Does the . character a reserved character?

Question

Hi,
I would like to search the following information in a full text index: *10.1139/f91-051* I am then submit the following query:
select * from mytable where contains(my_column,'"10.1139/f91-051" WITHIN doi') > 0
but it does not return any data. However, when I replace the *.* character by a _*, I can retrieve my document:
select * from mytable where contains(my_column,'"10_1139/f91-051" WITHIN doi') > 0
Does the *.* character a reserved character? But when I try to escape it, it does not work neither:
select * from mytable where contains(my_column,'"10\.1139/f91-051" WITHIN doi') > 0
What did I do wrong?
In the same way, what query should I submit to find all the I.B.M. words from my index for exemple?
Kind regards,
Fred
Edited by: user503159 on 31 juil. 2009 02:29

Answer 1

I'm unable to reproduce this exactly as you've stated.
A few things to note:
- The double quotes in your query have no effect. There is no "phrase" operator in Oracle Text, tokens that follow each other are a phrase by default
- The "-" character is a special operator, and needs to be escaped to be treated as a literal (you can precede it with a backslash or put braces (curly brackets) round the whole search term)
- "\" and "/" are break characters by default.
- "." is a join character ONLY within numbers. It's a break character within normal alphabetic text.
Run the folliowing:
-- drop table t;
create table t (x varchar2(2000));
insert into t values ('<doi>10.1139/f91-051</doi>');
create index ti on t(x) indextype is ctxsys.context parameters ('section group ctxsys.auto_section_group');
select token_text, decode(token_type, 0, 'Normal', 2, 'Zone Section') as "Token Type" from dr$ti$i order by token_type;The last query shows us what tokens are indexed. We should see:
TOKEN_TEXT                                                       Token Type
051                                                              Normal
F91                                                              Normal
10.1139                                                          Normal
DOI                                                              Zone SectionSo we can run a query without the break characters
SQL> select * from t where contains (x, '10.1139 f91 051 within doi') > 0;
X
<doi>10.1139/f91-051</doi>or we can run a query with the break characters in it, but with the "-" character escaped:
SQL> select * from t where contains (x, '10.1139/f91\-051 within doi') > 0;
X
<doi>10.1139/f91-051</doi>The query that you claimed worked does not work for me because the "-" operator is not escaped (I'm on 11g - it's possible the behaviour is different in earlier versions, but unlikely)
SQL> select x from t where contains(x,'"10_1139/f91-051" WITHIN doi') > 0;
no rows selectedAs for your question about searching for "I.B.M.", you can search for "I.B.M." or "I B M" and it will find it. You can also set "." to be a SKIPJOINS character, in which case you can search for "I.B.M." or "IBM" (but not "I B M").
Note there is a gray area for tokens which are partly numeric - "1.23ABC" will probably be indexed as a single token, whereas "ABC1.23" will probably be indexed as two tokens "ABC1" and "23". This is because for performance reasons the lexer does not backtrack - when it hits the "." in the first example it thinks it is handling a pure number and therefore treats the "." as a numeric join character.