HiTaS cannot load JJ-routine

[SimonBBK]

Please specify
SDC tool used: TauArgus
Version used: 4.2
Operating system used: Windows 10

Hello,

I am Simon and I am working for Deutsche Bundesbank. We are using TauArgus now for one year. We used to have version 4.1.7 and were able to flag tables. We changed to version 4.2 and now for most tables that worked with 4.1.7 we now receive the error "HiTaS cannot load JJ-routine".

This table works with 4.2
working.zip

This table produces the error under 4.2 but works under 4.1.7
notworking.zip

We now switched back to version 4.1.7 for the time being.

I hope this information helps to reproduce the error and hopefully not locating the mistake in front of the screen...

Best regards

Simon

screenvideo of table working in 4.1.7
https://user-images.githubusercontent.com/77275568/105852578-f2770280-5fe4-11eb-8fe6-9eeecfbcbaa0.mp4

screenvideo of same table not working in 4.2
https://user-images.githubusercontent.com/77275568/105852529-e2f7b980-5fe4-11eb-9bf9-49cfeaac87e4.mp4

screenvideo of a different table working in 4.2
https://user-images.githubusercontent.com/77275568/105852553-ec812180-5fe4-11eb-8789-84a8125509d5.mp4

[ppdewolf]

Additional question: what solver are you using? Cplex, Xpress, Free?

[SimonBBK]

Sorry for the late reply. Thought i activated to get informed if there is a reply.

We are using the free solver.

[jeroenborghuis]

Hi,
I receive the same error when using the runArgusBatchFile function (after using the createArgusInput function) in sdcTable (0.31), using CPLEX solver and method="MOD":
"Error in modular suppression procedure
HiTaS: couldn't load JJ-routines"
Can anyone help solve this issue?

R code:
dimList_sbi2d_ld <- list(sbi2d = dimsbi2d, Lndl = dimLD)

p_sbi2d_ld <- makeProblem(
data = ako,
dimList = dimList_sbi2d_ld,
numVarInd = "Tot_AK")

primSuppRules <- list()
primSuppRules[[1]] <- list(type = "p", n = 1, p = 10)

bO_md2 <- createArgusInput(
obj = p_sbi2d_ld,
typ = "microdata",
path = "F:/Desktop/Tau",
solver = "CPLEX",
method = "MOD",
SingleSingle = 1,
SingleMultiple = 1,
MinFreq = 0,
primSuppRules = primSuppRules,
responsevar = "Tot_AK",
licensefile = "Z:/TauArgus/4.2.0/access.ilm")

runArgusBatchFile(
batchF = "F:/Desktop/Tau/batch_0ik2r5liuI.arb", #bij iedere run 'createArgusInput' nieuwe batch file selecteren!
exe = "Z:/TauArgus/4.2.0/TauArgus.exe",
batchDataDir = NULL,
verbose = T)

Log file:
24-feb-2021 17:35:47 : "F:/Desktop/Tau/microdata_0ik2r5liuI.asc"
24-feb-2021 17:35:47 : "F:/Desktop/Tau/metadata_0ik2r5liuI.rda"
24-feb-2021 17:35:47 : "sbi2d""Lndl"|"Tot_AK"|""|""
24-feb-2021 17:35:47 : P(10,1)|
24-feb-2021 17:35:47 :
24-feb-2021 17:35:47 : Start explore file: F:/Desktop/Tau/microdata_0ik2r5liuI.asc
24-feb-2021 17:35:50 : Start computing tables
24-feb-2021 17:35:56 : Table: sbi2d x Lndl | Tot_AK has been specified
24-feb-2021 17:35:56 : Tables have been computed
24-feb-2021 17:35:56 : Micro data file read; processing time 9 seconds
24-feb-2021 17:35:56 : Tables from microdata have been read
24-feb-2021 17:35:56 : CPLEX,"Z:/TauArgus/4.2.0/access.ilm"
24-feb-2021 17:35:56 : MOD(1,1,1,1,0)
24-feb-2021 17:35:56 : Start of the modular protection for table sbi2d x Lndl | Tot_AK
24-feb-2021 17:35:57 : Error in modular suppression procedure
HiTaS: couldn't load JJ-routines
24-feb-2021 17:35:57 : Error in batch file
24-feb-2021 17:35:57 : End of TauArgus run

[SimonBBK]

Please specify
SDC tool used: TauArgus
Version used: 4.1.7
Supression: Modular, Solver Free
Operating system used: Windows 10

Hi,

I tend to get the same error message all the time....

I have the same error message now also with version 4.1.7.

Unfortunatly I cannot share the data here yet, as they are still "originals". I experimented a bit and it seems to be related to the patten or amount of cells that are flagged as "C". The very same data with one patten/amount of "C"-cells works but with a different one produces the error. The two tab-files differ only by the flagging not by the value or anything else.

[ppdewolf]

In case of the situation mentioned by @jeroenborghuis the problem turned out to be produced by the fact that the number of significant digits in his table was too large (13 significant digits of which two decimals). This is over the machine precision and thus sometimes produces non-additivity. When he divided all contributions by 1000 and removed the decimals, the Modular approach worked as expected.
@SimonBBK: Perhaps something similar is present in your data? Moreover: what do you mean by the "C"-flag?

[SimonBBK]

With C-falg I mean unsafe cells that have a frequency of 1 or 2.
In the data example for 4.2. the biggest value is 1452900988 and this is a "normal" size. I try to shrink it but this might cause additivity problems in most of our cases.

[KarinaKelleher]

Getting same message as reported here (HiTaS cannot load JJ-routines).

I decided to add to this thread rather than start a new one as getting same issue. We are planning on carrying out some testing, comparing capabilities of the free solvers and a commercial solver. As part of this we plan on looking at linked tables functionality - using modular suppression. However we are getting issues with modular suppression - even though secondary suppressions were required it was giving "0 cells suppressed".

Investigating further, we found an issue with modular suppression on single tables also (up to now we would have been using optimal solver) - We are getting the above message "Error in modular suppression procedure. HiTaS: couldn't load JJ-routines" on these.

I have completely simplified our files etc. to try to eliminate anything that could be causing this, as you can see in the relevant files attached.

I have tested this on:

TAU-Argus - Version 4.2.0 Build 5 and Version 4.2.4 Build 2 - Using free solver and using FICO Xpress solver and get this message in all cases.

Any ideas on what could be causing this issue?

Also I assume that others have carried out similar comparisons of the free and commercial solvers (we are trialing XPress) - As mentioned we are particularly interested in additional capabilities for linked tables but any pointers on what others consider benefits of commercial over free solvers would be also very much appreciated, Thanks.

TestDiff.zip

[ppdewolf]

See #262