Update xsv vs qsv stats comparison with qsv 28.0.0 benchmarks
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
136b49f
Update Benchmarks Data Schematic page (qsv 28.0.0)
21189b5
Troubleshooting: match the README; drop advice for errors that no longer occur
- SIGILL: lead with the AVX2 requirement (expected on non-AVX2 CPUs);
source builds are portable, not CPU-tuned; qsvdp and 23.0.x options
- from the README: PowerShell UTF-16LE / Excel first-column note, regex
look-around/backreference limits (validate --fancy-regex), OOM
prevention modes, the README's in-memory command list, sudo xattr
- stale indexes are rebuilt automatically ("please rebuild" error and
index --force don't exist); a leading UTF-8 BOM is stripped on read
- luau also missing from Linux ARM64 prebuilts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
4f3f4e9
FAQ: match the README's command symbols, variants and AI policy
- 🤯 list adds color, readstat, writestat; 🪄 list is the README's 12
commands (validate was not one)
- UI group is clipboard/color/lens/prompt; Windows x86_64 prebuilts lack
clipboard; pro lens is a separate Windows-only command
- Linux ARM64 prebuilts have a reduced feature set
- xsv/qsvlite wording and fork date from the README; csvkit ~10x
- Contributing: the README's AI Policy (credit the agent/model)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
77030a1
Comparison: correct xsv facts; current qsv numbers from the README
- xsv: 20 commands (was ~13); frequency/split/stats run in parallel
with an index (--jobs), so not "No" multithreading; last release
0.13.0, repo archived
- qsv 28: 80 commands, 28 default / 48 --everything stats columns,
validate up to 1,338,688 rows/sec
- fork date (Sept 2021) and qsvlite migration wording from the README
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
0542e31
Why qsv: match the README (tagline, goals/non-goals, current numbers)
- README tagline and a condensed Goals / Non-Goals section
- 80 commands (was 70+), five variants incl. qsvpy* (was four)
- validate up to 1,338,688 rows/sec (was 780,000); sub-second stats is
the default run, up to 48 measures with --everything
- geocoding: online providers too, not only the offline index
- csvkit ~10x, as on the Comparison page; qsv-cookiecutter (404)
replaced with qsv-client
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
c209273
Command Reference: match the README's symbols and legend
- 18 commands' symbols synced to the README's Available Commands table,
including the CKAN and Luau icons, 🎲 (seeded randomness) and 🌎
(geospatial); wiki-only symbols the README doesn't give were removed
- full 23-entry legend from the README (was an 11-entry subset)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3abaf6b
Getting Started, Binary Variants: match the README; which prebuilts have clipboard
- Getting Started: try-online, MCP/Cowork guide, qsv-client and Notebooks
pointers from the README; stats computes "up to 48" statistics, not 48
by default; lens also reads JSON arrays and Avro/IPC; note that most
prebuilts lack the ui feature, so clipboard is missing
- Binary Variants: ui is left out on macOS, Windows x86_64 and Linux
x86_64/ARM64; only Windows ARM64, IBM Power and IBM Z include it (per
the publish workflows)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
334950b
Installation: match the README's installation section
- qsv-client (Python) tip; Tool definitions section
- direct download links for qsv pro, prebuilt ZIPs, Easy Installer and
shell completions (qsv.dathere.com/download itself is a 404)
- PGO / CPU-optimization wording, sudo xattr, Homebrew completions
- cargo install builds the portable baseline, not a CPU-tuned binary
- s390x in the Linux list (ships in 28.0.0)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
27970ec
SQL-and-Polars: Polars 0b47007 bump (qsv #4993)
New sqlp SQL after 28.0.0: correlated IN with inequalities, mixed numeric
join keys, PERCENTILE_CONT/DISC, expression LIKE patterns, typed literals,
LIMIT subqueries. Result changes: IS DISTINCT FROM precedence, POSITION/
STRPOS of NULL, window row order. Known issue pola-rs/polars#30523.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
45411a1
Home: move the README badge row here; restore header placeholder
GitHub's wiki UI doesn't render _Header.md (checked logged in), so the
badges added in a44357a never showed. Put them at the top of Home.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
03fa7dd
Header: match the README's title and badge row
Replaces the placeholder comment. The MSRV badge links to the README's
section; the self-referential Wiki badge is left out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
a44357a
Footer: match the README (MIT only, third-party notices, AI Policy, DOI)
qsv is MIT-only since the Unlicense was removed after v21.1.1 (#4161);
the MIT link goes to COPYING, which explains the change. The old
UNLICENSE link pointed at a deleted file.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
e8bcd70
Sidebar: add README quick links; Benchmarks and describegpt model notes pages
Quick links mirror the qsv README's list (tour/notebooks/lessons,
benchmarks, gallery, Data Schematics, feature flags, goals, testing,
talks, "Have we achieved ACI?" series, sponsor).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
e539f82
Home, External Resources: add "Have we achieved ACI?" part 4 (Data Schematic post)
Matches the series list in qsv's README.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
6083e61
Release 28.0.0: Home banner, describegpt answer cache, benchmarks, CI hardening
- Home: 28.0.0 release banner and breaking changes (CSVW datatype object,
CKAN-group themes as Concepts)
- AI & Documentation: describegpt answer-cache section (#4970, #4971,
#4978, #4980); language-detection fixes (#4976, #4979)
- Benchmarks: benchmarks.sh 10.x on hyperfine 2.0, peak memory,
quarantine, geoclient benchmarks (#4960, #4964)
- Security: least-privilege GITHUB_TOKEN, SHA-pinned actions (#4985,
#4986, #4957)
- Visualization: Live Data Schematic gallery how-to (#4952), non-numeric
--color/--size error (#4959), height reporting since 28.0.0
- get --output overwrite permissions (#4948); geocode cache-info entry
counts (#4966, #4983); DCAT-US v3 validation note no longer "unreleased"
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
a446035
DCAT-US v3: document GSA's priority-dataset guidance (AI research, COVID-19)
0632a16
Recipe: DCAT-US v3 for data.gov: refresh outputs for 28.0.0
8fc9cf8
Release 28.0.0: replace the Unreleased (after 27.0.1) markers
9feb914
Metadata Profiling: geoconnex build needs feature_capable for qsv
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
16a9c12
profile docs: pre-release audit fixes (DCAT-US v3, Metadata Profiling)
- Default output is <file>.csv.metadata.json; fix jq examples.
- identifier fallback: input URL or seeded resource.url (absolute IRI).
- URL inputs always use the network (download, DCAT discovery, Last-Modified).
- Inferred tier now lists the estimates and LLM keywords/themes; measured vs
estimated wording matches the CHANGELOG.
- Output blocks: add formula_results, projection_warnings, dcat_discovered,
projection_provenance, llm_attribution.
- Shell-outs: add geocode (--locate) and describegpt (--infer).
- census --locate: 5% threshold applies to states too (heaviest always).
- MCP sampling claim made precise (--format json).
- LLM dictionary column annotations in provenance (#4981), override-only.
- Unreleased markers for #4973/#4974/#4975 and the who-wins section.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2831da8
DCAT-US v3 page: hashed duplicate-row count memory (#4975)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
5575e85
profile --locate census: region columns win over sampled coordinates (#4973)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
823e49d
profile: WKT POINT columns supply the coordinates (#4974)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bb32e11
profile --locate census: county Locations from a county column (#4973)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
6f9033e
DCAT-US v3 page: fix worked-example anchor
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
da6ecdb
DCAT-US v3 page: LA crime data (data.gov) worked example, release-build memory numbers
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
767a99b
chore: update serena config
[skip ci]
f621328
DCAT-US v3 page: record harvest.data.gov pass for current master output
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9682018
DCAT-US v3 page: current migration-step numbering, Distribution conformsTo, PR citations
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
186a2dc