Skip to content

pad memory and CPU profiles with None to stabilize schema - #74

Merged
nicholasjng merged 4 commits into
masterfrom
null-padding
Nov 19, 2026
Merged

nicholasjng merged 4 commits into
masterfrom
null-padding

Conversation

@nicholasjng

Copy link
Copy Markdown
Owner

This simplifies duckDB queries on JSON archives.

NS_PER_UNIT and Selector.includes had no callers left, and the
completion generator, path display and selector checks were helpers
used once, so reading their call sites meant a jump for no reuse. The
note on `host` sat under an unrelated constant in compare.py and now
lives with SKEW_FIELDS, the tuple it explains.
The console column sizing test repeated one in test_runner.py, the
slash-name and empty-run tests asserted behavior no code path can break
on its own, and the plain .jsonl DuckDB case tested DuckDB's reader
rather than anything the gzipped case leaves uncovered. They cost run
time without guarding mew.
…ut a measurement

DuckDB infers columns from the keys it sees, so a file no profiler
wrote to had no memory or cpu_profile column, and a query naming one
failed to bind instead of returning NULL. Padding with null, as skip
already is, keeps the columns in every file without implying a
measurement, where an empty object would infer a map that rejects
arithmetic. Readers checking for a block now test for None rather than
key presence.
Keeps the uv-lock hook on the uv release that wrote the lockfile, and
picks up the current ty and docs toolchain releases so local checks
match what CI resolves.
@nicholasjng
nicholasjng merged commit f551251 into master Nov 19, 2026
21 checks passed
@nicholasjng
nicholasjng deleted the null-padding branch November 19, 2026 13:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant