Pathfinder: templates appear to not be filled out properly #893
Comments
colleenXu
changed the title
tokebe
changed the title
|
@colleenXu
In Example 5 (Template B)
In Example 5 (Template C)
In Example 3 (Template C):
Maybe some sort of gene/protein conflation needs to be implemented in Pathfinder? Or should there be different behavior when no categories are found in the category table? |
|
Odd, I was under the impression that Gene/Protein conflation would happen prior to pathfinder handling, so Pathfinder template generation would have both to work with from the get go... If that's not the case, we'll want an extra Gene/Protein conflation pass prior to template building in Pathfinder (Or, alternatively, Gene/Protein expansion when generating the in-memory table from the hard table) If no categories are found, we'll want NamedThing, though again I was expecting that the internal query execution of a template would handle injecting that... |
|
From my POV, there's several things to investigate/untangle:
Helpful to look back at slide deck and info in #832 |
|
Andrew told me that we don't have to rush to fix this. It's fine to not get this done by the end of this phase. |
What's happening
(1) The filled-out templates written in the TRAPI/console logs look strange. They're missing category/predicate assignments. Is that okay? (these logs come from #862)
From running example 5: affects all 3 filled-out templates
From running example 3: affects only 3rd filled-out template ("C")
(2) Some of those filled-out templates terminate early because no matching MetaKG edges were found.
Logs
Example 5:
2024-10-24T07:38:11.514Z WARNING: [Template-2]: Query Edges [(reversed b_obj), (reversed un_b)] have no MetaKG edges. Your query terminates.2024-10-24T07:38:10.841Z WARNING: [Template-3]: Query Edges [(sub_c), (c_un)] have no MetaKG edges. Your query terminates.Example 3:
2024-10-24T07:51:30.533Z WARNING: [Template-3]: Query Edges [(sub_c), (c_un)] have no MetaKG edges. Your query terminates.(3) The templates that terminate early have odd console logs, where the
input_typeoroutput_typeof a QEdge to execute (qXEdge? APIEdge?) is an empty array.From Example 5, 2nd template
From Example 5, 3rd template
From Example 3, 3rd template
Data
Commentary
This is likely happening in CI + Test, since I found examples of it happening Friday 10/18 and it's happening in my local (all main branches) right now.
I'm not sure if this is "expected" behavior or not. I lean towards unexpected, because I think the category/predicate mappings do exist but aren't being used in some templates. Some I noticed:
The text was updated successfully, but these errors were encountered: