fix: cartesian products in get_evaluations #3081
Conversation
dosubot
bot
added
the
size:M
we don't have a set up for server integration tests in general |
| connectable, | ||
| index_col=["span_id", "document_position"], | ||
| ) | ||
| ).set_index(["span_id", "document_position"]) |
There was a problem hiding this comment.
Reason to prefer this to index_col?
There was a problem hiding this comment.
It didn't work. Don't know why
| .join_from(models.Span, models.Trace) | ||
| .join_from(models.Trace, models.Project) | ||
| .where(models.Project.name == project_name) | ||
| .where(models.DocumentAnnotation.annotator_kind == "LLM"), |
| .join_from(models.Span, models.Trace) | ||
| .join_from(models.Trace, models.Project) | ||
| .where(models.Project.name == project_name) | ||
| .where(models.DocumentAnnotation.annotator_kind == "LLM"), |
There was a problem hiding this comment.
So strange to me that sqlalchemy doesn't complain here.
I think we can seed data use SQLAlchemy and test the query at whatever level you find appropriate, something like here: https://github.com/Arize-ai/phoenix/blob/sql/tests/db/test_models.py, is this setup insufficient to catch this? |
yea, that would work |
No description provided.