How do I pass a value to a stored procedure? I am creating a stored procedure which appends a value to a datatype. The procedure has 2 in parameter (one and one ). The procedure is compiled but I'm having problem executing it. Below is a simplified version of the procedure and the error given when the procedure is executed. SQL> CREATE OR REPLACE PROCEDURE prUpdateContent ( 2 p_contentId IN NUMBER, 3 p_body IN ) 4 IS 5 v_id NUMBER; 6 v_orig ; 7 v_add ; 8 9 BEGIN 10 v_id := p_contentId; 11 v_add := p_body; 12 13 SELECT body INTO v_orig FROM test WHERE id=v_id FOR UPDATE; 14 15 DBMS_LOB.APPEND(v_orig, v_add); 16 commit; 17 END; 18 / Procedure created. SQL> exec prUpdateContent (1, 'testing'); BEGIN prUpdateContent (1, 'testing'); END; * ERROR at line 1: ORA-06550: line 1, column 7: PLS-00306: wrong number or types of arguments in call to 'PRUPDATECONTENT' ORA-06550: line 1, column 7: PL/SQL: Statement ignored
and we said...
Looks like you are attempting to append a TEXT (varchar2) to an
existing -- NOT appending a to a .
It'll just be:
ops$tkyte@dev8i> create table demo ( x int, y );
Table created.
ops$tkyte@dev8i>
ops$tkyte@dev8i> insert into demo values ( 1, 'Hello' );
1 row created.
ops$tkyte@dev8i>
ops$tkyte@dev8i> create or replace procedure lob_append( p_id in
number, p_text in varchar2 )
2 as
3 l_ ;
4 begin
5 select y into l_ from demo where x = p_id for
update;
6
7 dbms_lob.writeappend( l_, length(p_text),
p_text );
8 end;
9 /
Procedure created.
ops$tkyte@dev8i>
ops$tkyte@dev8i> select * from demo
2 /
X Y
----------
-----------------------------------------------------------------
---------------
1 Hello
ops$tkyte@dev8i> exec lob_append( 1, ' World' );
PL/SQL procedure successfully completed.
ops$tkyte@dev8i> select * from demo
2 /
X Y
----------
-----------------------------------------------------------------
---------------
1 Hello World
to do that.
Review & Followup
Rating: 5
Succinct as usual, Mr. Kyte October 25, 2001
Reviewer: Harrison from Haymarket ("Venice without the water") VA, USA
These things always look a lot harder in my mind than on your terminal. As Winston quoted another old man, "I've seen a lot of troubles in my life and most of them never happened" (except the recent rash of people writing to the data dictionary, but he never saw that. Obviously some people are not aware that every ringing phone is not announcing a raise).
Rating: 3
April 16, 2002
Reviewer: Joshua Perry from Las Vegas
When I do something similar in PLSQL (9i), I get an error!
Example:
BEGIN
DECLARE l_ ;
vTEXT_PART varchar2(20) := 'This is a test';
vTEXT_ID int := 0;
BEGIN
IF( vTEXT_ID = 0 )
THEN
INSERT INTO TEXT_TMP VALUES
( SEQ_TEXT_ID.NextVal, vTEXT_PART )
RETURNING TEXT_ID INTO vTEXT_ID;
ELSE
SELECT TEXT_DATA INTO l_
FROM TEXT_TMP WHERE XTEXT_ID = vTEXT_ID
FOR UPDATE ;
dbms_lob.writeappend( l_,
length(vTEXT_TEXT_PART), vTEXT_PART );
END IF;
END;
END;
First, I run with vTEXT_ID = 0, then after a row is inserted I take the returned
sequence and run again to try to append to that row and I get the error:
"The following error has occurred:
ORA-00600: internal error code, arguments: [ktsircinfo_num1], [2147483647], [0],
[0], [], [], [], []
ORA-06512: at "SYS.DBMS_LOB", line 722
ORA-06512: at line 13"
What am I doing wrong here?
Thanks
| Followup: |
After changing your code significantly to make it even compile:
ops$tkyte@ORA9I.WORLD> create table text_tmp( text_id int, text_data );
Table created.
ops$tkyte@ORA9I.WORLD> create sequence seq_text_id;
Sequence created.
ops$tkyte@ORA9I.WORLD>
ops$tkyte@ORA9I.WORLD>
ops$tkyte@ORA9I.WORLD> begin
2 DECLARE l_ ;
3 vTEXT_PART varchar2(20) := 'This is a test';
4 vTEXT_ID int := 0;
5 BEGIN
6 IF( vTEXT_ID = 0 )
7 THEN
8 INSERT INTO TEXT_TMP VALUES
9 ( SEQ_TEXT_ID.NextVal, vTEXT_PART )
10 RETURNING TEXT_ID INTO vTEXT_ID;
11 ELSE
12 SELECT TEXT_DATA INTO l_
13 FROM TEXT_TMP WHERE TEXT_ID = vTEXT_ID
14 FOR UPDATE ;
15
16 dbms_lob.writeappend( l_,
17 length(vTEXT_PART), vTEXT_PART );
18 END IF;
19 END;
20 END;
21 /
PL/SQL procedure successfully completed.
ops$tkyte@ORA9I.WORLD>
ops$tkyte@ORA9I.WORLD>
ops$tkyte@ORA9I.WORLD> BEGIN
2 DECLARE l_ ;
3 vTEXT_PART varchar2(20) := 'This is a test';
4 vTEXT_ID int := 1;
5 BEGIN
6 IF( vTEXT_ID = 0 )
7 THEN
8 INSERT INTO TEXT_TMP VALUES
9 ( SEQ_TEXT_ID.NextVal, vTEXT_PART )
10 RETURNING TEXT_ID INTO vTEXT_ID;
11 ELSE
12 SELECT TEXT_DATA INTO l_
13 FROM TEXT_TMP WHERE TEXT_ID = vTEXT_ID
14 FOR UPDATE ;
15
16 dbms_lob.writeappend( l_,
17 length(vTEXT_PART), vTEXT_PART );
18 END IF;
19 END;
20 END;
21 /
PL/SQL procedure successfully completed.
ops$tkyte@ORA9I.WORLD>
ops$tkyte@ORA9I.WORLD> select * from text_tmp;
TEXT_ID TEXT_DATA
----------
--------------------------------------------------------------------------------
1 This is a testThis is a test
I cannot reproduce.
You need to show an ENTIRE case, everything from start to finish. don't make me
"guess" |
Rating: 3
April 17, 2002
Reviewer: Sagi from India
I have a table LOGINFO. The Structure is:
LOGID VARCHAR2(8)
CALLDESC
CLOSEDESC
When I tried the following:
declare
v_calldesc ;
v_closedesc ;
BEGIN
SELECT calldesc, closedesc INTO v_calldesc, v_closedesc
from calllog where callid='00078349' for update ;
dbms_lob.writeappend(v_calldesc,LENGTH('APPENDING TO CALLDESC'), 'APPENDING
TO CALLDESC' ) ;
DBMS_LOB.writeappend(v_closedesc,LENGTH('APPENDING TO CLOSEDESC'), 'APPENDING
TO CLOSEDES') ;
DBMS_OUTPUT.PUT_LINE('Before commit') ;
COMMIT ;
DBMS_OUTPUT.PUT_LINE('After commit') ;
END ;
Encounted the below Error:
==========================
declare
*
ERROR at line 1:
ORA-21560: argument 2 is null, invalid, or out of range
ORA-06512: at "SYS.DBMS_LOB", line 722
ORA-06512: at line 9
It is working if I use one column but gives error if using two Columns. Why
is this?
Thanx in advance.
Regards,
| Followup: |
because argument 2 is out of range. You have DIFFERENT strings passed to length then to dbms_lob. .. TO CALLDESC ... .. TO CALLDESC ... bug in your code. |
Rating: 3
Better example and understanding April 17, 2002
Reviewer: Joshua Perry from Las Vegas
Sorry about the other example...
I have discovered that the problem lies in the temp table not the SQL. What I
failed to mention in the example is that I am using a Global Temporary table
built off an Object Type,
Example:
CREATE TYPE "USERNAME".TEXT_ROW_TYPE as object
("TEXT_ID" int, "TEXT_DATA" );
CREATE Global Temporary TABLE TEXT_TMP
of "USERNAME"."TEXT_ROW_TYPE"
ON COMMIT preserve ROWS;
The code that you provided works for the example:
CREATE Global Temporary TABLE TEXT_TMP
("TEXT_ID" int, "TEXT_DATA" )
ON COMMIT preserve ROWS;
but not when I use the Temp. Table derived from the Object "TEXT_ROW_TYPE," why?
Thanks for the help!
| Followup: |
How's about a full example -- from start to finish -- in the fashion I do them. A simple, small, yet 100% complete example we can run to see the behaviour. After a while I lose track of what problem exactly you are trying to solve. |
Rating: 3
Full Example April 17, 2002
Reviewer: Joshua Perry from Las Vegas
I was responding the the earlier example/response on this question, and assumed
you had that information available to to so. Here is an example in full:
This is how you created the temp table (from your earlier response):
create table text_tmp( text_id int, text_data );
This is where the problem lies:
CREATE TYPE "USERNAME".TEXT_ROW_TYPE as object
("TEXT_ID" int, "TEXT_DATA" );
CREATE Global Temporary TABLE TEXT_TMP
of "USERNAME"."TEXT_ROW_TYPE"
ON COMMIT preserve ROWS;
/
create sequence seq_text_id;
/
begin
DECLARE l_ ;
vTEXT_PART varchar2(50) := 'This is a test';
vTEXT_ID int := 0;
BEGIN
INSERT INTO TEXT_TMP
VALUES( SEQ_TEXT_ID.NextVal, vTEXT_PART )
RETURNING TEXT_ID INTO vTEXT_ID;
END;
END;
/
BEGIN
DECLARE l_ ;
vTEXT_PART varchar2(50) := 'This is a ANOTHER test';
vTEXT_ID int := 1;
BEGIN
SELECT TEXT_DATA INTO l_
FROM TEXT_TMP WHERE TEXT_ID = vTEXT_ID
FOR UPDATE ;
dbms_lob.writeappend( l_,
length(vTEXT_PART), vTEXT_PART );
END;
END;
It work fine with the first create statement, but fails on the second PLSQL
block with the Temp table based on the row object.
Thanks
| Followup: |
Reproduced, looks like it might be related to an in progress bug #1310111. My
suggestion is to contact support and provide this 100% reproducible and as small
as possible test case:
ops$tkyte@ORA817DEV.US.ORACLE.COM> CREATE or replace type TEXT_ROW_TYPE as
object("TEXT_DATA" );
2 /
Type created.
ops$tkyte@ORA817DEV.US.ORACLE.COM> CREATE Global Temporary TABLE TEXT_TMP of
"TEXT_ROW_TYPE" ON COMMIT preserve ROWS;
Table created.
ops$tkyte@ORA817DEV.US.ORACLE.COM> DECLARE
2 l_ ;
3 BEGIN
4 INSERT INTO TEXT_TMP VALUES( 'testing' );
5 SELECT TEXT_DATA INTO l_ FROM TEXT_TMP FOR UPDATE ;
6 dbms_lob.writeappend( l_, length('x'), 'x' );
7 END;
8 /
DECLARE
*
ERROR at line 1:
ORA-00600: internal error code, arguments: [ktsircinfo_num1], [2147483647], [0],
[0], [], [], [], []
ORA-06512: at "SYS.DBMS_LOB", line 722
ORA-06512: at line 6
I could file it ( i believe it to be a dup of the above bug# ) but it'll be
better if you do it. |
Rating: 5
Nice one April 18, 2002
Reviewer: Nsabasi Etteh from England
I am new to this and i never knew that you could pass XML through a stored procedure. I did it and it worked. Keep it up guys
Rating: 5
June 12, 2002
Reviewer: A reader
Hi Tom, I am passing 1-LOB from one procedure to another. Second procedure already has 2-. I want to append 1-LOB passed from procdedure to 2-LOB eg. 2-LOB := 2-LOB ||1-LOB; How can i do that. Thanks
| Followup: |
read about the dbms_lob package. PROCEDURE APPEND Argument Name Type In/Out Default? ------------------------------ ----------------------- ------ -------- DEST_LOB BLOB IN/OUT SRC_LOB BLOB IN PROCEDURE APPEND Argument Name Type In/Out Default? ------------------------------ ----------------------- ------ -------- DEST_LOB IN/OUT SRC_LOB IN will do that. |
Rating: 5
March 31, 2003
Reviewer: SY from AVO
hi,Tom, I face the same question.
1. I can append text to now, or use "writeAppend" raw data to BLOB.
or use " loadfromfile " to insert external text file into BLOB,
but it seems it can't write/update/append text string
into BLOB like the following :
procedure SP_update_textBLOB( p_id in number, p_text
in varchar2 )
as
l_blob blob;
begin
select blob into l_blob from report where id = p_id for update;
dbms_lob.writeappend( l_blob, length(p_text), p_text );
commit;
end;
Something wrong with my code (run in oracle 8.1.7)?
2. in fact, I want to convert a select statement result
into csv format, and insert this to a BLOB or column.
If I use dump_csv SP to output to OS, and load a real file from OS,
I can do it great. But if I direct write from select result
to BLOB, it seems I only can append one line once like
this:
for i in 1 .. 255 loop
begin
dbms_sql.define_column( l_theCursor, i, l_columnValue, 2000 );
l_colCnt := i;
l_v2:=l_separator || l_descTbl(i).col_name;
SP_updateBLOB(2, utl_raw.cast_to_raw(l_v2) );
l_separator := p_separator;
end;
end loop;
Can I write it all once ? instead write line one by one?
3. How could write a line end charater to BLOB/?
Thank you very much!:)
| Followup: |
1) blobs take raw. varchar2 is not raw. a varchar2 passed to a routine that
expects a raw expects that the varchar2 is in fact HEX codes that represent raw
data.
use utl_raw.cast_to_raw(p_text) instead of just p_text.
2)) well, just replace the || with dbms_lob.writeAppend
this routine would have the blob, you could:
loop
exit when ( dbms_sql.fetch_rows(l_theCursor) <= 0 );
l_separator := '';
for i in 1 .. l_colCnt loop
dbms_sql.column_value( l_theCursor, i,
l_columnValue );
dbms_lob.writeAppend( l_blob, length(l_separator||l_columnValue),
utl_raw.cast_to_raw(l_separator||l_columnValue) );
l_separator := p_separator;
end loop;
dbms_lob.writeAppend( l_blob, 1, '0A' ); /* newline */
end loop;
dbms_sql.close_cursor(l_theCursor);
but -- i have a feeling you would find that building a line and then appending
the line would be more efficient -- but DON'T select the blob out each time --
just do that once at the top of the loop.
3) see above, look for /* newline */ |
Rating: 5
March 31, 2003
Reviewer: SY from AVO
I checked the question "XML Generation/Manipulation using SQL", and found the
following question is exactly what I want, if change XML to CSV. :)
declare
2 l_ctx dbms_xmlquery.ctxHandle;
3 l_;
4 begin
5 l_ctx := dbms_xmlquery.newContext('select * from scott.emp');
6 insert into xmldocs values ('EMP TABLE', empty_())
7 returning xmldoc into l_;
8 dbms_xmlquery.getXml(l_ctx, l_);
9 end;
10 /
But, it seems there's no dbms_csvquery exist. :(
Rating: 4
April 01, 2003
Reviewer: SY from AVO
when I keep digging asktom.oracle.com & google , I find many treasure.
I can answer my 3 question now, everybody who have better advice don't hesitate
to tell me.
1.As blob stores binary data, to update/append it must through raw , can
update/append with text .
2. insert chr(13)||chr(10) can end the line.
3. I update Tom's dump_csv code to a sp to do as this xml function:
xml :
l_ctx := dbms_xmlquery.newContext('select * from scott.emp');
insert into xmldocs values ('EMP TABLE', empty_())
returning xmldoc into l_;
dbms_xmlquery.getXml(l_ctx, l_);
-------------------------------
csv:
CREATE OR REPLACE procedure dump_csv_
( p_id in number, p_query in varchar2, p_separator in varchar2 default ',' )
is
l_output utl_file.file_type;
l_theCursor integer default dbms_sql.open_cursor;
l_columnValue varchar2(2000);
l_status integer;
l_colCnt number default 0;
l_separator varchar2(10) default '';
l_cnt number default 0;
l_descTbl dbms_sql.desc_tab;
l_v2 varchar2(4000);
begin
--l_output := utl_file.fopen( p_dir, p_filename, 'w', 32000 );
dbms_sql.parse( l_theCursor, p_query,
dbms_sql.native );
dbms_sql.describe_columns( l_theCursor, l_colCnt, l_descTbl );
l_v2:=chr(13)||chr(10);
_update(p_id, l_v2 );
for i in 1 .. 255 loop
begin
dbms_sql.define_column( l_theCursor, i,
l_columnValue, 2000 );
l_colCnt := i;
l_v2:=l_separator || l_descTbl(i).col_name;
_append(p_id, l_v2 );
l_separator := p_separator;
exception
when others then
if ( sqlcode = -1007 ) then exit;
else
raise;
end if;
end;
end loop;
l_v2:=chr(13)||chr(10);
_append(p_id, l_v2 );
dbms_sql.define_column( l_theCursor, 1,
l_columnValue, 2000 );
l_status := dbms_sql.execute(l_theCursor);
loop
exit when ( dbms_sql.fetch_rows(l_theCursor) <= 0 );
l_separator := '';
for i in 1 .. l_colCnt loop
dbms_sql.column_value( l_theCursor, i,
l_columnValue );
l_v2:=l_separator || l_columnValue ;
_append(p_id, l_v2 );
l_separator := p_separator;
end loop;
l_v2:=chr(13)||chr(10);
_append(p_id, l_v2 );
l_cnt := l_cnt+1;
end loop;
dbms_sql.close_cursor(l_theCursor);
commit;
end dump_csv_;
/
procedure _update( p_id in number, p_text
in varchar2 )
as
l_ ;
begin
UPDATE report_doc SET docclob = EMPTY_()
where sid=p_id
RETURNING docclob into l_;
dbms_lob.writeappend( l_, length(p_text), p_text );
end;
procedure _append( p_id in number, p_text
in varchar2 )
as
l_ ;
begin
select docclob into l_ from report_doc where sid = p_id for update;
dbms_lob.writeappend( l_, length(p_text), p_text );
end;
e..... it's still much room to improve....
could anyone throw some light here? :) thanx in adv.
Rating: 5
April 01, 2003
Reviewer: SY
Wow, Tom's response much faster than I expected ! I even don't know you have replied . Thank you, Tom. With your suggestion, my code now look like : ... l_v2:=l_v2 || chr(13)||chr(10); dbms_lob.writeappend( l_, length(l_v2), l_v2 ); ... it's fast ! PS: Will it be faster if I write the lines to a temp in ram and in the end write the temp to the real ?
| Followup: |
no, it would not be (faster)... but hey -- as Im fond of saying -- benchmark it! |
Rating: 5
Thanks ! April 02, 2003
Reviewer: SY
Thank tom, your suggestion is alway the best. :)
I benchmark it and got surprise result, one is ten times faster than the other,
so I think maybe I didn't make the best code, and maybe set temp 'cache'
parameter to 'true' is really helpful.... I paste my code here ... it would be a
bit lengthiness.
I use 'set timing on' in sqlplus to test.
the faster one, use temp , cost 1 second . :
/*--------------------------------------------------*/
CREATE OR REPLACE procedure dump_csv_clob3
( p_id in number, p_query in varchar2 )
is
l_theCursor integer default dbms_sql.open_cursor;
l_columnValue varchar2(2000);
l_status integer;
l_colCnt number default 0;
l_separator varchar2(10) default '';
l_cnt number default 0;
l_descTbl dbms_sql.desc_tab;
p_separator varchar2(10) default ',';
l_v2 varchar2(4000) default '';
l_ ;
l4t_ ;
begin
dbms_sql.parse( l_theCursor, p_query, dbms_sql.native );
dbms_sql.describe_columns( l_theCursor, l_colCnt, l_descTbl );
DBMS_LOB.CREATETEMPORARY(l_, TRUE);
for i in 1 .. 255 loop
begin
dbms_sql.define_column( l_theCursor, i, l_columnValue, 2000 );
l_colCnt := i;
l_v2:=l_v2 || l_separator || l_descTbl(i).col_name;
l_separator := p_separator;
exception
when others then
if ( sqlcode = -1007 ) then exit;
else
raise;
end if;
end;
end loop;
l_v2:=l_v2 || chr(13)||chr(10);
dbms_lob.writeappend (l_, length(l_v2), l_v2);
l_v2:='';
dbms_sql.define_column( l_theCursor, 1, l_columnValue, 2000 );
l_status := dbms_sql.execute(l_theCursor);
loop
exit when ( dbms_sql.fetch_rows(l_theCursor) <= 0 );
l_separator := '';
for i in 1 .. l_colCnt loop
dbms_sql.column_value( l_theCursor, i, l_columnValue );
l_v2:=l_v2 || l_separator || l_columnValue ;
l_separator := p_separator;
end loop;
l_v2:=l_v2 || chr(13)||chr(10);
dbms_lob.writeappend (l_, length(l_v2), l_v2);
l_v2:='';
l_cnt := l_cnt+1;
end loop;
dbms_sql.close_cursor(l_theCursor);
UPDATE table1 SET _colume = EMPTY_()
where id=p_id
RETURNING _colume into l4t_;
DBMS_LOB.APPEND(l4t_, l_);
commit;
end dump_csv_clob3;
/
2nd: direct write, slower
/*--------------------------------------------------*/
CREATE OR REPLACE procedure dump_csv_clob2
( p_id in number, p_query in varchar2 )
is
l_theCursor integer default dbms_sql.open_cursor;
l_columnValue varchar2(2000);
l_status integer;
l_colCnt number default 0;
l_separator varchar2(10) default '';
l_cnt number default 0;
l_descTbl dbms_sql.desc_tab;
p_separator varchar2(10) default ',';
l_v2 varchar2(4000) default '';
l_ ;
begin
dbms_sql.parse( l_theCursor, p_query, dbms_sql.native );
dbms_sql.describe_columns( l_theCursor, l_colCnt, l_descTbl );
UPDATE table1 SET _column = EMPTY_()
where id=p_id
RETURNING _column into l_;
for i in 1 .. 255 loop
begin
dbms_sql.define_column( l_theCursor, i, l_columnValue, 2000 );
l_colCnt := i;
l_v2:=l_v2 || l_separator || l_descTbl(i).col_name;
l_separator := p_separator;
exception
when others then
if ( sqlcode = -1007 ) then exit;
else
raise;
end if;
end;
end loop;
l_v2:=l_v2 || chr(13)||chr(10);
dbms_lob.writeappend (l_, length(l_v2), l_v2);
l_v2:='';
dbms_sql.define_column( l_theCursor, 1, l_columnValue, 2000 );
l_status := dbms_sql.execute(l_theCursor);
loop
exit when ( dbms_sql.fetch_rows(l_theCursor) <= 0 );
l_separator := '';
for i in 1 .. l_colCnt loop
dbms_sql.column_value( l_theCursor, i, l_columnValue );
l_v2:=l_v2 || l_separator || l_columnValue ;
l_separator := p_separator;
end loop;
l_v2:=l_v2 || chr(13)||chr(10);
dbms_lob.writeappend (l_, length(l_v2), l_v2);
l_v2:='';
l_cnt := l_cnt+1;
end loop;
dbms_sql.close_cursor(l_theCursor);
commit;
end dump_csv_clob2;
/
Rating: 3
SY, I got different result April 08, 2003
Reviewer: A reader
In my case, writing a real is slightly faster than using temp , Oracle9iR2. Maybe you should check the relative settings.
Rating: 4
April 20, 2003
Reviewer: A reader
hi ,all , how to ' create a temp in ram ' ? It seems that temp is not stored in ram .....
| Followup: |
PROCEDURE CREATETEMPORARY Argument Name Type In/Out Default? ------------------------------ ----------------------- ------ -------- LOB_LOC IN/OUT CACHE BOOLEAN IN DUR BINARY_INTEGER IN DEFAULT it'll cache what it can/should. |
Rating: 4
Copying Varchar2 to CLOB February 12, 2004
Reviewer: Yogesh from Pune, India
I have a table with a VARCHAR2 column. I want to change its datatype to , But as this is not possible I added one new column as . How can I copy data from varchar column to ?
| Followup: |
update set _column = varchar2_column; ops$tkyte@ORA9IR2> alter table emp add ename_; Table altered. ops$tkyte@ORA9IR2> update emp set ename_ = ename; 14 rows updated. ops$tkyte@ORA9IR2> alter table emp drop column ename; Table altered. ops$tkyte@ORA9IR2> alter table emp rename column ename_ to ename; Table altered. ops$tkyte@ORA9IR2> set linesize 60 ops$tkyte@ORA9IR2> desc emp Name Null? Type ----------------------------- -------- -------------------- EMPNO NOT NULL NUMBER(4) JOB VARCHAR2(9) MGR NUMBER(4) HIREDATE DATE SAL NUMBER(7,2) COMM NUMBER(7,2) DEPTNO NUMBER(2) ENAME |
Rating: 4
ORA-03120 February 13, 2004
Reviewer: Yogesh from Pune, India
Thanks Tom, I did the same thing. But when I use to select rows from that table I was getting SQL> select * from test; ERROR: ORA-03120: two-task conversion routine: integer overflow My Server is sun Solaris and I was trying this from my XP client. When I saw the explanation, it was saying something about heterogeneous system. So I logged in to my Solaris server using telnet and did the same thing. It worked. I could not understand, if something is working from server, why it should fail from client. Your comments Please.
| Followup: |
what version of plus do you have on windoze and which version of the database do you have on sun? |
Rating: 3
ORA-03120 February 16, 2004
Reviewer: Yogesh from Pune, India
SQL*Plus: Release 8.1.7.0.0 Oracle8i Enterprise Edition Release 8.1.7.0.0 - 64bit Production
| Followup: |
are you utf8 in the server. |
Rating: 4
UTF8 February 17, 2004
Reviewer: Yogesh from Pune, India
Yes you are right, NLS_CHARACTERSET is set to UTF8
Rating: 4
Clob retrival and keyword search February 24, 2004
Reviewer: djecal from Richmond,VA
Hi Tom, is it a way to run a keyword search in file data type? After
writing a procedure to load a file can we load more than one file? I failed
to retrieve file using dbms_lob.read (). These are my statements:
select dbms_lob.read (state_code,seq_num,resp) from narrative_tb;
ERROR at line 1:
ORA-00904: invalid column name
SQL> select dbms_lob.read(resp,4000,1) from narrative_tb;
select dbms_lob.read(resp,4000,1) from narrative_tb;
*
ERROR at line 1:
ORA-00904: invalid column name
SQL> select dbms_lob.substr (resp,4000,1) from narrative_tb;
The last query was successful. Can you help me with query in case the file is
over 4000? Also, how would search for keyword?
Thank you
| Followup: |
No idea what you are trying to do with read there.. select _column from table; will work or select dbms_lob.substr( col_column, for_char, from_char ) from t; will get a substring of it. if the file is over 4000, just select _column from t; in sqlplus use "set long 500000" to set the max that plus will display (only applies to sqlplus) |
Rating: 5
keyword search on multiple flat files in oracle February 24, 2004
Reviewer: Dams from Orlando,FL
Hello, we have 1600 files stored on a directory of our server. What will be the easiest way to do a keyword search through all thes files? Thanks
| Followup: |
Oracle Text, just index them. Then you can use SQL.http://otn.oracle.com/pls/db92/db92.docindex?remark=homepage#index-TEX |
Rating: 5
keyword search February 25, 2004
Reviewer: Dams from Orlando, FL
Hi Tom,
can you help me out with this please?
I wanted to create this index in Oracle InterMedia and got this error message.
I can I solve it?
CREATE INDEX MASTERDB.MASTERDB_IDX
ON MASTERDB.NARRATIVE_TB(RESP) INDEXTYPE IS CTXSYS.CONTEXT
PARAMETERS (' DATASTORE CTXSYS.FILE_DATASTORE')
/
*
ERROR at line 2:
ORA-04045: errors during recompilation/revalidation of CTXSYS.CONTEXT
ORA-29835: ODCIGETINTERFACES routine does not return required interface(s)
| Followup: |
looks like an upgrade or install issue -- please contact support for assistance on that one. |
Rating: 3
combining file_datastore with user_datastore March 26, 2004
Reviewer: Parag from Munich, Germany
Hi Tom,
We are running full text search based on a file_datastore for the files residing
on the filesystem. Moreover, we have some simple OR conditions with main file
content clause,for search of other attributes like:
select .. from text_files
where contains(file, 'plan') > 0
or filename like 'plan'
or extension = 'plan'
As now the search requirenments grows, we have to add lot of other attributes to
the search criteria(owner, description, categories, keywords ...)
So on a trial basis, I have created a User_datastore for all the other
attributes. New query would look like:
select ..
from text_files, file_attributes
where text_files.objectid = file_attributes.objectid
and (contains(file, 'plan') > 0
or
contains(fileAttr, 'plan') > 0
)
Now the question is, is it possible to combine these two datastores together so
that I can issue only one contains clause?
(I tried appending the File_datastores's 'file' column(on which the index is
created) in the procedure created for User_datastore. Procedure creation went
fine but select statement did not returned any result.
I can provide my test case as well if you want to see, but before that, I just
want to know wheather such thing is possible at all or not?)
)
thanks a lot!
| Followup: |
| http://asktom.oracle.com/pls/ask/f?p=4950:8:::::F4950_P8_DISPLAYID:440419921146 |
Rating: 2
No results by combining two datastores! March 30, 2004
Reviewer: parag from Germany
Hi Tom,
Here is my test code for creating a File_datastore and User_datastore.
conn test/test
drop table file_content;
create table file_content(id int primary key, filePath varchar2(255));
insert into file_content values(1, 'C:\data\abc.html');
-- abc.html is result of conversion from
-- abc.doc to abc.html by ctxhx
-- stored in C:\data directory
begin
ctx_ddl.drop_preference('FILE_PREFERENCE');
end;
/
begin
ctx_ddl.create_preference('FILE_PREFERENCE','FILE_DATASTORE');
end;
/
CREATE INDEX IDX_FILE_CONTENT ON file_content(filePath)
INDEXTYPE IS CTXSYS.CONTEXT
parameters ('datastore FILE_PREFERENCE filter ctxsys.null_filter');
drop table tt;
create table tt (id number(10), varchar2_col varchar2(50), dummy_col char(1) );
insert into tt values(1,'jungle', '');
insert into tt values(2, 'forest', '');
commit;
grant select any table to ctxsys;
conn ctxsys/ctxsys
create or replace procedure pr_one(rid in rowid, tlob in out ) is
begin
for c1 in ( select varchar2_col, nvl(dummy_col,' ') dummy_col, filepath
from test.tt, test.file_content
where test.tt.id = test.file_content.id
and test.tt.rowid = rid)
loop
dbms_lob.writeappend(tlob, length(c1.varchar2_col)+1,
c1.varchar2_col||' ');
dbms_lob.writeappend(tlob, length(c1.filepath)+1, c1.filepath||' ');
dbms_lob.writeappend(tlob, length(c1.dummy_col), c1.dummy_col);
end loop;
end;
/
grant execute on pr_one to test;
conn test/test@parag2
exec ctx_ddl.drop_preference('one_user_datastore');
exec ctx_ddl.drop_preference('one_lexer');
begin
ctx_ddl.create_preference( 'one_user_datastore', 'user_datastore' );
ctx_ddl.set_attribute( 'one_user_datastore', 'procedure','pr_one' );
end;
/
begin
ctx_ddl.create_preference( 'one_lexer', 'BASIC_LEXER' );
ctx_ddl.set_attribute( 'one_lexer', 'base_letter', 'YES');
ctx_ddl.set_attribute( 'one_lexer', 'mixed_case', 'NO' );
end;
/
create index idx_tt on tt(dummy_col)
indextype is ctxsys.context
parameters( 'datastore one_user_datastore lexer one_lexer');
-- test File_datastore index individually
-- 'training' is a word appearing in abc.html file
Prompt 'Testing FILE DATASTORE with contains(filepath,'training')'
Select * from file_content
where contains(filepath,'training') > 0;
/*
Result I get
ID FILEPATH
-- -----------------
1 C:\data\abc.html
*/
prompt 'Testing USER_DATASTORE'
Select * from tt
where contains(dummy_col,'jungle') > 0;
/*
ID VARCHAR2_COL D
---- ------------- -
1 jungle
*/
prompt 'Testing Combined FILE_DATASTORE AND USER_DATASTORE'
Select * from tt
where contains(dummy_col,'training') > 0;
/*
no rows selected
*/
So if I try to search a by putting the real path which is in filepath column
then I get the result:
Select * from tt
where contains(dummy_col,'abc') > 0;
ID VARCHAR2_COL D
----- -------------------------------------------------- -
1 jungle
Index on File_datastore is not getting used at all!
How can solve this problem?
| Followup: |
of course not? why would it?
You indexed the return value of a function. that function indexed:
dbms_lob.writeappend(tlob, length(c1.varchar2_col)+1,
c1.varchar2_col||' ');
dbms_lob.writeappend(tlob, length(c1.filepath)+1, c1.filepath||' ');
dbms_lob.writeappend(tlob, length(c1.dummy_col), c1.dummy_col);
so, when did the word 'training' ever get returned by that function? you
indexed the NAME of a file. You did not index the CONTENTS of the file.
Perhaps, you meant to load the file into the blob for indexing? like a
dbms_lob.loadfromfile call would do?? |
Rating: 3
Combined datastore April 02, 2004
Reviewer: Parag from Germany
well, as I see from dbms_lob.loadfromfile which copy the content of the file into our temp . And the content will be returned by the function to check against contains clasue. This will surely give my results. but how is the index working then? If I have million number of files then function will load all the files, concat the content to other attributes and return the resulting string to check whether it contains the specified term. This is too expensive, and I guess that's why FILE_DATASTORE is presented so as to keep the file content-data as indexed one.Any query against this index will not check the whole content of file but just a indexed data. What I need is to join the indexed data from file contents to the other attributes!!
| Followup: |
it'll be the same the document is in the file system -- the contents of the document in the index. whether you do it yourself in the function OR you let text read the document -- the same stuff ends up in the index!! the document -- it is always in the file system, the index -- that is always in the database. |
Rating: 4
combined datastore April 02, 2004
Reviewer: Parag from Germany
ok, just forget the previous comment. I got it what you mean by load the file before indexing.(function based index..) But I see one problem here. Normally file is a static data on our platform, that means once it is uploaded it is used by all the users on the platform. It is not changed. BUT, metadata of a file(which is also part of index), like categories, keywords, permissions they change all the time. This means, this function based index will always be synchronized whenever there is a change in metadata, right? Do you see any other way of implementing the same?
| Followup: |
correct -- the index would be syncronized. and the problem with that would be what exactly? (other than you get the right answer..... when you query....) |
Rating: 3
Datastore April 05, 2004
Reviewer: Parag from Germany
(Wish you a nice holydays!!) In case you see this,what do you mean by: (other than you get the right answer..... when you query....) ?? If the file is too big, synchronization of index could be expensive in that case. Moreover as I said, we do not expect the file to be changed, but the metadata. So my question was that is there a way to index the file content only once and use it with other metadata in a function?
| Followup: |
what i mean is - if you update the data - and we DID NOT sync the index - then you get the wrong answer (as text would not find documents it should and it would find documents it shouldnt) You are indexing the "row", either the entire row or nothing, in your case. |
Rating: 4
Datastore April 05, 2004
Reviewer: Parag from Germany
Thanks a lot Tom. I see now there is no way I can avoid synchronizing of index every time the metadata changed. But as I already have documents parsed (to convert them to htmls and that was also a expenssive task) on my hard disk, hopefully, synchronizing of index should be fast enough. (for a really big file, god help my users!!) thanks again!
April 05, 2004 -- Thanks for the question regarding "Performance poor after analyzeing table", version 8.1.7
736
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
