How to compare single byte and multi bytes string? - oracle

Question

I have two strings, one is input from client and other is data in table. Two strings seem to be the same, but has different hex value when I try CAST_TO_RAW.
SELECT UTL_RAW.CAST_TO_VARCHAR2('3539352F47502D0A41544258484E') INPUT,
UTL_RAW.CAST_TO_NVARCHAR2('003500390035002F00470050002D000D000A00410054004200580048004E') DTA
FROM DUAL;
/*input and data seem same*/
Consider that two string is the same. How I pass this case and compare them in query like this:
SELECT A.DATA,
A.ORTHER_COL
FROM MYTABLE A
WHERE A.DATA = INPUT;
I tried TO_SINGLE_BYTE but it not work (because of its different LENGTHB):
SELECT *
FROM DUAL
WHERE TO_SINGLE_BYTE(UTL_RAW.CAST_TO_VARCHAR2('3539352F47502D0A41544258484E')) =
TO_SINGLE_BYTE(UTL_RAW.CAST_TO_NVARCHAR2('003500390035002F00470050002D000D000A00410054004200580048004E'));
/*return null*/

Answer 1

The two string are not the same; the second one has an extra 000D in the middle:
'3539352F47502D0A41544258484E'
^^^^
'003500390035002F00470050002D000D000A00410054004200580048004E'
^^ ^^ ^^
If they were actually the same you could compare them with implicit conversion (adding the 0D to the first string, but you may prefer to remove it from the second):
SELECT *
FROM DUAL
WHERE UTL_RAW.CAST_TO_VARCHAR2('3539352F47502D0D0A41544258484E') =
UTL_RAW.CAST_TO_NVARCHAR2('003500390035002F00470050002D000D000A00410054004200580048004E');
D
-
X
Or explicitly cat to nvarchar2:
SELECT *
FROM DUAL
WHERE cast(UTL_RAW.CAST_TO_VARCHAR2('3539352F47502D0D0A41544258484E') as nvarchar2(2000)) =
UTL_RAW.CAST_TO_NVARCHAR2('003500390035002F00470050002D000D000A00410054004200580048004E');
D
-
X
or the other way:
SELECT *
FROM DUAL
WHERE UTL_RAW.CAST_TO_VARCHAR2('3539352F47502D0D0A41544258484E') =
cast(UTL_RAW.CAST_TO_NVARCHAR2('003500390035002F00470050002D000D000A00410054004200580048004E') as varchar2(4000));
D
-
X
It's possible you could see some oddities from the casting I suppose, depending on the raw data and the two DB character sets.
From Oracle 12c you could potentially use a UCA linguistic collation that ignores the difference between LF and CRLF, e.g.:
alter session set nls_sort = 'UCA0700_ORADUCET_S1';
alter session set nls_comp = 'LINGUISTIC';
SELECT *
FROM DUAL
WHERE UTL_RAW.CAST_TO_VARCHAR2('3539352F47502D0A41544258484E') =
UTL_RAW.CAST_TO_NVARCHAR2('003500390035002F00470050002D000D000A00410054004200580048004E');
D
-
X
You'd need to see what effect it had on performance though, and whether the other ignorable characters would cause false matches. If you only want to ignore discrepancies between LF/CRLF then you're probably stuck with sanitising the data to be consistent.