Skip to content

Copies of the model in other repositories

The BER feature model is maintained here. Two other repositories carry a copy or a re-implementation of it, and each can drift from this one. This page says which version each follows, how it differs, and how to bring it back in line. Checked 2026-10-02.

bridge-central-schema

Where https://github.com/doe-bridge/bridge-central-schema/pull/5, Add the BER feature model as an imported module (draft), files src/ber_central_schema/schema/ber_feature_model.yaml and attributes.yaml
Follows This repository's v0.1.0 release (commit 4aaca94), as stated in the copy's own notes. PR 5 head checked: 5e2a4b2.
Who updates it Whoever opens the pull request that changes the model here also opens or updates the matching bridge-central-schema pull request, and each links the other (see the pull request template).

Differences between the PR 5 copy at 5e2a4b2 and v0.1.0, from diff on 2026-10-02:

difference why follow-up
Slot name is renamed display_name, with alias: name The central root schema has its own name slot, and LinkML refuses two imported slots with one name. Data keys stay name. A modeling decision for the maintainers of both schemas; see overlaps and conflicts.
Prefix EDAM.DATA becomes EDAM, and enum meanings become EDAM:data_NNNN The LinkML Python generator can't handle the dotted prefix (https://github.com/linkml/linkml/issues/3458). The IRIs are the same. https://github.com/turbomam/feature-table-corpus/pull/163 (open, not merged, as of 2026-10-02) renames the prefix here to EDAM_DATA instead, because this repository's linter flags EDAM as non-canonical, and adds a test that imports the model under a stub root schema. If both merge as they stand, the two copies spell the prefix differently for the same IRIs.
tree_root: true removed from Dataset The central root schema chooses its own tree root. None needed.
Evidence and decision-history notes removed or shortened, and the descriptions of Feature and feature_id shortened The central copy keeps definitions only; the evidence stays here. None needed, but a resync must keep these edits.

Bringing the copy back in line

  1. Tag a release here.
  2. In a bridge-central-schema branch, copy model/schema/ber_feature_model.yaml and model/schema/attributes.yaml from the tag.
  3. Reapply the edits in the table above, and update the copy's notes to the new tag.
  4. Run that repository's tests, then diff against the tag and check that only the listed differences remain.
  5. Link the two pull requests to each other.

nmdc-lakehouse

Where https://github.com/microbiomedata/nmdc-lakehouse, src/nmdc_lakehouse/feature_convert.py, documented in its docs/nmdc_feature_tables.md
Follows No pinned version. It re-implements the model's rules without importing this repository. https://github.com/microbiomedata/nmdc-lakehouse/issues/382 asks for the README to say which version it follows.
Who updates it A model change here that affects NMDC files needs a matching pull request there, as https://github.com/microbiomedata/nmdc-lakehouse/pull/374 and https://github.com/microbiomedata/nmdc-lakehouse/pull/377 were.

Its column differences are listed under nmdc-lakehouse in the overlaps page. https://github.com/microbiomedata/nmdc-lakehouse/issues/381 asks for a CI test there that validates its output with the installable validator, pinned to v0.1.0, so drift fails there instead of in BERDL.