Open NicoLaval opened 4 months ago
NicoLaval
commented
4 months ago Substring syntax (In Trevas param2 is the start_index, and param3 the length. The user test seems to assume, that param2 is the start_index and param3 the end_index (inclusive))
@noahboerger could you confirm substr doesn't cause exception throw but a wrong result because of the previous remark?
noahboerger
commented
4 months ago The first note on substr was just that it is currently not supported in the shape DS_r := substr( DS_1, 1, 2).
The different syntax also seems to be the correct one according to the reference manual (p. 76 f.). So i would propose to close the issue as it is only an issue for our used testcases.
NicoLaval
commented
4 months ago This syntax has to work, and is implemented in Trevas I think. The condition on the handled dataset is: dataset with only string measures.
@noahboerger, can you provide a failed example with the stack trace please?
noahboerger
commented
4 months ago @NicoLaval you are right, with a dataset with only string measures it is working also on my side. The dataset i have tested it with had the wrong format
| Id_1 | Id_2 | num_1 | Text_1 |
|---|---|---|---|
| IDENTIFIER | IDENTIFIER | MEASURE | MEASURE |
| STRING | STRING | DOUBLE | STRING |
The concrete test case i have executed is the one from BdI "11. conditional/if-then-else_2", which is trying to execute it with a dataset that is also containing a measure with an additional double type. This testcase has failed with the following error:
The test from BdI assumes that the additional non string measure is kept without being manipulated. But I think your implementation is right, as the table of VTL-ML-Operators in the reference manual (p. 17), requires the input to be in the following form:
op ::
dataset { measure<string> _+ }
| component<string>
| string
pattern1, pattern2 ::
component<string>
| stringso i think it the correct behaviour to fail here.
I did not dive so deep into this test case when i executed it the first time, but then it is a wrong test case. I am not sure if the trevas error message could be a bit more clear, as the engine still tries to execute the replace-operator even with a mismatching input dataset.
@noahboerger you reported
substris not supported.I thought it was, for scalars, components and datasets.
Could you report a failed example please?