0075_fix_proposal_snapshot_commit_naming.py
file-level
1
files
1
commits
0
hotspots
0
🧊 dead
0
💥 blast risk
| 1 | """Fix musehub_proposals from/to_snapshot_id naming bug -- add from/to_commit_id |
| 2 | |
| 3 | musehub#144 review of proposal #9 traced a real naming bug back to #0049 |
| 4 | ("proposal_snapshot_anchors"): that migration's own docstring says it |
| 5 | "captures the HEAD commit ID of each branch" but named the columns |
| 6 | from_snapshot_id/to_snapshot_id. Every row ever written to these columns |
| 7 | holds a commit ID, never a Muse Snapshot (manifest) ID -- confirmed via |
| 8 | musehub_proposals.create_proposal, which has always populated them from |
| 9 | branch.head_commit_id. |
| 10 | |
| 11 | Fix: add from_commit_id/to_commit_id (String(128), nullable), backfill them |
| 12 | from the existing from_snapshot_id/to_snapshot_id values (which already hold |
| 13 | commit IDs for every existing row), then correct from_snapshot_id/ |
| 14 | to_snapshot_id in place to hold the real snapshot_id looked up from each |
| 15 | referenced commit. Existing rows whose commit_id no longer resolves in |
| 16 | musehub_commits (e.g. GC'd/unreachable history) are left with a null |
| 17 | snapshot_id rather than a guessed value. |
| 18 | |
| 19 | Revision ID: 0075 |
| 20 | Revises: 0074 |
| 21 | """ |
| 22 | from __future__ import annotations |
| 23 | |
| 24 | from typing import Sequence, Union |
| 25 | |
| 26 | from alembic import op |
| 27 | import sqlalchemy as sa |
| 28 | |
| 29 | |
| 30 | # revision identifiers, used by Alembic. |
| 31 | revision: str = '0075' |
| 32 | down_revision: Union[str, None] = '0074' |
| 33 | branch_labels: Union[str, Sequence[str], None] = None |
| 34 | depends_on: Union[str, Sequence[str], None] = None |
| 35 | |
| 36 | |
| 37 | def upgrade() -> None: |
| 38 | op.add_column('musehub_proposals', sa.Column('from_commit_id', sa.String(length=128), nullable=True)) |
| 39 | op.add_column('musehub_proposals', sa.Column('to_commit_id', sa.String(length=128), nullable=True)) |
| 40 | |
| 41 | # Backfill the new, correctly-named columns from the existing (mislabeled) |
| 42 | # columns -- every existing row's from_snapshot_id/to_snapshot_id is |
| 43 | # already a commit ID. |
| 44 | op.execute( |
| 45 | "UPDATE musehub_proposals " |
| 46 | "SET from_commit_id = from_snapshot_id, to_commit_id = to_snapshot_id" |
| 47 | ) |
| 48 | |
| 49 | # Now correct from_snapshot_id/to_snapshot_id in place to hold the real |
| 50 | # snapshot_id of the commit they reference. Rows whose commit no longer |
| 51 | # resolves (unreachable/GC'd history) fall back to NULL rather than a |
| 52 | # stale commit ID masquerading as a snapshot ID. |
| 53 | op.execute( |
| 54 | "UPDATE musehub_proposals p " |
| 55 | "SET from_snapshot_id = c.snapshot_id " |
| 56 | "FROM musehub_commits c " |
| 57 | "WHERE c.commit_id = p.from_commit_id" |
| 58 | ) |
| 59 | op.execute( |
| 60 | "UPDATE musehub_proposals p " |
| 61 | "SET from_snapshot_id = NULL " |
| 62 | "WHERE p.from_commit_id IS NOT NULL " |
| 63 | "AND NOT EXISTS (SELECT 1 FROM musehub_commits c WHERE c.commit_id = p.from_commit_id)" |
| 64 | ) |
| 65 | op.execute( |
| 66 | "UPDATE musehub_proposals p " |
| 67 | "SET to_snapshot_id = c.snapshot_id " |
| 68 | "FROM musehub_commits c " |
| 69 | "WHERE c.commit_id = p.to_commit_id" |
| 70 | ) |
| 71 | op.execute( |
| 72 | "UPDATE musehub_proposals p " |
| 73 | "SET to_snapshot_id = NULL " |
| 74 | "WHERE p.to_commit_id IS NOT NULL " |
| 75 | "AND NOT EXISTS (SELECT 1 FROM musehub_commits c WHERE c.commit_id = p.to_commit_id)" |
| 76 | ) |
| 77 | |
| 78 | |
| 79 | def downgrade() -> None: |
| 80 | # Restore the pre-fix semantics: from/to_snapshot_id hold the commit ID. |
| 81 | op.execute( |
| 82 | "UPDATE musehub_proposals " |
| 83 | "SET from_snapshot_id = from_commit_id, to_snapshot_id = to_commit_id" |
| 84 | ) |
| 85 | op.drop_column('musehub_proposals', 'to_commit_id') |
| 86 | op.drop_column('musehub_proposals', 'from_commit_id') |