[explainability] python script to calculate namescore of slt files #31877
Conversation
mgree
added
T-UX
|
Oh, this is great! I think the result is so low only because we don't have humanization turned on in most of our EXPLAIN statements in slts. It defaults to turned on in staging/prod, but otherwise it defaults to off. The option is (There are several such EXPLAIN options. This is the tracking issue: https://github.com/MaterializeInc/database-issues/issues/8097) |
|
Isn't humanized expressions on by default? It's what the docs say, though maybe default |
That is only in staging/prod. It's off in tests, see details in the above-linked https://github.com/MaterializeInc/database-issues/issues/8097
No, those options that don't occur in the |
Oh! I see---I misunderstood. I'll turn it on explicitly for tests in this PR, then. |
mgree
force-pushed
the
explain-namescore
branch
from
694bffe to
a329c8d
Compare
|
Okay, you're absolutely right: we do much better when we actually ask for names. 😁 |
ggevay
left a comment
ggevay
left a comment
There was a problem hiding this comment.
Looks good! We could maybe run Nightly before merging, just to be safe.
|
The testdrive failure is an unrelated timeout (also an issue on main see https://github.com/MaterializeInc/database-issues/issues/9095)---the salient bits look good! Merging. |
2 participants
#31802 introduces the concept of a
NAMESCOREmetric: the percentage of column references inEXPLAINoutput which have associated names.This PR introduces a python script,
namescore, that implements that metric.Output on old SLT files (without `WITH (humanized expressions)`)
Output on SLT files (adding `WITH (humanized expressions)` explicitly)
Motivation
[design] better column names #31802
https://github.com/MaterializeInc/database-issues/issues/8960
Checklist
$T ⇔ Proto$Tmapping (possibly in a backwards-incompatible way), then it is tagged with aT-protolabel.