First-use review: v0.12.0
Observed on 2026-09-12 UTC against the published usdata[pandas]==0.12.0
package and the maintained getting-started walkthrough.
The walkthrough, five newest-dataset live checks, and refreshed
climate-anomalies notebook passed.
No SDK or adapter changes were needed.
Method and results
The published package was installed into a new temporary virtual environment
outside the repository on Linux with Python 3.14.7. The imported package came
from that environment's site-packages, with no editable installation.
The walkthrough used fresh, isolated temporary cache directories; user caches
were untouched. The Python and manifest blocks were extracted from the maintained
guide. The release baseline was bc785de (v0.12.0).
| Step | Observed result |
|---|---|
| Search precipitation in Oklahoma | Completed successfully; supported station and other weather datasets returned. |
Inspect noaa:ghcn-daily |
Dataset information and accepted parameters displayed successfully. |
Fetch station USW00013967, May 6–7, 2024, PRCP,TMAX |
One 209-byte CSV downloaded. |
| Execute the guide's Python reading example | Two observation rows opened successfully. |
| Pull and verify the guide's manifest | Lockfile written; all cached assets matched. |
| Restore into a second empty cache | Pinned source downloaded again; restored bytes passed verification. |
| Read and verify the restored inputs with DNS and socket connections blocked | Both succeeded without network access. |
Newest-dataset checks
The checkout's bounded live tests were run with explicit live opt-in and pytest's temporary caches. All five passed, with no skips:
- Climate normals: airport station discovery; a daily February 27–March 1 window including leap day; all twelve monthly rows and exact-byte restoration.
- HURDAT2: download and parse the current whole Atlantic file; infer the local reader and verify Hurricane Ida's identity, coordinate signs, and peak wind.
The climate-anomalies notebook was executed in a fresh kernel and temporary working directory with isolated inputs. Its saved outputs were refreshed only after successful execution. It downloaded both sources, aligned twelve months, handled the documented temperature-unit exception, produced both plots, verified the lockfile, and confirmed a subsequent pull reused the verified cache.
The calculated summary reported 10 warmer-than-normal months and 942.4 mm of 2024 precipitation against 924.3 mm of summed monthly normals, a +18.1 mm difference. These are observations from one station and one execution, not a regional estimate, climate trend, or statistical significance test.
Limits
This review is a dated upstream-health and first-use snapshot. It covers one Linux/Python installation, the primary station walkthrough, climate normals, Atlantic HURDAT2, and the climate example. It does not replace the full scheduled live suite or the offline platform/dependency matrix. Upstream services can fail or revise files later. Keep manifests, lockfiles, and cached bytes together; checksums detect changes but do not archive unavailable upstream revisions.
The earlier v0.10.0 review retains its original evidence and example corrections.