. then the parser would be faced with choosing between the above interpretation and
4. the alternative of casting the numeric constant to integer and applying the integer + integer operator.
5. Lacking any knowledge of which choice to prefer, it would give up and declare the query ambiguous.
6. The fact that only one of the two casts isimplicitis the way in which we teach the parser to prefer resolution of
7. a mixed numeric-and-integer expression as numeric;
8. there isno built-in knowledge about that.

因此，建议谨慎使用AS IMPLICIT。建议使用AS IMPLICIT的CAST应该是非失真转换转换，例如从INT转换为TEXT，或者int转换为numeric。

而失真转换，不建议使用as implicit，例如numeric转换为int。

1. Itis wise to be conservative about marking casts asimplicit.
2. An overabundance of implicit casting paths can cause PostgreSQL to choose surprising interpretations of commands,
3. or to be unable to resolve commands at all because there are multiple possible interpretations.
4. A good rule of thumb is to make a cast implicitly invokable only for information-preserving
5. transformations between types in the same general type category.
6. For example, the cast from int2 to int4 can reasonably be implicit,
7. but the cast from float8 to int4 should probably be assignment-only.
8. Cross-type-category casts, such as text to int4, are best made explicit-only.

注意事项 + 例子

不能嵌套转换。例子

1、将text转换为date

错误方法

1. create or replace function text_to_date(text) returns date as $$
2. select cast($1 as date);
3. $$ language sql strict;
4. create cast (text as date)withfunction text_to_date(text)asimplicit;

嵌套转换后出现死循环

1. postgres=#select text '2017-01-01'+1;
2. ERROR: stack depth limit exceeded
3. HINT:Increase the configuration parameter "max_stack_depth"(currently 2048kB), after ensuring the platform's stack depth limit is adequate.
4. CONTEXT: SQL function "text_to_date" during startup
5. SQL function "text_to_date" statement 1
6. SQL function "text_to_date" statement 1
7. SQL function "text_to_date" statement 1
8. ......

正确方法

1. create or replace function text_to_date(text) returns date as $$
2. select to_date($1,'yyyy-mm-dd');
3. $$ language sql strict;
4. create cast (text as date)withfunction text_to_date(text)asimplicit;

1. postgres=#select text '2017-01-01'+1;
2. ?column?
3. ------------
4. 2017-01-02
5. (1 row)

我们还可以直接使用IO函数来转换：

1. postgres=# create cast (text as date)with inout asimplicit;
2. CREATE CAST
3. postgres=#select text '2017-01-01'+1;
4. ?column?
5. ------------
6. 2017-01-02
7. (1 row)

参考

https://www.postgresql.org/docs/10/static/sql-createcast.html

本文为云栖社区原创内容，未经允许不得转载，如需转载请发送邮件至yqeditor@list.alibaba-inc.com；如果您发现本社区中有涉嫌抄袭的内容，欢迎发送邮件至：yqgroup@service.aliyun.com 进行举报，并提供相关证据，一经查实，本社区将立刻删除涉嫌侵权内容。