I have observed an abnormal behaviour in Extract Document Data activity in document understanding.
When using this activity, and choosing IXP model then create output variable, if type was a IDocumentData collection of GptIxp14, later I published a new version and found below error:
Argument ‘ExtractionResults’: BC30512: Option Strict On disallows implicit conversions from ‘IDocumentData(Of GptIxp15)’ to ‘IDocumentData(Of GptIxp14)’. The selected value is incompatible with the property type.
So as per error it expects IDocumentData(Of GptIxp15) and only workaround I see is create new variable and map everywhere used.
This is not the right behaviour and very hard to maintain.
I can explain this, its a similar root cause to something that can happen to ready an Excel table.
With various activities they generate a custom data type based on the data you are reading, with the Excel situation I mentioned its based on the spreadsheet table columns etc, for IXP, its based on the fields you use in the extraction.
The number at the end is fairly meaningless. Its just a counter.
If the data type changes, it means you have likely updated the IXP fields, causing the taxonomy to change and then the data type to be remade. This then results in a new data type with a new number of the end. You do then have to remake the variables etc to select the new data type. You can turn strong data typing off, which suppresses this issue by making all the data types a generic one, but I wouldnt suggest this as you lose alot of the benefits from working with the strong data type.
yes, I did make a slight change in the field - changed name of one field. Post that model version changed from version 14 to version 15, also the data type required also changes from GtpIxp14 to GptIxp15, so this I felt it’s unusual behaviour.
Thats not unusual at all. I am not sure how you expect the code to work if it infact doesn’t make a new version of the concrete data type, otherwise the field name would mismatch, aswell as the data type.
Actually I feel unusual because every time I publish new version of model, I need to recreate new variable, map it again and redeploy even though if it is just updated prompts and no field updates.
Yeah, I understand why you’d want that to stay the same data type, but as you see it regenerates. I think the code the actually compare if an update is needed vs just regenerating is much more complex.
Hi Rohit,
I totally agree.
In my experience even though the Taxonomy fields are not changed, changes to the IXP “Instruction / Description” to get a better value into the same field still result in a change to the “GtpIxp” reference - which causes code changes to be necessary. It means work to apply the change through the code if features such as “Action Center” are used.
I can appreciate the Taxonmy structure changes but I think UiPath have it wrong for the extraction instructions to cause such an impact.