Repository navigation
docs(research): Ascent logical tuples via DR indices, search masks and BYODS - #1706
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configuration
Warning Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
AdaWorldAPI
changed the title
AdaWorldAPI
changed the title
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e5c681cb97
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
… cites Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G22yT6htkcdyXsihxxXdrg
2 participants
Purpose
Documentation-only research plan answering the operator's Ascent counterproposal: a logical tuple is an addressed projection, not a materialized value object. Ascent's logical tuple construction may lower into native DR folds or into a prebound index that already represents the join, without importing its tuple-at-a-time executor.
Two files
.arxiv/2023_oopsla_byods_ascent_custom_relations_dr_excerpt.md: primary source OOPSLA 2023 BYODS,(D,inj,γ), semi-naive,eqreland index-sharing, plus architectural interpretation and falsifiers..claude/plans/ascent-byods-dr-fold-polyfill-v1.md: A0–A6 cold compiler plan, counterexamples, snapshot/index-seal gates, proof/effect separation and matched-workload economics.Three distinct logical tuple cases
$hashA[$i].Value:$hashB[$i].Value, read borrowed lane values ati; no tuple allocation, search, join or mask intersection.ibyLeftMask & RightMaskon a common attested coordinate; fold over original lanes. The AND plane counts matches of coordinate slots, not arbitrary projected result tuples.ADMISSION=B.Limits: non-functional
ymultiplicity may need factorized fan-out (Σ_y countLeft[y] × countRight[y]for bag counts);DISTINCT(x,z)requires de-dup acrossyor a named refusal. Index preparation/attestation and authorization remain separately costed.Existing ruling and cold-cost status
The 2026-10-21 Ascent BYODS SKIP as an unmodified hot tuple executor remains correct. It does not falsify a compiler rewrite of Ascent's logical join before default iterator/head-emission codegen. A1 audits that distinction; A4 compares direct/selection/index-as-join lowering against existing RelAlg/DR.
CLAUDE.mdalready includes the 2026-10-22 cold exception: cold proof/compiler/differential-oracle costs are reported in µs, labeled reported, not gates. No P0 policy change requested or made. Matched cold+warm and sharing/nonsharing economics remain a separate falsifiable question from hot-fold ns/µs admission.Ownership, safety and status
EXO-79/84/85/86/87/102/46/73 and egglog/math-foundry owners unchanged. No extra engine/AST, P0 modification, dependencies, runtime code, ABI or CI. Existing witnessed-coordinate and same-table-admission constraints remain.
OPEN/DRAFT, not merged. Cross-session handoff and corrections on EXO-50.