ecs: a lookup through an index shouldn't conflict with a writer whose rows are exclusive to the index's
メンテナーはふだん 1 日以内に返信
まだ誰も着手していません。
評価
- 難易度
- 4/5
- 見積もり時間
- 3〜5日
- 初心者へのやさしさ
- 22/100
- issue の種類
- 機能追加
- 明瞭さ
- おおむね明確
- 活発さ
- 活発
- 技術スタック
- typescript
- 領域
- data
調査の方向性
Start from the scheduler's access-conflict check that #246 extended, the code that emits the message Systems ... both access ... and one writes it. The request is to treat a lookup through an index as carrying that index's components.with and not, so a writer whose not excludes the index's with no longer conflicts on the key. Read the reply on #246 first, since it sets which lookup kinds may skip the check. Done means a schedule with showHover and extrudeSelection builds without an ordering, and the non-unique and up cases keep their current behaviour or are explicitly excluded. The issue leaves soundness for non-unique and up lookups open, so settle that before writing code.
索引モデルが issue の本文から書いたものです。
説明
Summary
A lookup through an index conflicts on its key with any system that writes that key, even when the writer's rows can never be in the index, because the index's own with/not is disjoint from the writer's. #246 made that disjointness count for two systems' rows (one requires a component the other's not lists); an index a system looks through carries the same information statically (its components.with / not) and isn't used.
Where this came from (mecs)
The editor draws a hover mark: an entity with [hovered, cell, face, faceColor] that follows the pointer. showHover moves it with commands.add(row.entity, cell, ...), so the mark's face is drawn one sync point later than the pointer. I'd like it to write row[cell] = ... directly so it moves in the frame the pointer does.
cell is also the key of voxelAt, the unique index of voxels (Index.create({ key: [cell], unique: true, components: { with: Voxel } })) that extrudeSelection and selectFace look up by. The mark is not a voxel (Voxel includes voxel; the mark has no voxel), so nothing showHover writes can ever appear in voxelAt. With 0.3.0 I added not: [voxel] to showHover's rows, which resolved its conflict with deriveCells (rows of with: Voxel), but the build still fails:
Systems "extrudeSelection" and "showHover" both access "cell" and one writes it; order them with before or after.
extrudeSelection only looks cell up through voxelAt, whose with is Voxel, and showHover's rows have not: [voxel]: the two sets are exclusive by construction.
What I need
Treat a lookup through an index as if its rows were the index's components (with, not) for the exclusivity check #246 added: a lookup of an index whose with includes a component the writer's not lists doesn't conflict with that writer on the index's key. (The issue's reply said "lookups, up, index and order keys still conflict"; I'm asking for the index case to follow the same rule when the index declares exclusivity, since a schedule is built without a world and the index's declaration is all it has.)
If it can't be sound for non-unique or up lookups, restricting it to indexes is enough: it is the case that hurts.
Cost of the workaround
Ordering the two systems (after) adds a tier boundary and a sync point to a frame that is otherwise idle (#245 measured about 0.1 µs each, now less), and the order is arbitrary between two systems that have nothing to do with each other.
- 主要言語
- TypeScript
- スター
- 12
- フォーク
- 8
- 平均マージ
- 11時間 2分
- マージ済み PR(30日)
- 23
環境構築
- Dockerfile・Docker Compose ファイルなし
- プルリクエストのテンプレートなし
- コントリビューションガイドを読む
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
adobe/data のほかの issue
-
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
メンテナーはふだん 1 日以内に返信
-
ecs: let kernels issue commands through helpers (component lists, symbol-keyed values, void helpers)オープン
難易度 5/5 1週間以上 初心者へのやさしさ 15/100
メンテナーはふだん 1 日以内に返信
-
難易度 4/5 3〜5日 初心者へのやさしさ 30/100
メンテナーはふだん 1 日以内に返信
-
ecs: an architecture standard (layout, naming, exports, checker, diagrams) for the new ECS samples, and how to distribute it対応中かも このイシューにリンクされたプルリクエストがオープン中、またはマージ済みです。 オープン
難易度 5/5 1週間以上 初心者へのやさしさ 8/100
メンテナーはふだん 1 日以内に返信
-
難易度 5/5 1週間以上 初心者へのやさしさ 15/100
メンテナーはふだん 1 日以内に返信
似ている issue
-
[Bug] Pod detail shows CPU usage ten times too high, and throws on decimal memory suffixes対応中かも @RaphaelFakhri が今日担当しました。 オープン
難易度 2/5 1〜3時間 初心者へのやさしさ 84/100
メンテナーはふだん 11 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 84/100
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 76/100
remotion-dev/remotion#12957 ·
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 66/100
tyrasd/overpass-turbo#871 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 67/100
メンテナーはふだん 1 日以内に返信