Is there an Oracle Function to determine the number of repeat character - sql

Question

I need to validate the number of repeat character in a email.
I try the next code who give me the percentage of repeat character, but only work if character are next to each other. So one posibily its order the email by character to get my result.
SELECT
round(((REGEXP_COUNT(regexp_replace(SUBSTR('999824123#HOTMAIL.COM',1,INSTR('989824123#HOTMAIL.COM', '#', 1)-1), '(.)\1+','&'),'&')+length(SUBSTR('989824123#HOTMAIL.COM',1,INSTR('989824123#HOTMAIL.COM', '#', 1)-1)) - length(regexp_replace(SUBSTR('989824123#HOTMAIL.COM',1,INSTR('989824123#HOTMAIL.COM', '#', 1)-1), '(.)\1+','\1')))* 100)/length(SUBSTR('989824123#HOTMAIL.COM',1,INSTR('989824123#HOTMAIL.COM', '#', 1)-1)),2) AS PORCENTAJE_IGUAL
FROM DUAL A;
I expect 60% of repeat character for this email 989824123#HOTMAIL.COM. not incluing domain.
please Help.
PD: sorry for the bad english

Answer 1

Numbers 9, 8, 2 repeats in email, so we have 6 characters (9, 9, 8, 8, 2, 2) which repeats and 3 unique (1, 3, 4). 6/9 gives us 66,67%.
You can use this query to count this:
with
t(email) as (select '989824123#hotmail.com' from dual),
a(email) as (select substr(email, 1,instr(email, '#', 1)-1) from t),
l as (select substr(email, level, 1) ltr from a connect by level <= length(email))
select sum(case when cnt <> 1 then cnt end) / sum(cnt)
from (select ltr, count(1) cnt from l group by ltr)
I cut domain, then in subquery l I divided string into one-letter rows, rest was only to count non-unique chars and divide by number of all chars.
edit:
how do you apply something like this in a update or select for a large
scale data base with many email?
You can create function:
create or replace function rpt_similarity(i_email in varchar2) return number is
v_email varchar2(100);
v_ret number;
begin
v_email := substr(i_email, 1, instr(i_email, '#', 1) - 1);
with l as (
select substr(v_email, level, 1) ltr
from dual
connect by level <= length(v_email))
select sum(case when cnt <> 1 then cnt end) / sum(cnt)
into v_ret
from (select ltr, count(1) cnt from l group by ltr);
return v_ret;
end;
and use it like here:
select rpt_similarity('abxabc#pqr.com') from dual;
or:
select rpt_similarity(email) from your_table;
Also you can use above solution in select directly, without function, here is the example:
create table test(id, email) as (
select 101, '989824123#hotmail.com' from dual union all
select 102, 'hsimpson#gmail.com' from dual union all
select 103, 'msimpson#gmail.com' from dual union all
select 104, 'bsimpson121314#hotmail.com' from dual union all
select 105, 'abxabx#hotmail.com' from dual );
with
a(id, email) as (select id, substr(email, 1,instr(email, '#', 1)-1) from test),
l as (
select id, email, substr(email, level, 1) ltr from a
connect by level <= length(email)
and prior id = id and prior sys_guid() is not null)
select id, email, sum(case when cnt <> 1 then cnt end) / sum(cnt)
from (select id, email, ltr, count(1) cnt from l group by id, ltr, email)
group by id, email;
connect by queries tends to be slow for large sets of data. Maybe you can adapt your regexp functions and it will be faster. I tried to do it, but your regexp_replace changes 99 into $ and 999 also into one $.