Fix LT-22249: Don't modify existing analysis on flextext import#442
Conversation
|
I haven't completely evaluated my logic related to the morpheme matching, this is working for the initial test case but needs more testing of different scenarios before we consider it complete. |
jtmaxwell3
left a comment
jtmaxwell3
left a comment
There was a problem hiding this comment.
When I looked at this code today, the bug was that FindExistingGloss found an analysis that matched the wordform and the gloss but not the morphemes, and that the code that followed FindExistingGloss changed the morphemes. Don't you have the same problem if FindExistingAnalysisStack fails?
|
The bug is in UpdateToWordGloss, not FindExistingGloss. Also, I don't see how FindExistingAnalysisStack verifies that the morphemes are in the right order or that there aren't extra morphemes or glosses in the analysis. |
jtmaxwell3
left a comment
jtmaxwell3
left a comment
There was a problem hiding this comment.
I read through all of the code and I didn't see any problems. I'm trusting the test cases to catch anything I might have missed in the code.
* Reuse wordforms that are matches but create new anaysis for them * Add Normalization before string comparison * Split Morpheme and glossing logic Goal: If the imported data is a subset of an existing analysis then use the existing analysis, otherwise create a new analysis and add it to a matching Wordform.
jasonleenaylor
force-pushed
the
bugfix/flextextImport
branch
from
e408947 to
1de62b6
Compare
* Reuse wordforms that are matches but create new anaysis for them * Add Normalization before string comparison * Split Morpheme and glossing logic Goal: If the imported data is a subset of an existing analysis then use the existing analysis, otherwise create a new analysis and add it to a matching Wordform. # Conflicts: # Src/LexText/Interlinear/BIRDInterlinearImporter.cs
2 participants
This change is