Repository navigation
cleanup: revert lesson workarounds once minigraf#287, #297, and #298 are fixed #40
Description
Background
Three bugs in `@minigraf/browser` 1.2.0 required workarounds in the lesson `starterCode` and e2e test helper (commit `4d9730f`):
-
bug: EavtKey and AevtKey missing value bytes causes silent data loss when batching multi-valued facts minigraf#287 — batch transact with same entity+attribute loses all but one fact.
- Workaround in `l1-s3`: split `[:alice :friend :bob]` and `[:alice :friend :carol]` into separate `(transact ...)` calls.
-
bug: literal keyword bound in rule call resolves as entity UUID, silently mismatches keyword values in rule body minigraf#298 — literal keyword bound at rule call site resolves as entity UUID, silently mismatches `Value::Keyword` facts in value-position rule body patterns → empty results.
- Workaround in `o3-s2`, `o3-s3`: look up the target entity via a fact pattern (e.g., `[?alice :employee/title "CEO"]`) and pass the bound variable into the rule instead of a literal keyword.
-
bug: magic sets rewriting rejects intermediate variables when first rule arg is bound by literal keyword minigraf#297 — magic sets rewriting flags intermediate variable `?mid` as "Unbound variable in rule head" when the first rule arg is bound by a literal keyword, causing errors/panics in recursive rules that call other user-defined rules.
- Workaround in `sl3-s3`, `t3-s3`: same pattern-bind approach as above (e.g., `[?united :club/name "FC United"]`).
Also, `tests/e2e/steps.test.ts` had to be updated to use the renamed WASM file (`minigraf_wasm_bg.wasm` instead of `minigraf_bg.wasm`). This is a 1.2.0 packaging change and does not need reverting.
Action required
Once all three upstream bugs are fixed and a patch release ships:
- Revert `l1-s3` `starterCode` to use a single batch transact with all three facts (fixes #287).
- Revert `o3-s2` and `o3-s3` to use literal keyword bindings in rule calls (e.g., `(reports-to ?emp :alice)`) (fixes #298).
- Revert `sl3-s3` and `t3-s3` to use literal keyword bindings in rule calls (e.g., `(transfer-chain :united ?club)`) (fixes #297).
- Bump `@minigraf/browser` to the patched version and run CI to confirm.
Activity
[-]cleanup: revert lesson workarounds once minigraf#297 and minigraf#287 are fixed[/-][+]cleanup: revert lesson workarounds once minigraf#287, #297, and #298 are fixed[/+]chore: upgrade @minigraf/browser to 2.0.1 and revert #298 workaround (#…
adityamukho commentedon Sep 26, 2026
Reopening. #43 only did part of this issue.
Status after #43 (
@minigraf/browser2.0.1)(reports-to ?emp :alice)and(reports-to ?emp :bob)again.Unbound variable in rule head: ?midwhen a recursive rule calls another user-defined rule (transfer-link,same-line). project-minigraf/minigraf#409 fixes this and is merged. It is in the v2.0.2 milestone. The pattern-bind workaround stays until 2.0.2 is on npm.Step 4 (upgrade the dependency) is done. The app is on 2.0.1.
Left to do
@minigraf/browser2.0.2 ships, upgrade to it. Then revert sl3-s3 and t3-s3 to(transfer-chain :united ?club)and(reachable :north ?station), and run the e2e step tests to confirm.Edited: minigraf#297 is now fixed upstream by minigraf#409 (v2.0.2 milestone).
Generated by Claude Code
chore: upgrade @minigraf/browser to 2.0.2 and revert #297 workaround
chore: bump @minigraf/browser to 2.0.6
adityamukho commentedon Oct 14, 2026
Reopening. #44 auto-closed this, but it only covered the minigraf#297 part (sl3-s3, t3-s3 reverted; app now on
@minigraf/browser2.0.6).Left to do: minigraf#287 is still a known issue in 2.0.6. Its fix needs file format v8 and ships in minigraf v3.0.0. When v3.0.0 is on npm, upgrade and revert l1-s3 to one batch transact.
Generated by Claude Code
adityamukho commentedon Oct 24, 2026
Status 2026-10-24, with #47 and #48:
@minigraf/browserto 2.0.10 and reverts thesl3-s4workaround for minigraf#608 (fixed in v2.0.7). That workaround was added in fix: lesson starter code that gave wrong answers or errors #46, not here, so fix(lessons): revert the sl3-s4 workaround for minigraf#608; @minigraf/browser 2.0.10 #48 doesn't close any item of this issue.l1-s3item is still blocked. On 2.0.10, onetransactof[:alice :friend :bob] [:alice :friend :carol]still returns only:carol. That's minigraf#287 / minigraf#371, a known v2.x issue fixed only in v3.0.0.l1-s1andl1-s2into separate transacts for the same bug. When v3.0.0 is on npm, those can go back to one batch transact along withl1-s3.l1-s1/l1-s2/l1-s3, re-run the lesson audit.Generated by Claude Code