Oracle how to instr for tab or end of line using regexp_instr - sql

Question

I have the following code
declare
l_clob clob;
l_line varchar2(32767);
l_field varchar2(32767);
l_line_start pls_integer := 1;
l_line_end pls_integer := 1;
l_field_start pls_integer := 1;
l_field_end pls_integer := 1;
begin
select response_clob
into l_clob
from xxhr.xxhr_web_service_response
where response_id = 290;
-- Loop through lines.
loop
l_line_end := dbms_lob.instr(l_clob, chr(10), l_line_start, 1);
l_line := dbms_lob.substr(l_clob, l_line_end - l_line_start + 1, l_line_start);
-- If this is a line with fields and not web service garbage.
if substr(l_line, 1, 1) = '"' then
l_field_start := 2;
-- Loop through fields.
loop
l_field_end := instr(l_line, chr(9), l_field_start, 1);
l_field := substr(l_line, l_field_start, l_field_end - l_field_start);
dbms_output.put(l_field || ',');
l_field_start := l_field_end + 1;
exit when l_field_end = 0;
end loop;
dbms_output.put_line('');
end if;
l_line_start := l_line_end + 1;
exit when l_line_end = 0;
end loop;
end;
with which I'm trying to parse this clob test data:
LINE_TEXT
"PERSON_ID_NUMBER 30000 1223445454"
"PERSON_DOB 30000 01-01-1900"
The clob data is tab separated and has a chr(10) at the end. I'm not familiar with regexp_instr, but currently I'm only using instr to search for the tab separators; so it's missing the end of line field and producing:
PERSON_ID_NUMBER,30000,,
PERSON_DOB,30000,,
How can I change the instr into a regexp_instr to also look for the end of line character in addition to the tab and then correctly pick up the last field?
I need the function to be performant, since it is parsing large files.

Answer 1

You can split the line of CLOB column by converting to char, and then apply regexp_substr() as
with t as
(
select level as row_num, regexp_substr(to_char(t.line_text),'^.*$',1,level,'m') as str
from tab t
connect by level <= length (to_char(t.line_text))
- length (replace (to_char(t.line_text), chr (10))) + 1
)
select row_num, regexp_replace(str,'[[:space:]]+',',') as str
from t;
ROW_NUM STR
------- -----------------------------------------
1 PERSON_ID_NUMBER,30000,1223445454
2 PERSON_DOB,30000,01-01-1900
Demo
Edit : even works without to_char() conversion, if your CLOB is huge then you need to split step by step by substr(str,1,4000), substr(str,4001,8000) ...
with t as
(
select level as row_num, regexp_substr(substr(t.line_text,1,4000),'^.*$',1,level,'m') str
from tab t
connect by level <= length (substr(t.line_text,1,4000))
- length (replace(substr(t.line_text,1,4000), chr (10))) + 1
)
select row_num, regexp_replace(substr(str,1,4000),'[[:space:]]+',',') as str
from t

Answer 2

Fixed it with:
declare
l_clob clob;
l_line varchar2(32767);
l_field varchar2(32767);
l_line_start pls_integer := 1;
l_line_end pls_integer := 1;
l_field_start pls_integer := 1;
l_field_end pls_integer := 1;
begin
select response_clob
into l_clob
from xxhr.xxhr_web_service_response
where response_id = 290;
-- Loop through lines.
loop
l_line_end := dbms_lob.instr(l_clob, chr(10), l_line_start, 1);
l_line := dbms_lob.substr(l_clob, l_line_end - l_line_start + 1, l_line_start);
-- If this is a line with fields and not web service garbage.
if substr(l_line, 1, 1) = '"' then
l_field_start := 2;
-- Loop through fields.
loop
l_field_end := instr(l_line, chr(9), l_field_start, 1);
l_field := substr(l_line, l_field_start, l_field_end - l_field_start);
dbms_output.put(l_field || ',');
exit when l_field_end = 0;
l_field_start := l_field_end + 1;
end loop;
l_field := substr(l_line, l_field_start);
dbms_output.put_line(l_field);
end if;
l_line_start := l_line_end + 1;
exit when l_line_end = 0;
end loop;
end;