如何 select 从 XML clob 列开始计数？

Question

我正在尝试 select XML(CLOB) 列中的标签数。

我试过

select regexp_count(diagram, 'userTask id=', 1, 'c') as "User Tasks",
       regexp_count(diagram, 'task id=', 1, 'c') as "Task"
from process_table

它有效，但是我需要比使用 regexp_count 更快地获得输出。

我试过：

select count(xt.task),
count(xt.userTask)
from process_table process
cross join xmltable(
            xmlnamespaces(default 'http://www.omg.org/spec/BPMN/20100524/MODEL'),
            '//definitions/process' passing xmltype(process.diagram) 
            columns
            task varchar2(20) path 'task',
            userTask varchar2(60) path 'userTask'
            ) xt

但是，我收到一个错误 ORA-19279: XPTY0004 - XQuery dynamic type mismatch: expected singleton sequence - got multi-item sequence

Answer 1

有几种方法可以做到这一点。一种是使用 XMLTable 查找任一子节点，获取节点名称，并计算每个节点出现的次数：

select
  count(case when xt.name = 'userTask' then name end) as userTasks,
  count(case when xt.name = 'task' then name end) as tasks
from process_table process
cross join xmltable(
  xmlnamespaces(default 'http://www.omg.org/spec/BPMN/20100524/MODEL'),
  '//definitions/process/(userTask|task)' passing xmltype(process.diagram) 
  columns
    name varchar2(20) path 'name(.)'
) xt

或者您可以使用 FLWOR 表达式同时获取两种子节点类型的计数：

select userTasks, tasks
from process_table process
cross join xmltable(
  xmlnamespaces(default 'http://www.omg.org/spec/BPMN/20100524/MODEL'),
  'let $u := count(//definitions/process/userTask)
   let $s := count(//definitions/process/task)
   return <x><u>{$u}</u><s>{$s}</s></x>'
  passing xmltype(process.diagram) 
  columns
    userTasks number path 'u',
    tasks number path 's'
) xt

但我不确定最终会比第一个选项更快。

另一种选择是使用单独的XML查询每个节点的 XPath 计数来检查：

select
  xmlquery('declare default element namespace "http://www.omg.org/spec/BPMN/20100524/MODEL";
      count(//definitions/process/userTask)'
    passing xmltype(diagram)
    returning content) as userTasks,
  xmlquery('declare default element namespace "http://www.omg.org/spec/BPMN/20100524/MODEL";
      count(//definitions/process/task)'
    passing xmltype(diagram)
    returning content) as tasks
from process_table

db<>fiddle 使用一个简单的 XML CLOB 抛出您看到的错误，以及这三种方法（包括将 XML 查询结果转换为数字）。

我也很想看看其他人的想法。

XPath counts returns ORA-06502: PL/SQL: numeric or value error when there are more than 10 rows. When I add where clause works perfectly

When I add id = some_number, or when 10 rows are displayed it works.

如果处理的行有空值，这将出错diagram；但您可以通过添加 where diagram is not null.

来排除这些

db<>fiddle

Answer 2

我认为将 XML 存储为 XMLTYPE 而不是 CLOB 会更好。然后你可以创建一个虚拟列，你甚至可以在它上面创建索引。像这样：

CREATE TABLE PROCESS_TABLE (
    DIAGRAM XMLTYPE,
    TASK_COUNT      INTEGER GENERATED ALWAYS AS (
        TO_NUMBER(CAST(XMLQUERY('count(/definitions/process/task)' PASSING BY VALUE "DIAGRAM" RETURNING CONTENT) AS VARCHAR2(100)))
    ) VIRTUAL,
    USER_TASK_COUNT INTEGER  GENERATED ALWAYS AS (
        TO_NUMBER(CAST(XMLQUERY('count(/definitions/process/userTask)' PASSING BY VALUE "DIAGRAM" RETURNING CONTENT) AS VARCHAR2(100)))
    ) VIRTUAL
)
XMLTYPE DIAGRAM STORE AS SECUREFILE BINARY XML; 

CREATE INDEX IND_TASK_COUNT ON process_table (TASK_COUNT);

如何 select 从 XML clob 列开始计数？

How to select count from XML clob column?

sql

oracle

sql-optimization