Release Comparison

What changed since the last release?

A gate report answers “is this release ready?” It does not answer “what got worse since 0.9?” LocalizeShip 1.0.0 has no dedicated comparison window. It gives you deterministic JSON reports, which are straightforward to diff.

Release gates report

The honest scope

Capability	Status in 1.0.0
Compare two workspaces side by side in the Editor	Not shipped
Compare two JSON gate reports	Possible with your own script
Compare two HTML reports visually	Possible, but noisy
Diff the workspace asset in version control	Works (plain serialized asset)
Time-series dashboard	Not shipped

Do not describe this page as a product feature. It is a workflow built on exported artifacts.

1. Export both reports

Run Release Board → Release Gates → Export JSON for each release. The filename includes the release version:

LocalizeShip_Report_v0.9.0.json
LocalizeShip_Report_v1.0.0.json

Keep a copy in a release-reports/ folder or as a CI artifact. The timestamp inside the report is UTC ticks, so two reports from the same version are distinguishable.

2. Compare with PowerShell

$old = Get-Content .\LocalizeShip_Report_v0.9.0.json | ConvertFrom-Json
$new = Get-Content .\LocalizeShip_Report_v1.0.0.json | ConvertFrom-Json

"Verdict: {0} -> {1}" -f $old.OverallStatus, $new.OverallStatus

foreach ($check in $new.Checks) {
    $prev = $old.Checks | Where-Object { $_.CheckName -eq $check.CheckName }
    "{0}: {1} -> {2} ({3} violations)" -f `
        $check.CheckName, $prev.Passed, $check.Passed, $check.Violations.Count
}

$oldViolations = $old.Checks.Violations
$newViolations = $new.Checks.Violations

"New violations:"
$newViolations | Where-Object { $_ -notin $oldViolations } | Sort-Object -Unique

"Resolved violations:"
$oldViolations | Where-Object { $_ -notin $newViolations } | Sort-Object -Unique

OverallStatus is an integer:

Value	Status
`0`	Passed
`1`	Failed
`2`	IncompleteCoverage

Because violation strings contain entry IDs and locales, set comparison is meaningful as long as the message format has not changed between versions.

3. Compare the human-readable reports

Open both HTML files and compare the four cards:

Translations Approval Matrix — did missing or stale counts move?
Visual QA Matrix & Geometry — did unvisited tiles shrink? Did error findings appear?
Voice Coverage & Revisions — are the same lines stale, or new ones?
Subtitle Timing & Reading Speed — did CPS violations appear after a text change?

The HTML contains details lines with the counters, so you can read the scale without opening the JSON.

4. Diff the workspace asset

The workspace is a serialized Unity asset. If it lives in version control, git diff shows exactly what changed between releases:

git diff v0.9.0 v1.0.0 -- Assets/Decnet/LocalizeShip/Samples/RPGSample/RPGSampleWorkspace.asset

Expect large diffs. The asset contains arrays of approvals, findings and revision history. The useful signal is usually:

a new SourceRevision with a higher revision number;
approval state transitions (Stale, NeedsReview);
new or resolved VisualFinding entries;
new VoiceEntry statuses.

A YAML-aware diff or a small editor script that prints counts per state is easier to read than the raw diff.

5. What a regression looks like

Change	Signal
Source edited after translation	`Stale` count up, voice `StaleSourceRevision` up
New UI screen without a capture	`unvisitedTiles` up, verdict Possibly `IncompleteCoverage`
Translator introduced a bad tag	Visual or text rule issue, not gate — gate does not check tags
New locale added	Missing translations jump, verdict `IncompleteCoverage`
Subtitle text lengthened	CPS violations appear
Approval record lost	Missing count up

Notice that a regression in text rules or tags will not appear in the gate report. Compare those with your own validation step.

6. Keep the comparison honest

Compare reports produced by the same gate version. A change in check logic can look like a regression.
Compare the same locale set. Adding a locale changes every count.
Do not compare a full report with a partial one produced after editing the workspace asset by hand.
If the verdict improved, confirm that coverage grew rather than the policy being relaxed.

A release-to-release checklist

Export the gate report for the candidate.
Diff it against the previous release report.
List new violations by category and owner.
Check that resolved violations were fixed, not merely suppressed or deleted.
Archive both reports with the build.
Record the verdict change in the release notes.