首先让我说，根据我的研究，我什至不确定在postgres中是否可行！话虽这么说...
我正在使用具有“可配置字段”的保险数据库。许多表具有通用名称的列（字面上为field1，field2，field3等）。我正在尝试编写一个函数，该函数将解析配置表，为所需的特定数据类型确定表名和列名，然后构建，执行并返回该查询的结果。诀窍在于，结果集中的列数将根据所需数据的类型而变化。它可以是2列，十几列或任意数量。它也可以是任意数量的记录。
使用配置表映射这些列中存储的数据。因此，其中一个配置表告诉我该表的名称和该列的名称，另一个配置表告诉该字段的“名称”（从字面上看，这是普通人为数据库列命名的名称）。表中的一行指定“类型”。 “类型”不是指数据类型，而是指与其他“类型”完全不相关的不同类型的数据。将“类型”想像为通常在其自己的数据库表或列中使用的类型。无论如何，数据本身存储在这些随机分配的列之一中。同一表中的同一列可以（并且确实）具有其他，完全
可以并且将来将定义新的数据类型。我已经研究了多态表，但是看来您必须将结果集定义为函数定义的一部分。在Sybase中，这非常容易，因为Sybase函数可以从字面上返回任何结果集，而无需声明列和类型。我的问题不是产生所需的SQL语句或执行它。我还没有走那么远，但这似乎相对简单，但是我不知道如何返回执行的任何SQL语句。
对于我的POC，我编写了以下代码：
select trim('select ' || bfb.table_name || '.Policy_ID, ' || string_agg('' || bfb.table_name || '.' || bfb.column_name || ' as ' || bf.name , ',' ORDER BY replace(bo.code, 'SISCH', ''))) AS selectstatement,
'from ' || '' || bfb.table_name AS fromstatement,
-- We want the query limited only to the policy for which we are running this procedure
'where ' || '' || bfb.table_name || '.Policy_ID = ' || '1000301378' ||
-- And we only want the t_business_obect.code value that the cursor is on
' and ' || '' || bfb.table_name || '.schedule_info_code = ''' || bo.code || ''' UNION ALL ' AS wherestatement,
-- I am stripping off the prefix so we are left with only the row number (on a schedule))
replace(bo.code, 'SISCH', '') AS coderownum
from t_business_model bm join t_business_object bo on bm.model_id = bo.model_id
join T_BUSINESS_FIELD_BINDING bfb on bo.object_id = bfb.object_id
join T_BUSINESS_FIELD bf on bfb.field_id = bf.field_id
where bfb.Binding_Data_ID = 6000001969
-- Trying to code this to work on either an entire schedule or a single column
and (bo.code LIKE coalesce('SISCH', '') || '%')
group by bfb.table_name, bo.object_id, replace(bo.code, 'SISCH', ''), bo.code
order by bfb.table_name, replace(bo.code, 'SISCH', '')
哪个（当我从结果中删除最后一个UNION ALL时）产生以下SQL语句：
select T_AS_SCHEDULE_INFO.Policy_ID, T_AS_SCHEDULE_INFO.FIELD199 as NumofLawyers,T_AS_SCHEDULE_INFO.FIELD200 as PriorActsYearscovered,T_AS_SCHEDULE_INFO.FIELD198 as NumofLawyersLPL from T_AS_SCHEDULE_INFO where T_AS_SCHEDULE_INFO.Policy_ID = 1000301378 and T_AS_SCHEDULE_INFO.schedule_info_code = 'SISCH1' UNION ALL
select T_AS_SCHEDULE_INFO.Policy_ID, T_AS_SCHEDULE_INFO.FIELD199 as NumofLawyers,T_AS_SCHEDULE_INFO.FIELD200 as PriorActsYearscovered,T_AS_SCHEDULE_INFO.FIELD198 as NumofLawyersLPL from T_AS_SCHEDULE_INFO where T_AS_SCHEDULE_INFO.Policy_ID = 1000301378 and T_AS_SCHEDULE_INFO.schedule_info_code = 'SISCH2' UNION ALL
select T_AS_SCHEDULE_INFO.Policy_ID, T_AS_SCHEDULE_INFO.FIELD200 as PriorActsYearscovered,T_AS_SCHEDULE_INFO.FIELD199 as NumofLawyers,T_AS_SCHEDULE_INFO.FIELD198 as NumofLawyersLPL from T_AS_SCHEDULE_INFO where T_AS_SCHEDULE_INFO.Policy_ID = 1000301378 and T_AS_SCHEDULE_INFO.schedule_info_code = 'SISCH3' UNION ALL
select T_AS_SCHEDULE_INFO.Policy_ID, T_AS_SCHEDULE_INFO.FIELD198 as NumofLawyersLPL,T_AS_SCHEDULE_INFO.FIELD200 as PriorActsYearscovered,T_AS_SCHEDULE_INFO.FIELD199 as NumofLawyers from T_AS_SCHEDULE_INFO where T_AS_SCHEDULE_INFO.Policy_ID = 1000301378 and T_AS_SCHEDULE_INFO.schedule_info_code = 'SISCH4' UNION ALL
select T_AS_SCHEDULE_INFO.Policy_ID, T_AS_SCHEDULE_INFO.FIELD199 as NumofLawyers,T_AS_SCHEDULE_INFO.FIELD198 as NumofLawyersLPL,T_AS_SCHEDULE_INFO.FIELD200 as PriorActsYearscovered from T_AS_SCHEDULE_INFO where T_AS_SCHEDULE_INFO.Policy_ID = 1000301378 and T_AS_SCHEDULE_INFO.schedule_info_code = 'SISCH5' UNION ALL
select T_AS_SCHEDULE_INFO.Policy_ID, T_AS_SCHEDULE_INFO.FIELD198 as NumofLawyersLPL,T_AS_SCHEDULE_INFO.FIELD199 as NumofLawyers,T_AS_SCHEDULE_INFO.FIELD200 as PriorActsYearscovered from T_AS_SCHEDULE_INFO where T_AS_SCHEDULE_INFO.Policy_ID = 1000301378 and T_AS_SCHEDULE_INFO.schedule_info_code = 'SISCH6'
我希望我的函数能够执行该代码并返回结果集，而不管查询可能返回了多少列。我并不担心数据类型，因为第一列将始终是ID，其他列都将存储为文本。
最后，我试图生成一个结果集。在上面的示例中，结果集将有4列：policy_id，numoflawyers，prioractsyearscovered和numoflawyerslpl。