Exception with llun handling for non-text data

[antoonbronselaer]

Problem

When running llunatic on this dataset but stored in a postgres with integer values, I get an exception because at some point, llun values are attempted to be inserted in the variable column of the work schema.

For example, the statement:

INSERT INTO work.allergen_dataset__hazelnut (tid, step, hazelnut, cellGroupId, cellOrigValue) VALUES (890, 'r.e18_0_f#.e1_0_f#.e3_0_f#.e5_0_f#.e20_0_f#.e7_0_f#.e21_0_f#.e10_0_f#.e12_0_f#.e14_0_f#.e16_0_f#.e19_0_f#.e2_0_f#.e4_0_f#.e6_0_f#.e8_0_f#.e11_0_f#.e9_0_f#', _L9789, '_L9789', '0');

causes an error because the datatype of column "hazelnut" is integer and it tries to insert _L9789, which is not an integer

Proposed solution

Llunatic could store text representations of these values (it does so already with cellorigvalue I think) and then parse to the correct datatype whenever necessary.

[donatellosantoro]

Hi Antoon,
thanks for your report.

In the source schema in PostgreSQL what is the type of the column 'hazelnut'?
A quick fix should be change all the types in 'text'.
In alternative you could try to enable the 'UseDictionaryEncoding' feature in the configuration.

P.s.: can you please share the whole scenario (csv data + scenario.xml) so we can reproduce the issue? Thanks

[antoonbronselaer]

Hi Donatello,

thanks for the quick reply!

As mentioned in the original message, the datatypes are integer. I realize that casting to text works (I tried that before I opened the issue) but I thought I'd be nicer if Llunatic handles this. I have not tried the feature 'UseDictionaryEncoding' but I will have a look at it.

The dataset can be found in the link of the first post. I also attach the scenario file that causes the error. I'll add it as a text file because Github prevents me from adding XML files.

allergen-chase-fo.txt