Closed yonzarecki closed 7 years ago
Moving to Sprint 2, as this issue is not critical to our commitments for Sprint 1.
@tonylekhtman Can you help me get started ? Where do I need to go if i just wanna send some SQL queries and look at the results ?
The best way in my opinion is via mysql workbench. It presents the results in a nice tabular form.
You can also send queries via the teminal. You need to write mysql -u
I got it working, looking d:
From looking at the questions I can see that most exceptions are under a "< code> >" tags, this can help with exception detection and extractions. (not always tho)
There are also many duplicates in the DB (this may result in duplicate results in query time)
Another idea is to keep an index of all rows seen, and give lower significance to rows not in the index (probably user-specific) in the distance functions.
I suppose this is already implemented, but most newlines in the DB are represented by " ", if it doesn't it can help with exception parsing.
We should look more into the exception trace print types. There are several of them, yet they all represent the same thing. A few examples are 268 vs. 261 vs. 262 etc.
A good idea is maybe to analyse these types better and make a "generic" print for us to compare, this way we won't miss a good answer because of different printing styles.
I think I'm done for now, this gave us enough work for the time being. I'll open issues about these topics tomorrow.
As suggested by Yossi, this is a good time to start looking at our stack-traces data, and look for hints within it. Can we see some similarity between traces that we can translate into better performing algorithms ?
We should add prints and sample instances from the db and see if we can realize something important from looking at our data.