Commit bee1750
slides: replace lottery-table slide with two SQL-example slides (win + loss)
Replace the 'lottery in numbers' table with two slides, each showing
the actual SQL of one ClickBench query:
- Slide 3 'When pushdown wins big': Q23 (SELECT * ... WHERE URL LIKE
'%google%' ... LIMIT 10), main 3 612 ms → +pd 121 ms, 30× faster.
- Slide 4 'When pushdown loses big': Q30 (... WHERE SearchPhrase <> ''
GROUP BY ...), main 276 ms → +pd 547 ms, 1.98× slower.
Both queries hit the same ClickBench dataset — same flag, opposite
outcomes. Speaker notes give the mechanism: row-level eval + sparse
RowSelection → page-skipping (win) vs mandatory unselective filter
with no row-skipping payoff and an extra column read (loss).
Renumber slide-comment headers to keep them sequential (1..10).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>1 parent 74e9ac3 commit bee1750
2 files changed
Lines changed: 130 additions & 89 deletions
0 commit comments