morleycb
commented
4 months ago Hey Dirk,
This is due to the same underlying issue as issue 1254. At the time my analysis was the following (this is for label groups, but the same concept applies to Lexemes):
The label group "ref" (like a non system specific id) "LabelGroup_RXh0ZXJuYWxfenpfU1VS" should get decoded into the actual label group name. I just did a super quick test to see what "RXh0ZXJuYWxfenpfU1VS" would get decoded into (using Encoding.UTF8.GetString(Convert.FromBase64String("RXh0ZXJuYWxfenpfU1VS"))) and it produces three parts: "External", "zz", "SUR". So I'm guessing that the label group name someone created is "External_zz_SUR".
So the bug here is that the collaboration code is using _ as a general purpose delimiter in these refs, not taking into account that one of the underlying values being encoded might contain an _ itself.
Here are few of relatively simple fix ideas off the top of my head:
- Tweak the label group create dialog to not allow underscore characters to be in a label group name
- Change the collaboration code to use a different delimiter that we know won't be in any label group, label, semantic domain, lexeme, meaning, lexicon translation, i.e. all the entity types that use this encode/decode logic.
- I think the 'encode' and 'decode' methods used by all the specific entities are in a class called GeneralModelBuilder. This is a better long term fix, but any existing collaboration data already containing refs encoded using "_" won't be decodeable anymore. This secondary issue could be mitigated by implementing the decode method to have 'fallback' logic. E.g.
- First try to decode using the new delimiter. If the resulting number of parts doesn't match what was expected, try the older "_" as the delimiter, and if neither match, throw the PropertyResolutionException.
- Change the decode logic for entity types that only have a single value contained in their 'ref' (i.e. label, label group and semantic domain): just simply don't call 'split' when the expected number of parts is 1. This approach could still leave a potential a future issue, though, with lexeme, meaning or lexicon translation if say a lexeme's lemma contains an "_" character. This would be super easy. Just change the existing var parts = decodedString.Split(delimiter); to something like var parts = (expectedNumberOfParts > 1) ? decodedString.Split(delimiter) : [ decodedString ];
Description:
A user who has a project with active modification to the glosses is throwing the following Collab Import error:
Steps To Reproduce:
Found in Version:
1.4.1.4
Fixed in Version:
(Clear Dashboard version in which the bug was fixed.)
Test Cases:
(Linked Test Cases)