TI crashes TM1 server and inconsistent feeders

[harrytm1]

Hi,

I'm currently running 9.5.1 on my local PC which is on Windows 7 and Excel 2003.

I have a cube that allows cash flow planning. Actual and forecast data is by dates. As forecasting is done every week, the version is "Week Version". Here are some rules for this cube:

# All data in actualised dates in Actual Week Version to be copied to corresponding dates in all Forecast Week Versions.
# 'WV ACT' is the Actual version. 'Last Actual Date Serial' is the TM1 Date Serial for the 1st day of the forecast week.
# Each Date element has its TM1 Date Serial number as an attribute.

[] = N:
IF( ATTRN('Week Version', !Week Version,'Type') = 1,
IF(ATTRN('Date', !Date, 'Date Serial') < ATTRN('Week Version', !Week Version, 'Last Actual Date Serial'),
['WV ACT'],
CONTINUE),
CONTINUE);

# 'Cash Available' is the c-element in Cash Management dimension.
# Ensure that consolidation is along Cash Management dim.
['Cash Available'] = ConsolidateChildren('Cash Management');

Problem 1:
The feeder for the 1st rule is:
# Feed "WV ACT" version to all forecast week versions
['WV ACT'] => ['WV_GRP001'];

where 'WV_GRP001' is the parent of ALL Forecast Week Version elements. When the rule is saved, everything seems to be working in the Cube View i.e. all data in WV ACT version appear in the dates with Date Serial less than that of the Forecast Week Version element. E.g. if the Forecast Week is "27 Dec 2010", all dates prior to 27 Dec 2010 will be pulled from WV ACT version.

However, when I deliberately delete the data in a cell in WV ACT version and the date is prior to 27 Dec 2010, then enter another value in another cell in another date (still prior to 27 Dec) in WV ACT, I noted that in the Forecast Week version, the rule-derived corresponding value is NOT FED. I'm not sure if the rule or feeder is causing this problem. The weird thing is, why is it working earlier and not now? I need to save the rule again to get it working. but the moment I repeat the clear and re-enter test case again, the no-feeding error happens.

Problem 2:
Please refer to attached image. When I run a TI that copies forecast data from one version to another, the TM1 server crashes. The message is "Runtime Error! Program C:\Program Files\Cognos\TM1\bin\tm1s.exe
This application has requested the Runtime to terminate it in an unusual way. Please contact the application's support team for more information."

FYI, the TI first creates a new Forecast Week version element and then copies the data from a source version into the new element. Also, the errors above happen for both original 9.5.1 and after I had applied Hotfix 11.

Any advise is greatly appreciated! Thanks!

[harrytm1]

By the way, the TI works only if I delete the cube rule before running the TI. I was hoping to find a function to delete the rule in TI script, import the data, then load the rule from a file again. But I can't find the function to delete.

Also, I have changed the rule by replacing [] with a measure element just to test the effect. The corresponding feeder was changed too. It works as planned. So is it because of [], plus the presence of other rules that affect the calculation of c-elements (mainly due to the handling of opening balance and ending balance base on the period that is selected) that cause the no-feeding issue?

FYI, the changed rule below DID NOT help the TI issue, the server still crashes. Only deleting the rule helps. I have also attached an image of the Week Version dimension for your reference. Thanks!

# All data in actualised dates in Actual Week Version to be copied to corresponding dates in all Forecast Week Versions.
['CMMEA101'] = N:
IF( ATTRN('Week Version', !Week Version,'Type') = 1,
IF(ATTRN('Date', !Date, 'Date Serial') < ATTRN('Week Version', !Week Version, 'Last Actual Date Serial'),
['WV ACT', 'CMMEA101'],
CONTINUE),
CONTINUE);

Feeders;
# Feed "WV ACT" version to all forecast week versions
['WV ACT', 'CMMEA101'] => ['WV_GRP001', 'CMMEA101'];

[tomok]

Also, I have changed the rule by replacing [] with a measure element just to test the effect. The corresponding feeder was changed too. It works as planned. So is it because of [], plus the presence of other rules that affect the calculation of c-elements (mainly due to the handling of opening balance and ending balance base on the period that is selected) that cause the no-feeding issue?

You understand that empty brackets, [], means the entire cube don't you? You should ALMOST never use this in a rule. In your example it is resulting in a circular reference. Basically, you are saying that if the week version type is 1 and the date is in a certain range then make EVERYTHING equal to WV_ACT, including WV_ACT, a circular reference.

[tomok]

By the way, the TI works only if I delete the cube rule before running the TI. I was hoping to find a function to delete the rule in TI script, import the data, then load the rule from a file again. But I can't find the function to delete.

it's amazing what you can find in the TM1 reference guide!

RuleLoadFromFile
This is a TM1 TurboIntegrator function, valid only in TurboIntegrator processes. This function creates a TM1 rule for a specified cube from a text file. The text file must be formatted according to TM1 rules conventions. Each rule statement must conclude with a semi-colon (;) and comments must be prefixed with the # character. If a rule already exists for the specified cube, the existing rule is overwritten by the rule created by RuleLoadFromFile.

Simply call this with an empty file, which will remove the rule, process your data, and the call it with a file with the rule statements in it.

[harrytm1]

Thanks, tomok.

I doubt it will result in circular reference as "WV ACT" version's "Type" attribute is 0, not 1. Hence, the rule does not apply to it. Type 1 attribute is only available for Forecast Week versions.

Yes, I did use the RuleLoadFromFile function before I posted here earlier. It did not seem to be working. I shall try again nonetheless.

If all else fails, I shall remove the rule and then create a TI to copy the Actual data to a selected Forecast Week version. The only thing is, any changes to actual data will not be updated automatically to the Forecast versions.

[harrytm1]

Hi,

I have attached the error log after running the TI. I have created a TI to:
1) in the Prolog, first load a dummy rule file using RuleLoadFromFile function. The text file only contains a "#".
2) in the Epilog, check that the new Week Version element does not exist in the dimension using DIMIX = 0. If DIMIX = 0, then create the new element and attributes by executing a process "Wee Version Dim_Add Element".
3) Next, execute another process "Data Copy_Cash Management".
4) Reload the original Rule Worksheet.

In seems from the error log, that the error occurs at Step 3 in the Epilog when the Data Copy process is running. There is a stack overflow. Also, I noted that the RuleLoadFromFile in Step 1 did not complete even when the TI proceeds to Epilog. This is evidenced from the Rule Editor which still shows the original codes.