Using convert function on nvarchar2

Question

Hi experts,
I am having a bit of a problem with the convert function. We use convert to compare street- and citynames and ignore any special characters, such as é ç etc.
Something like:
select ...
from   ...
where  convert(new_street_name, 'US7ASCII') = convert(existing_street_name, 'US7ASCII')
This works fine if the datatype is varchar2, for instance:
SQL> select convert('äàáâçëèéêïìíîöòóôüùúûÿ','US7ASCII') text from dual;
TEXT
aaaaceeeeiiiioooouuuuy
If the datatype if nvarchar2 however, the result is not as expected:
SQL> select convert(cast('äàáâçëèéêïìíîöòóôüùúûÿ'as nvarchar2(64)),'US7ASCII') text from dual;
TEXT
慡慡捥敥敩楩楯潯潵畵
The NLS character settings on our database (10.2.0.4) are:
NLS_CHARACTERSET       AL32UTF8  Character set
NLS_NCHAR_CONV_EXCP    FALSE     NLS conversion exception
NLS_NCHAR_CHARACTERSET AL16UTF16 NCHAR Character set
I have tried several combinations... but no luck so far. Is it possible to use convert on an nvarchar2 to go from é to e?
Maybe it is better just to use the translate function and define each conversion explicitly. Convert seemed a nice option because it works without any additional parameters... on a varchar2 at least
Thanks!

Answer 1

The usage of convert is not encouraged by the docs and in my opinion it's rather by accident that this works in your specific case than something other.
What's going on?
Convert returns the char-datatype of the input.
We can use simple to_char to use the convert funtion in the way you intend.
(You shoud take care when handling n_char data in sql statements, especially in 10 g enviromnments. You have to set the environment parameter ORA_NCHAR_LITERAL_REPLACE=TRUE and use the n-prefix. Take a look at the globalization guide for more details.)
CREATE TABLE  "TESTNCHAR"
   ( "ID" NUMBER,
"STR" NVARCHAR2(30),
"STR2" VARCHAR2(300)
insert into testnchar values (1, n'ßäàáâçëèéêïìíîöòóôüùúûÿ','ßäàáâçëèéêïìíîöòóôüùúûÿ')
select
id
,str,str2
,dump(str,1010) dmp
,dump(str2,1010) dmp2
,dump(convert(str,'US7ASCII')) dc
,dump(convert(str2,'US7ASCII')) dc2
,convert(to_char(str),'US7ASCII') c
,convert(str2,'US7ASCII') c2
from testnchar
ID
STR
STR2
DMP
DMP2
DC
DC2
C
C2
1
ßäàáâçëèéêïìíîöòóôüùúûÿ
ßäàáâçëèéêïìíîöòóôüùúûÿ
Typ=1 Len=46 CharacterSet=AL16UTF16: 0,223,0,228,0,224,0,225,0,226,0,231,0,235,0,232,0,233,0,234,0,239,0,236,0,237,0,238,0,246,0,242,0,243,0,244,0,252,0,249,0,250,0,251,0,255
Typ=1 Len=46 CharacterSet=AL32UTF8: 195,159,195,164,195,160,195,161,195,162,195,167,195,171,195,168,195,169,195,170,195,175,195,172,195,173,195,174,195,182,195,178,195,179,195,180,195,188,195,185,195,186,195,187,195,191
Typ=1 Len=23: 63,97,97,97,97,99,101,101,101,101,105,105,105,105,111,111,111,111,117,117,117,117,121
Typ=1 Len=23: 63,97,97,97,97,99,101,101,101,101,105,105,105,105,111,111,111,111,117,117,117,117,121
?aaaaceeeeiiiioooouuuuy
?aaaaceeeeiiiioooouuuuy
We can see that this already fails for ß.
To give you an idea of alternative approaches:
create table test_comp (nid number, str1 varchar2(300), str2 varchar2(300))
insert into test_comp values (1, 'ßäàáâçëèéêïìíîöòóôüùúûÿ','ssäàáâçëèéêïìíîöòóôüùúûÿ')
insert into test_comp values (2, 'ßäàáâçëèéêïìíîöòóôüùúûÿ','säàáâçëèéêïìíîöòóôüùúûÿ')
select
from test_comp
where
str2 like str1
no data found
select
from test_comp
where
upper(str2) like NLS_UPPER(str1, 'NLS_SORT = XGERMAN')
NID
STR1
STR2
1
ßäàáâçëèéêïìíîöòóôüùúûÿ
ssäàáâçëèéêïìíîöòóôüùúûÿ