Score every date and write the feature-level interop CSVs - #296
Conversation
9533159 to
9d9eb5c
Compare
9d9eb5c to
f040675
Compare
0f5cc78 to
338c5f1
Compare
DanielRyanSmith
left a comment
DanielRyanSmith
left a comment
There was a problem hiding this comment.
A published row
is not rescored, so --force is the only way to correct one.
This from the PR description is incorrect, right? The scores are regenerated after every script run, and also there's no --force flag here
| csvRecord.push(csvField(scored.versions[i])); | ||
| csvRecord.push(formatScore(scored.scores.get(products[i]))); |
There was a problem hiding this comment.
browser scores are looked up by name (scored.scores.get(products[i])), but versions are indexed by array position (scored.versions[i]). I don't know if wpt.fyi guarantees that runs is sorted alphabetically by browser name, so version strings could end up under the wrong browser columns.
Maybe key versions by browser name rather than array index:
const versions = new Map(runs.map(run => [run.browser_name, run.browser_version]));
and look it up with scored.versions.get(products[i])
There was a problem hiding this comment.
That makes sense, thank you for the review!
| * date, appended to the scores already published on gh-pages, and a | ||
| * for one date. |
There was a problem hiding this comment.
| * date, appended to the scores already published on gh-pages, and a | |
| * for one date. | |
| * date, appended to the scores already published on gh-pages, and a | |
| * feature breakdown for one date. |
There was a problem hiding this comment.
Thank you for catching this!
338c5f1 to
e269f99
Compare
Each date is scored against the manifest from that date's own wpt revision, so a run is never measured against a catalogue it did not run from. A build writes one aggregate row per date plus that date's per-feature breakdown, rescoring every date as browser-specific-failures.js does.
e269f99 to
a57c4dc
Compare
2 participants
Each date is scored against the manifest from that date's own wpt revision, so
a run is never measured against a catalogue it did not run from. A build writes
one aggregate row per date plus that date's per-feature breakdown, rescoring
every date as browser-specific-failures.js does.