UFAL/Health report bugs and improvements

[milanmajchrak]

The original issue:
ufal#1328

Another idea: Add a health report check in the end of the email/report which does not have any diff info.

All related issues and PR:
ufal#1328
ufal#1334
#1254

[Paurikova2]

@milanmajchrak In that PR there are only improvements from Milan Kuchtiak. We should think about what else from that issue we want and create a new issue and PR.

[Kasinhou]

Summary: health-report + report-diff (state per PR dataquest-dev/DSpace#1254)

Context

This PR addresses issues reported in:

• ufal/clarin-dspace#1328 (health-report + report-diff UX/parameters)
• ufal/clarin-dspace#1334 (report-diff not working correctly with different sets of checks)

---

1) health-report

Issues (before PR)

• The legacy command healthcheck was still shown in the CLI, even though health-report had been introduced.
• -c could not reliably handle selection of multiple checks.
• Weak input validation (-c, -f) and ambiguous error behavior.
• Inconsistent switches:
  • -i instead of the standard -h
  • -o instead of the proposed -r/--report

Changes in PR

• Removed legacy healthcheck command from dspace/config/launcher.xml.
• -i replaced by -h/--help.
• -o replaced by -r/--report.
• Improved validation:
  • -c must be an integer within the range of available checks.
  • -f must be a positive integer.
• Tests in HealthReportIT added/extended (help, validation, report file output, check selection).

What may still be worth fine-tuning

• Consistently revisit the historical report type label (healthcheck) and email template naming (kept for now for backward compatibility).
• Verify UX of error messages (to be fully unambiguous for both CLI and REST execution).

---

2) report-diff

Issues (before PR)

• Key bug from #1334: comparison relied on fixed indices (/checks/0, /checks/1, ...), so reports with different check selections produced:
  • massive null values,
  • misleading "Changed" entries virtually everywhere.
• --check was semantically unclear and in the old implementation tied to fragile logic via args.
• Output did not communicate which checks could not actually be compared.

Changes in PR

• -i replaced by -h/--help.
• Comparison is now based on the intersection of checks by name (check.name), not by index position.
• Report normalization introduced before diffing:
  • only shared checks are compared.
• Skipped Checks section added:
  • explicitly lists checks that appeared in only one of the reports.
• report-diff-fields.json reworked from fixed indices to name-based selectors:
  • e.g. /checks/[name=Item summary]/...
• Resolver added for [name=...] selector paths.
• API layers for findByLastModifiedAndCheckType (DAO/service) removed, as they were tied to the old fragile logic.
• Tests in ReportDiffIT extended (help, skipped checks, intersection behavior, formatting).

What remains open (from issue/comments)

• Selecting reports by ID / simple date (YYYY-MM-DD) instead of a strict timestamp.
• Better UX around --dates (clearer "from/to" selection).
• Possible rename of report-diff -> health-report-diff (suggested in comments, not part of this PR).
• Confirm/validate email sending behavior across all code paths (an issue was reported in comments).

---

3) Recommended next steps

• Add user-friendly report selection (ID or YYYY-MM-DD) for --from/--to.
• Improve --dates output (e.g. also print the internal report ID).
• Decide on naming (report-diff vs health-report-diff) based on UI/UX.
• Add integration tests for edge-case combinations of check selection + --check filter.
• After PR merge, add a short "migration note" to the docs (changed arguments -h, -r, intersection behavior).

---

[Kasinhou]

@milanmajchrak what do you think about this (formatting, order,...)