Implemented in #61 — measured, and it reaches the floor
Same four-way comparison re-run against the production corpus with the shipped
implementation as one of the arms. DUCKDB_THREADS=2,…
Implementation notes (from the profiling above)
Where the IN list is built today
.sql_lit_chr()—R/spending.R:737. Renders'a','b',...; harmless in itself..build_verb_sql()…
Closing in favour of #58, which is where the rollup win actually is.
Profiling (detail in the comment above) showed this issue targets ~1% of the runtime: the
many-to-many left_join is 6 ms…
Measured: this is worth more than the issue claims, and it is the fix for #59 too
Same aggregate over the 20,106-government type = city cohort (FY2022), with the cohort
expressed four ways…
Measured before implementing, and the premise here does not hold
This issue proposes moving the aggregation and crosswalk join into SQL, on the theory that "the many-to-many join is the part…
Evidence from the cog-api side: the overhead is remote I/O, not compute
While doing a performance pass on cog-api (which is a thin wrapper over these verbs), I
measured the same call shapes…
@/tmp/claude-1000/-home-jared-Nextcloud-Civilytics-Code-Civilytics-cog-explorer/4bbb145c-ac06-4d88-84be-1e596655c718/scratchpad/c46.md