Skip to content

Avoid excessive (and incorrect) dct:modified changes #345

Description

@dalito

In voc4cat PR processing, voc4cat transform --prov-from-git updates dct:modified dates for ALL concepts based on git history, even when only a few concepts actually changed. This happens because:

  1. The CI round-trip (xlsx -> ttl -> split -> merge) rewrites ALL concept files via voc4cat-merge (git merge-file --theirs), stripping per-concept dct:modified dates (since xlsx doesn't preserve them)
  2. --prov-from-git then re-adds dates from git history for ALL files
  3. Each CI commit becomes the "most recent commit" for all files, causing a cascading date shift on every CI run

Result: The PR nfdi4cat/voc4cat#255 had 540 out of 566 changed files with only a dct:modified date change -- pure noise that obscures the 26 real changes.

Metadata

Metadata

Assignees

Labels

bugSomething is not working as expected.enhancementRequest for changing or improving existing features

Type

No type

Projects

Status
Completed

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions