I have a column in a table with simple sql queries and I want to regexp_substr the table name from them. Examples of texts:
SELECT PT.PT_PARTY_NAME VALUE,PT.PT_PARTY_NAME LABEL
FROM DWH_OWNER.DWH_ACCOUNTS ACC,
DWH_OWNER.DWH_PARTIES PT
WHERE ACC.ACC_SOURCE_ID = :P_DOMVAL_REF1
AND ACC.ACC_PT_KEY = PT.PT_KEY
or
SELECT DISTINCT QRMPT_PRODUCT VALUE, QRMPT_PRODUCT LABEL FROM
DWH_OWNER.DWH_QRM_PRODUCT_TYPES ORDER BY VALUE
So, I need to find the "FROM", skip some whitespaces then get the first word. I'm not interested for the moment in the second table.
What I've done:
select
sql,
regexp_substr(upper(sql), '(\s)FROM(\s)*([[:alnum:]]|\.|_)*') tablename
from my_table_with_queries;
Output:
FROM DWH_OWNER.DWH_ACCOUNTS
FROM
DWH_OWNER.DWH_QRM_PRODUCT_TYPES
What's wrong with my qyery: It outputs the tablename with the "FROM " before the table name. I want the table name directly, without whitespaces.
Desired output:
DWH_OWNER.DWH_ACCOUNTS
DWH_OWNER.DWH_QRM_PRODUCT_TYPES
EDIT:
I managed to do regexp_substr(tablename, '(\w|_|\.)+', 1,2)
over the previous step to do the tablename clean. But It is possible to get the desired output with a single regexp?
Here should be an sqlfiddle, but the site does not work in this moment. The query:
with a as (
select 'SELECT PT.PT_PARTY_NAME VALUE,PT.PT_PARTY_NAME LABEL
FROM DWH_OWNER.DWH_ACCOUNTS ACC,
DWH_OWNER.DWH_PARTIES PT
WHERE ACC.ACC_SOURCE_ID = :P_DOMVAL_REF1
AND ACC.ACC_PT_KEY = PT.PT_KEY' sql from dual
union all
select 'SELECT DISTINCT QRMPT_PRODUCT VALUE, QRMPT_PRODUCT LABEL FROM
DWH_OWNER.DWH_QRM_PRODUCT_TYPES ORDER BY VALUE' from dual
)
select
regexp_substr(upper(sql), '\sFROM\s*(\w|\.|_)*') tablename, sql
from a;
Rather than try to write your own parser, you could let Oracle parse it for you via explain plan
, and then look at the plan table to see which objects it refers to:
declare
text varchar2(4000) := 'SELECT PT.PT_PARTY_NAME VALUE,PT.PT_PARTY_NAME LABEL
FROM DWH_OWNER.DWH_ACCOUNTS ACC,
DWH_OWNER.DWH_PARTIES PT
WHERE ACC.ACC_SOURCE_ID = :P_DOMVAL_REF1
AND ACC.ACC_PT_KEY = PT.PT_KEY';
begin
execute immediate 'explain plan for ' || text;
end;
/
select distinct object_owner, object_name
from plan_table
where object_type = 'TABLE';
OBJECT_OWNER OBJECT_NAME
------------------------------ ------------------------------
DWH_OWNER DWH_ACCOUNTS
DWH_OWNER DWH_PARTIES
As @Aleksej suggested, if the optimiser only uses an index (so the execution plan shows index access/scan without hitting the table, because all the relevant columns are in the index) then the plan table only reports the index. You could allow for that by joining to the index view; if it hits the table too it'll just report it for both:
select distinct case when pt.object_type = 'INDEX' then ai.table_owner
else pt.object_owner end as owner,
case when pt.object_type = 'INDEX' then ai.table_name
else pt.object_name end as table_name
from plan_table pt
left join all_indexes ai on ai.owner = pt.object_owner
and ai.index_name = pt.object_name
where pt.object_type in ('TABLE', 'INDEX');
You also need to make sure the plan table is empty before each explain plan call and query to avoid confusion, or set a statement ID so you can identify which tables related to the current query.
来源:https://stackoverflow.com/questions/35295458/oracle-get-table-name-from-sql-text