从Oracle查询中提取表引用

Question

What I am Doing:

I am not very experienced with Linux but am using it for an internal api for my company. Through the API I have an array of 79 (expecting a total of 550) PL/SQL statements for which I need to get each table referenced in each statement. I have tried reducing each statement to the text between the FROM and WHERE clause, however some statements have from in quotes in their SELECT statement and there are a large number of sub-queries so not every item following FROM is a table.

Question:

Is there a parser or a known series of regex patterns I could use to parse out the table references? I know this pattern would be vast and am hoping someone has done this work already and if so, does anyone know where it is documented?

Here is what I have tried so far:

#1 This returns the IDs for each profile I need to reference
IDs=(`echo $(APICommand_GetIDs) | jq '.[] | select(.schedule.type!="NOT_SCHEDULED") | .profileId' |sort | uniq`)

#2 Itterate through the IDs and get the sql statement for each ID. Also has a percentage complete counter
profiles=(); i=0; for ID in "${IDs[@]}"; do profiles+=("`echo $(APICommand_GetProfiles) | jq '.sql'`"); let i+=1; status=`expr $i \* 100 / ${#IDs[@]}`; printf "$status \r"; done; printf "$complete \n";

#3 Itterate through each profile to get line between "FROM" and "WHERE"
echo ${profiles[@]} | sed 's/\\r\\n/ /g' | sed 's/\\t/ /g' | sed 's/%//g' | sed 's/  / /g' | sed 's/.*FROM //' | sed 's/WHERE.*//'

However #3 returns the PL/SQL statement in multiple lines which makes it difficult to parse. I'm also having difficulty coming up with the regex patterns to only get proper table references.

Answer 1

It appears as if you are trying to use an OS script to find these dependencies. If these are pl/sql statements, and they are in a procedure, package, or function, you can use the table PUBLIC_DEPENDENCY to find all dependencies to the objects. Link the object_id's in this table to DBA_OBJECTS.

Here is a SQL example showing the dependencies on my package 'ROUTINE_PKG'.

WITH cte (owner, object_type, object_name
        , object_id, lvl)
     AS (SELECT owner
              , object_type
              , object_name
              , object_id
              , 1 lvl
           FROM dba_objects
          WHERE owner = 'BRIANL'
            AND object_name = 'ROUTINE_PKG'
         UNION ALL
         SELECT do.owner
              , do.object_type
              , do.object_name
              , do.object_id
              , lvl + 1
           FROM public_dependency pd
                INNER JOIN cte
                    ON pd.referenced_object_id = cte.object_id
                inner join dba_objects  do on do.object_id = pd.object_id where cte.lvl <= 5)
SELECT distinct *
  FROM cte order by lvl, owner, object_name;


OWNER  OBJECT_TYPE  OBJECT_NAME OBJECT_ID LVL  
BRIANL PACKAGE      ROUTINE_PKG 148593     1  
BRIANL PACKAGE BODY ROUTINE_PKG 148594     1  
BRIANL PROCEDURE    ROUTINE3    148590     2  
BRIANL PACKAGE BODY ROUTINE_PKG 148594     2  
BRIANL PROCEDURE    ROUTINE2    148591     3  
BRIANL PROCEDURE    ROUTINE1    148592     4