github-search-cli 0.7.0 → 0.8.1

package/src/ranking.ts

@@ -1,57 +0,0 @@
-/**
- * Ranking layer: order normalized {@link Repo} results deterministically.
- *
- * All strategies are pure functions of the repo data plus a stable tie-breaker
- * (repo id) so that identical inputs always produce identical output. A future
- * AI reranker can be added as an additional strategy or as a post-step without
- * changing the rest of the pipeline.
- */
-import type { Repo, SortStrategy } from "./types.ts";
-
-/**
- * Rank repos according to the chosen strategy.
- * The sort is stable and deterministic: when two repos compare equal on the
- * primary key, we fall back to descending stars, then descending id.
- */
-export function rankRepos(repos: Repo[], strategy: SortStrategy): Repo[] {
-  const copy = [...repos];
-  switch (strategy) {
-    case "stars":
-      copy.sort(byKey((r) => r.stars));
-      break;
-    case "forks":
-      copy.sort(byKey((r) => r.forks));
-      break;
-    case "updated":
-      copy.sort(byKey((r) => Date.parse(r.updatedAt)));
-      break;
-    case "best-match":
-    default:
-      // GitHub already returns best-match order; preserve it unchanged.
-      // Re-sorting by score would be redundant for first-page results and
-      // misleading for paginated (page 2 repos would intermix with page 1).
-      // ponytail: if deterministic output is needed, sort by id:
-      // copy.sort((a, b) => a.id - b.id);
-      break;
-  }
-  return copy;
-}
-
-/** Comparator factory: descending by `key`, then descending stars, then
- *  ascending id (stable, deterministic tie-break). */
-function byKey(key: (r: Repo) => number): (a: Repo, b: Repo) => number {
-  return (a, b) => key(b) - key(a) || b.stars - a.stars || a.id - b.id;
-}
-
-/**
- * A simple, explainable composite score used for the (optional) "best" blend.
- * Not used by default but available for an AI/rerank layer later.
- * Range is roughly 0..1 after normalization by the caller.
- */
-export function compositeScore(r: Repo): number {
-  // Log-scaled stars + recency bonus. Deterministic and transparent.
-  const starScore = Math.log10(r.stars + 1);
-  const recencyMs = Date.now() - Date.parse(r.pushedAt);
-  const recencyScore = Math.max(0, 1 - recencyMs / (1000 * 60 * 60 * 24 * 365));
-  return starScore * 0.7 + recencyScore * 0.3;
-}