I have code that is randomly getting "ORA-22814: attribute or element value is larger than specified in type" with a bulk collect into.
I suspect this is a data issue since it's random but PL/SQL is not my strong suit but I maintain an application that has a great deal of it in processing logic.
This is an Oracle 12c database.
Any help would be appreciated!
This is the block of code:
SELECT mxpv_activityupdate_O (
projectid,
task_id,
task_code,
pvtask_code,
task_name,
CASE
WHEN wbs_t0_id IS NOT NULL AND scgemergent_flag = 'Y'
THEN
wbs_t0_id
ELSE
wbs_pv_id
END,
act_start_date,
CASE
WHEN act_end_date IS NOT NULL AND (invalidactfinish_flag = 'Y')
THEN
NULL
ELSE
act_end_date
END,
NULL,
CASE WHEN invalidactuals_flag = 'Y' THEN NULL ELSE act_work_qty END,
MAX (oid),
scgemergent_flag,
invalidactuals_flag,
invalidactfinish_flag)
BULK COLLECT INTO u_all_tab
FROM (SELECT projectorganization,
projecttype,
projectid,
task_id,
task_code,
pvtask_code,
pvact_start_date,
task_name,
act_start_date,
act_end_date,
CASE
WHEN pvact_start_date IS NULL
AND act_start_date IS NULL
AND NVL (act_work_qty, 0) <> 0
THEN
'Y'
ELSE
'N'
END
invalidactuals_flag,
act_work_qty,
oid,
wbs_t0_id,
wbs_pv_id,
CASE
WHEN pvact_start_date IS NULL
AND (act_start_date IS NOT NULL --or nvl(act_work_qty,0)<>0
)
AND projectorganization = 'SCG'
AND projecttype = 'Daily'
AND pvtarget_start_date > v_t0_start_midnight + 7
THEN
'Y'
ELSE
'N'
END
scgemergent_flag,
CASE
WHEN act_end_date IS NOT NULL
AND NVL (act_start_date, pvact_start_date) IS NOT NULL
AND act_end_date <
NVL (act_start_date, pvact_start_date)
THEN
'Y'
ELSE
'N'
END
invalidactfinish_flag
FROM (SELECT oid,
projecttype,
wbs_t0.wbs_id wbs_t0_id,
--emergentwork,
projectorganization,
projectid,
task_id,
pvwbs_id wbs_pv_id,
activityid task_code,
task_code pvtask_code,
pvact_start_date,
pvtarget_start_date,
CASE
WHEN pvact_end_date IS NULL THEN activityname
ELSE NULL
END
task_name,
CASE
WHEN actualstart IS NOT NULL
AND ( pvact_start_date IS NULL
OR actualfinish IS NOT NULL)
THEN
actualstart
ELSE
NULL
END
act_start_date,
--actualstart act_start_date,
CASE
WHEN pvact_end_date IS NULL
AND ( pvact_start_date IS NOT NULL
OR actualstart IS NOT NULL)
THEN
actualfinish
ELSE
NULL
END
act_end_date,
NULL target_start_date,
CASE WHEN mxactualunits.proj_catg_short_name = 'Y' --and nvl(pvact_work_qty, 0) <> nvl(actuallaborhours,0)
THEN NVL (actuallaborhours, 0) ELSE NULL END
act_work_qty
FROM --1) activityid and proj id matches
(SELECT oid,
pv.task_id,
mx.proj_id,
mx.projectid,
pv.task_code,
pv.task_name,
pv.wbs_id pvwbs_id,
pv.act_start_date pvact_start_date,
pv.act_end_date pvact_end_date,
pv.target_start_date pvtarget_start_date,
mx.activityid,
mx.activityname,
mx.actualstart,
mx.actualfinish,
pv.act_work_qty pvact_work_qty,
mx.actuallaborhours,
mx.projectorganization,
mx.projecttype,
mx.process_start_ts --,
FROM mxpv_activity_tmp mx, mxpv_task_tmp pv
WHERE mx.task_id = pv.task_id
UNION
--3) PM Number/Job Plan number matches
SELECT oid,
jp.task_id,
jp.proj_id,
jp.proj_short_name,
jp.task_code,
jp.task_name,
jp.wbs_id,
jp.act_start_date pvtarget_start_date,
jp.act_end_date pvtarget_end_date,
jp.target_start_date pvtarget_start_date,
jp.activityid,
jp.activityname,
jp.actualstart,
jp.actualfinish,
jp.act_work_qty pvact_work_qty,
actuallaborhours,
projectorganization,
projecttype,
p_process_start_ts --,
FROM --3a) PM Number matches
(SELECT * FROM TABLE (pm_tab)
UNION
--3a) Job Plan Number matches
SELECT * FROM TABLE (jp_tab)) jp)
update_activity,
(SELECT projpcat.PROJ_ID, pcval.PROJ_CATG_SHORT_NAME
FROM privuser_PROJPCAT projpcat,
privuser_PCATTYPE pctype,
privuser_PCATVAL pcval
WHERE projpcat.PROJ_CATG_ID = pcval.PROJ_CATG_ID
AND pcval.PROJ_CATG_TYPE_ID =
pctype.PROJ_CATG_TYPE_ID
AND pctype.PROJ_CATG_TYPE = 'MX Actual Units')
mxactualunits,
(SELECT w.wbs_id, w.proj_id
FROM privuser_projwbs w, privuser_phase ph
WHERE w.phase_id = ph.phase_id
AND ph.phase_name = '0') wbs_t0
WHERE --update_activity.proj_id = actstrtoffset.proj_id (+)
update_activity.proj_id = wbs_t0.proj_id(+)
AND update_activity.proj_id = mxactualunits.proj_id(+)
AND update_activity.process_start_ts =
p_process_start_ts--and delete_ts is null
))
GROUP BY projectid,
task_id,
task_code,
pvtask_code,
task_name,
wbs_t0_id,
wbs_pv_id,
act_start_date,
act_end_date,
act_work_qty,
invalidactuals_flag,
invalidactfinish_flag,
scgemergent_flag,
pvact_start_date;
The type is defined as follows:
TYPE PV_APP_DATA.MXPV_ACTIVITYUPDATE_O AS OBJECT(
proj_short_name VARCHAR2 (100),
task_id NUMBER,
task_code VARCHAR2 (120 CHAR),
pvtask_code VARCHAR2 (120 CHAR),
task_name VARCHAR2 (120 CHAR),
wbs_id NUMBER,
act_start_date DATE,
act_end_date DATE,
target_start_date DATE,
act_work_qty NUMBER,
oid NUMBER,
scgemergent_flag VARCHAR2 (10),
invalidactuals_flag VARCHAR2 (100),
invalidactfinish_flag VARCHAR2 (100)
)
Please post if you need more information I wasn't sure what could be useful.
update 1 7/18/17
I modified the query with substr with no luck in fixing the problem:
select mxpv_activityupdate_O(
substr(projectid,0,99),
task_id,
substr(task_code,0,119),
substr(pvtask_code,0,119),
substr(task_name,0,119),
case when wbs_t0_id is not null and scgemergent_flag='Y' then wbs_t0_id
else wbs_pv_id end,
act_start_date,
case when act_end_date is not null and
(invalidactfinish_flag='Y'
)
then null else act_end_date end,
null,
case when invalidactuals_flag = 'Y' then null else act_work_qty end,
max(oid),
substr(scgemergent_flag,0,9),
substr(invalidactuals_flag,0,99),
substr(invalidactfinish_flag,0,99)
)
bulk collect into u_all_tab
from
(SELECT
projectorganization,
projecttype,
substr(projectid,0,99) as projectid,
task_id,
substr(task_code,0,119) as task_code,
substr(pvtask_code,0,119) as pvtask_code,
pvact_start_date,
substr(task_name,0,119) as task_name,
act_start_date,
act_end_date,
case when pvact_start_date is null and act_start_date is null
and nvl(act_work_qty, 0) <> 0 then 'Y' else 'N' end invalidactuals_flag,
act_work_qty,
oid,
wbs_t0_id,
wbs_pv_id,
case when pvact_start_date is null and
(act_start_date is not null --or nvl(act_work_qty,0)<>0
)
and projectorganization='SCG' and projecttype='Daily'
and pvtarget_start_date>v_t0_start_midnight+7
then 'Y' else 'N' end scgemergent_flag,
case when act_end_date is not null and nvl(act_start_date, pvact_start_date) is not null
and act_end_date<nvl(act_start_date, pvact_start_date) then 'Y' else 'N' end
invalidactfinish_flag
FROM
"if I want to force to correct length what's the best approach?"
You need to compare the length of the Tables' columns with the length of the Type's target attributes, and see which attributes are shorter than the columns you're trying to fit into them. Use the data dictionary.
For the Type Attributes use USER_TYPE_ATTRS (or ALL_TYPE_ATTRS if it's a different schema):
select attribute_name, attr_type_name, length, scale, precision
from user_type_attrs
where type_name = 'MXPV_ACTIVITY_UPDATE_O'
/
For the Table Columns it will be a similar query based on USER_TAB_COLS.
Once you have established the errant columns you will need to alter the type to enlarge the attributes. Alternatively apply SUBSTR() function to the columns in the query projection, to cut them down to size. Which is actually better depends on the circumstances of your application, although usually we would prefer the approach which doesn't lose data.
Overall Task :- I need to retrieve data from 45 fields in system A and dump that data into a temp table which is then picked up by a unix process which produces an xml data file to be imported into system B.
Specific Question : What would be the best way of retrieving the data to be written into the 45 fields. Majority of the data is independent and can't be retrieved using a single statement. The way i currently retrieve this data is as follows (example below)
My temp tables hold the affected properties ID that i need to extract data for. i.e PROP_LIST_TEMP and ASSOC_PROP_TEMP.
SELECT SUBSTR (pro.pro_propref, 1, 25) UPRN,
(SELECT SUBSTR (adr_building, 1, 100)
FROM addresses, address_usages
WHERE aus_adr_refno = adr_refno
AND aus_aut_far_code = 'PHYSICAL'
AND aus_aut_fao_code = 'PRO'
AND (aus_end_date IS NULL OR aus_end_date > SYSDATE)
AND aus_pro_refno = pro.pro_refno)
BUILDING_NAME,
(SELECT CASE
WHEN (adr_street_number like 'BLOC%'
OR adr_street_number like '%-%'
OR adr_street_number like '%/%')
THEN NULL
ELSE regexp_replace (adr_street_number, '[^[:digit:]]+')
END
FROM addresses, address_usages
WHERE aus_adr_refno = adr_refno
AND aus_aut_far_code = 'PHYSICAL'
AND aus_aut_fao_code = 'PRO'
AND (aus_end_date IS NULL OR aus_end_date > SYSDATE)
AND aus_pro_refno = pro.pro_refno)
STREET_NUMBER,
(SELECT CASE
WHEN (adr_street_number like 'BLOC%'
OR adr_street_number like '%-%'
OR adr_street_number like '%/%')
THEN SUBSTR (adr_street_number, 1, 20)
ELSE REGEXP_REPLACE (adr_street_number, '[^[:alpha:]]+', '')
END
FROM addresses, address_usages
WHERE aus_adr_refno = adr_refno
AND aus_aut_far_code = 'PHYSICAL'
AND aus_aut_fao_code = 'PRO'
AND (aus_end_date IS NULL OR aus_end_date > SYSDATE)
AND aus_pro_refno = pro.pro_refno)
STREET_NUMBER_SUFFIX,
(SELECT SUBSTR (ptv_pty_code, 1, 3)
FROM prop_type_values
WHERE ptv_refno = pro.pro_hou_ptv_refno)
HOUSE_TYPE
FROM properties pro
WHERE pro_refno IN (select * from PIMSS_PROP_LIST_TEMP
UNION
select * from PIMSS_ASSOC_PROP_TEMP)
AND pro.pro_hou_hrv_hot_code IN
(SELECT frv_code
FROM first_ref_values
WHERE frv_frd_domain IN ('ASS_OWN_REF')
AND frv_current_ind = 'Y');
Since the where clauses of the subqueries in the select statement are identical, you could simply pull that out into the where clause, like so:
SELECT SUBSTR (pro.pro_propref, 1, 25) UPRN,
SUBSTR (addr.adr_building, 1, 100) BUILDING_NAME,
CASE WHEN (addr.adr_street_number like 'BLOC%'
OR addr.adr_street_number like '%-%'
OR addr.adr_street_number like '%/%')
THEN NULL
ELSE regexp_replace (addr.adr_street_number, '[^[:digit:]]+')
END STREET_NUMBER,
CASE WHEN (addr.adr_street_number like 'BLOC%'
OR addr.adr_street_number like '%-%'
OR addr.adr_street_number like '%/%')
THEN SUBSTR (addr.adr_street_number, 1, 20)
ELSE REGEXP_REPLACE (addr.adr_street_number, '[^[:alpha:]]+', '')
END STREET_NUMBER_SUFFIX,
(SELECT SUBSTR (ptv_pty_code, 1, 3)
FROM prop_type_values
WHERE ptv_refno = pro.pro_hou_ptv_refno) HOUSE_TYPE
FROM properties pro,
(select adr_building,
adr_street_number
FROM addresses, address_usages
WHERE aus_adr_refno = adr_refno
AND aus_aut_far_code = 'PHYSICAL'
AND aus_aut_fao_code = 'PRO'
AND (aus_end_date IS NULL OR aus_end_date > SYSDATE)) addr
WHERE pro.pro_refno = aus_pro_refno
and pro_refno IN (select * from PIMSS_PROP_LIST_TEMP
UNION
select * from PIMSS_ASSOC_PROP_TEMP)
AND pro.pro_hou_hrv_hot_code IN (SELECT frv_code
FROM first_ref_values
WHERE frv_frd_domain IN ('ASS_OWN_REF')
AND frv_current_ind = 'Y');
You might possibly need an outer join if there's a chance that no rows could be returned from the addr subquery.
I am creating a view and the same view converted into materialized view in the same system. But doing the same thing in another system I got error ORA-01722: invalid number when creating the materialized view. Why?
create materialized view MV_EMP_VALI
refresh complete with rowid start with SYSDATE+1/24 AS
(select * from V_CHA1);
View:-
CREATE OR REPLACE VIEW V_CHA1 AS(SELECT EMPNO,
MONTHYEAR,
to_number(SUM(CPFEMO)) AS EMOLUMENTS,
to_number(SUM(CPEPF)) AS EMPPFSTATUARY,
to_number(SUM(AEMO)) AS AEMO,
to_number(SUM(APEPF)) AS APEPF,
MAX(recsts) AS recsts
FROM ((SELECT RECDATE,
(CASE WHEN (REPFEMOFLAG='N') THEN
round(NVL(trim(EMO), 0))
ELSE
round(NVL(REVISEMO, 0)) END ) as CPFEMO,
round(NVL(trim(EPF), 0)) AS CPEPF,
0 as AEMO,
0 as APEPF,
'' as recsts,
EMPNO
FROM EMP_VALI
WHERE EFLAG = 'Y' AND SFLAG = 'N' AND EMPNO IS NOT NULL and
RECDATE >'01-Apr-2011')
union all
(SELECT NDT.RECDATE AS RECDATE,
sum(round(NVL(trim(NDT.EMO), 0))) as CPFEMO,
sum(round(NVL(trim(NDT.EPF), 0))) as CPEPF,
0 as AEMO,
0 AS APEPF,
NDT.EMPNO
FROM EMP_VALI VAL, EMP_SUPP NDT
WHERE VAL.EMPNO = NDT.EMPNO AND VAL.EFLAG = NDT.EFLAG AND
VAL.EFLAG = 'Y' AND VAL.SFLAG = 'Y' AND
NDT.SLIFLAG='N' and
VAL.EMPNO is not null and
NDT.RECDATE = VAL.RECDATE
GROUP BY NDT.RECDATE, NDT.EMPNO) UNION ALL
(SELECT DT.RECPAIDDATE AS RECDATE,
0 as CPFEMO,
0 as CPEPF,
sum(round(NVL(trim(DT.EMO), 0))) as AEMO,
sum(round(NVL(trim(DT.EPF), 0))) AS APEPF,
max('') as recsts,
DT.EMPNO
FROM EMP_VALI VAL, EMP_SUPP DT
WHERE VAL.EMPNO = DT.EMPNO AND VAL.EFLAG = DT.EFLAG AND
VAL.EFLAG = 'Y' AND VAL.SFLAG = 'Y' AND
VAL.EMPNO IS NOT NULL and dt.RECDATE=val.RECDATE AND DT.SFLAG IS NOT NULL AND DT.SFLAG not in ('N','F')
GROUP BY DT.RECPAIDDATE, DT.EMPNO)UNION ALL
(SELECT DT.RECPAIDDATE AS RECDATE,
SUM((CASE
WHEN (DT.ECR4FLAG = 'C') then
round(NVL(trim(DT.EMO), 0))
else
0
end)) as CPFEMO,
sum((CASE
WHEN DT.ECR4FLAG = 'C' then
round(NVL(trim(DT.EPF), 0))
else
0
end)) as CPEPF,
sum((CASE
WHEN DT.ECR4FLAG = 'A' then
round(NVL(trim(DT.EMO), 0))
else
0
end)) as AEMO,
sum((CASE
WHEN DT.ECR4FLAG = 'A' then
round(NVL(trim(DT.EPF), 0))
else
0
end)) as APEPF,
max(EMPRECOVERYSTS) as recsts,
DT.EMPNO
FROM EMP_VALI VAL, EMP_SUPP DT
WHERE VAL.EMPNO = DT.EMPNO AND VAL.EFLAG = DT.EFLAG AND
VAL.EFLAG = 'Y' AND VAL.SFLAG = 'Y' AND
VAL.EMPRECSTS = 'DEP' AND VAL.EMPNO IS NOT NULL and
dt.RECDATE = val.RECDATE AND DT.SFLAG IS NOT NULL AND
DT.SFLAG in ('F')
GROUP BY DT.RECPAIDDATE, DT.EMPNO))
GROUP BY RECDATE, EMPNO)
/
It's hard to tell from the statement, but if I had to guess, I put my money on the expression:
RECDATE >'01-Apr-2011'
assuming the column RECDATE is actually of type DATE. Therefor Oracle tries to convert the character value '01-Apr-2011' to a DATE as well. As you did not specify an format mask for this, the default NLS settings are used. If they define a number for the month then the above value would fail conversion.
You should never rely on implicit data type conversion. Especially not with dates. Use an ANSI literal instead:
RECDATE > DATE '2011-04-01'
or use the to_date() function with a format mask:
RECDATE > to_date('01-Apr-2011', 'dd-mon-yyyy')
Note that this could still fail for certain settings of NLS_LANG. In French you would need to specify 'Avr' instead of 'Apr'. So unless you are absolutely certain you can control all NLS_XXX settings all the time I'd strongly suggest to use month numbers instead. If you are more comfortable using to_date() than ANSI literals, you can use:
RECDATE > to_date('01-04-2011', 'dd-mm-yyyy')
Edit
if it's not the date column you need to check any other column for implicit data conversions.
These expression:
sum(round(NVL(trim(DT.EMO), 0)))
round(NVL(trim(EMOLUMENTS), 0))
round(NVL(trim(DT.EMO), 0))
round(NVL(trim(DT.EPF), 0))
look suspicious. If the columns in there are real numbers, then trim() is invalid and useless. If those are not numbers they could cause that error depending on the content of the column.
This expression to_number(SUM(CPFEMO)) is also useless as sum() will already return a number there is no reason to call to_number on a number(). Although I doubt it could raise your error you should still avoid it as it does not make any sense.
Following is sample query.
CREATE PROCEDURE GetModel
(
#brandids varchar(100), -- brandid="1,2,3"
#bodystyleid varchar(100) -- bodystyleid="1,2,3"
)
AS
select * from model
where brandid in (#brandids) -- use a UDF to return table for comma delimited string
and bodystyleid in (#bodystyleid)
My requirement is that if #brandids or #bodystyleid is blank, query should return all rows for that condition.
Please guide me how to do this? Also suggest how to write this query to optimize performance.
You'll need dynamic SQL or a split function for this anyway, since IN ('1,2,3') is not the same as IN (1,2,3).
Split function:
CREATE FUNCTION dbo.SplitInts
(
#List VARCHAR(MAX),
#Delimiter CHAR(1)
)
RETURNS TABLE
AS
RETURN ( SELECT Item = CONVERT(INT, Item) FROM (
SELECT Item = x.i.value('(./text())[1]', 'int') FROM (
SELECT [XML] = CONVERT(XML, '<i>' + REPLACE(#List, #Delimiter, '</i><i>')
+ '</i>').query('.') ) AS a CROSS APPLY [XML].nodes('i') AS x(i)) AS y
WHERE Item IS NOT NULL
);
Code becomes something like:
SELECT m.col1, m.col2 FROM dbo.model AS m
LEFT OUTER JOIN dbo.SplitInts(NULLIF(#brandids, ''), ',') AS br
ON m.brandid = COALESCE(br.Item, m.brandid)
LEFT OUTER JOIN dbo.SplitInts(NULLIF(#bodystyleid, ''), ',') AS bs
ON m.bodystyleid = COALESCE(bs.Item, m.bodystyleid)
WHERE (NULLIF(#brandids, '') IS NULL OR br.Item IS NOT NULL)
AND (NULLIF(#bodystyleid, '') IS NULL OR bs.Item IS NOT NULL);
(Note that I added a lot of NULLIF handling here... if these parameters don't have a value, you should be passing NULL, not "blank".)
Dynamic SQL, which will have much less chance of leading to bad plans due to parameter sniffing, would be:
DECLARE #sql NVARCHAR(MAX);
SET #sql = N'SELECT columns FROM dbo.model
WHERE 1 = 1 '
+ COALESCE(' AND brandid IN (' + #brandids + ')', '')
+ COALESCE(' AND bodystyleid IN (' + #bodystyleid + ')', '');
EXEC sp_executesql #sql;
Of course as #JamieCee points out, dynamic SQL could be vulnerable to injection, as you'll discover if you search for dynamic SQL anywhere. So if you don't trust your input, you'll want to guard against potential injection attacks. Just like you would if you were assembling ad hoc SQL inside your application code.
When you move to SQL Server 2008 or better, you should look at table-valued parameters (example here).
if(#brandids = '' or #brandids is null)
Begin
Set #brandids = 'brandid'
End
if(#bodystyleid = '' or #bodystyleid is null)
Begin
Set #bodystyleid = 'bodystyleid'
End
Exec('select * from model where brandid in (' + #brandids + ')
and bodystyleid in (' + #bodystyleid + ')')