special characters and characters capture from a string oracle

Question

I am trying to capture the table names starting with some patterns and find the total counts for each group but i would specifically like to capture ,below is an example

I think the best possible way to capture the above is using REGEXP ?

Query 1 - Initial

SELECT
   owner AS schema_name, 
   object_name, 
   object_type,
   REGEXP_REPLACE(OBJECT_NAME, '^([A-Z0-9$]{1,})_.*', '\1') as BEGINNING,
   count(*),
   round(100*ratio_to_report(count(*)) over (), 4) percentage 
FROM
   dba_objects 
GROUP BY
   owner,
   object_name,
   object_type,
   REGEXP_REPLACE(OBJECT_NAME, '^([A-Z0-9$]{1,})_.*', '\1') 
ORDER BY
   percentage desc;

Query 1 - existing result

SCHEMA OBJECT_NAME       OBJECT_TYPE  BEGINNING COUNT(*) PERCENT
STG    AB_01_CUST_ENRLMT TABLE          AB          1    .00001
STG    DKS_SD_PRDCT_DHSS TABLE          DKS         1    .00001
STG    ABC10_CUST_ENRLMT  TABLE         ABC10       1    .00001

Query 1 - expected results

 SNO SCHEMA OBJECT_NAME       OBJECT_TYPE  BEGINNING             COUNT(*) PERCENT
  1  STG    AB_01_CUST_ENRLMT TABLE        AB_01                 1       .00001
  2  STG    DKS_SD_PRDCT_DHSS TABLE        DKS_SD                1       .00001
  3  STG    ABC10_CUST_ENRLMT TABLE        ABC10_CUST            1       .00001
  4  STG    #Tableau_6_sid:15 TABLE        #Tableau_6            4000    1.5
  5  STG    /157d_PoI_12 TABLE             /157d_PoI             5000    1.6
  6  STG    JAVA/157d_Ph TABLE             JAVA/157_Ph           5000    1.6
  7  STG    STU$BA_COENT_123 TABLE        STU$BA_COENT           5000    1.5

how to achieve as per expected results

After suggestions i have incorporated the below logic and re ran , unfortunately i am not getting the desired result

Query 2 - Modified

SELECT
   owner AS schema_name,
   object_name,
   object_type,
   REGEXP_REPLACE(OBJECT_NAME, '^([A-Z0-9$]{1,})_([A-Z0-9$]{1,})_.*', '\1_\2')as BEGINNING,
   count(*),
   round(100*ratio_to_report(count(*)) over (), 4) percentage 
FROM
   dba_objects 
GROUP BY
   owner,
   object_name,
   object_type,
   REGEXP_REPLACE(OBJECT_NAME, '^([A-Z0-9$]{1,})_.([A-Z0-9$]{1,})_.*', '\1_\2')
ORDER BY
   percentage desc;

Query 2 - Results - Not as expected

 SNO SCHEMA OBJECT_NAME       OBJECT_TYPE  BEGINNING           COUNT(*) PERCENT
 1   STG    AB_01_CUST_ENRLMT TABLE        AB_1                  1      .00001
 2   STG    DKS_SD_PRDCT_DHSS TABLE        DKS_D                 1      .00001
 3   STG    ABC_25_ENRLMT     TABLE        ABC_5                 1      .00001
  4  STG    #Tableau_6_sid:15 TABLE        #Tableau_6            4000    1.5
  5  STG    /157d_PoI_12 TABLE             /157d_PoI             5000    1.6
  6  STG    JAVA/157d_Ph TABLE             JAVA/157_Ph           5000    1.6
  7  STG    STU$BA_COENT_123 TABLE        STU$BA_COENT           5000    1.5
 

what is that i need to correct in the REGEX logic ? , i have come across some special cases

Query 2 - Results - Not as expected

Answer 1

Your regex is fetching the first occurrence of (char digit) pattern. That is; ([A-Z0-9$]{1,})

The (...) bracets are used for grouping and in the third parameter of the REGEXP_REPLACE you are using this one group as replacement of the entire original string. There are other ways to solve this, but if you wish to continue using the same approach, you can extend your regex to the following;

REGEXP_REPLACE(OBJECT_NAME, '^([A-Z0-9$]{1,})_.([A-Z0-9$]{1,})_.*', '\1_\2')

Edited regex;

REGEXP_REPLACE(OBJECT_NAME, '^([A-Z0-9$]{1,})_([A-Z0-9$]{1,})_.*', '\1_\2')

This should replace the string with the (first_group)_(second_group) which is mentioned as \1_\2 in the REGEXP_REPLACE parameter.

Answer 2

REGEXP_REPLACE() might be applied with two kind of patterns depending on the number of underscores whether less than or equal to two or more :

SELECT object_name,
       CASE 
       WHEN REGEXP_COUNT(object_name,'_') <= 2 THEN
            REGEXP_REPLACE(object_name,'(*.)(_.*){1,}(_.*)$', '\1\2')
       ELSE REGEXP_REPLACE(object_name,'^([[:alnum:]]{1,})_([[:alnum:]]{1,}).*', '\1_\2')     
        END AS new_object_name
  FROM user_objects

Demo