mirror of
https://github.com/samsonjs/spirit-tracker.git
synced 2026-03-25 09:25:51 +00:00
UX Improvements
This commit is contained in:
Brennan Wilkes (Text Groove)
parent
37796cbcbc
commit
b2a4afa890
1 changed files with 394 additions and 8 deletions
|
|
@ -2,12 +2,16 @@
|
|||
"use strict";
|
||||
|
||||
/*
|
||||
Print local link URLs for SKUs with largest rank discrepancy between AB and BC lists.
|
||||
Print local link URLs for SKUs with largest rank discrepancy between AB and BC lists,
|
||||
but ONLY when there exists another *different* listing (not in same linked group)
|
||||
with a reasonably high similarity score by name.
|
||||
|
||||
Usage:
|
||||
node scripts/rank_discrepency_links.js \
|
||||
--ab reports/common_listings_ab_top1000.json \
|
||||
--bc reports/common_listings_bc_top1000.json \
|
||||
--meta viz/data/sku_meta.json \
|
||||
--min-score 0.75 \
|
||||
--top 50 \
|
||||
--base "http://127.0.0.1:8080/#/link/?left="
|
||||
|
||||
|
|
@ -18,6 +22,8 @@
|
|||
const fs = require("fs");
|
||||
const path = require("path");
|
||||
|
||||
/* ---------------- IO ---------------- */
|
||||
|
||||
function readJson(p) {
|
||||
return JSON.parse(fs.readFileSync(p, "utf8"));
|
||||
}
|
||||
|
|
@ -26,17 +32,21 @@ function parseArgs(argv) {
|
|||
const out = {
|
||||
ab: "reports/common_listings_ab_top1000.json",
|
||||
bc: "reports/common_listings_bc_top1000.json",
|
||||
meta: "", // optional sku_meta containing {links:[{fromSku,toSku}], ignores:...}
|
||||
top: 50,
|
||||
minDiscrep: 1,
|
||||
includeMissing: false,
|
||||
base: "http://127.0.0.1:8080/#/link/?left=",
|
||||
minScore: 0.75, // similarity threshold for "reasonably high"
|
||||
};
|
||||
for (let i = 0; i < argv.length; i++) {
|
||||
const a = argv[i];
|
||||
if (a === "--ab" && argv[i + 1]) out.ab = argv[++i];
|
||||
else if (a === "--bc" && argv[i + 1]) out.bc = argv[++i];
|
||||
else if (a === "--meta" && argv[i + 1]) out.meta = argv[++i];
|
||||
else if (a === "--top" && argv[i + 1]) out.top = Number(argv[++i]) || out.top;
|
||||
else if (a === "--min" && argv[i + 1]) out.minDiscrep = Number(argv[++i]) || out.minDiscrep;
|
||||
else if (a === "--min-score" && argv[i + 1]) out.minScore = Number(argv[++i]) || out.minScore;
|
||||
else if (a === "--include-missing") out.includeMissing = true;
|
||||
else if (a === "--base" && argv[i + 1]) out.base = String(argv[++i] || out.base);
|
||||
}
|
||||
|
|
@ -55,19 +65,375 @@ function buildRankMap(payload) {
|
|||
return map;
|
||||
}
|
||||
|
||||
function pickName(row) {
|
||||
if (!row) return "";
|
||||
return String(row.name || row.title || row.productName || row.displayName || "");
|
||||
}
|
||||
|
||||
/* ---------------- sku_meta grouping (optional) ---------------- */
|
||||
|
||||
function normalizeImplicitSkuKey(k) {
|
||||
const s = String(k || "").trim();
|
||||
const m = s.match(/^id:(\d{1,6})$/i);
|
||||
if (m) return String(m[1]).padStart(6, "0");
|
||||
return s;
|
||||
}
|
||||
|
||||
class DSU {
|
||||
constructor() {
|
||||
this.parent = new Map();
|
||||
this.rank = new Map();
|
||||
}
|
||||
_add(x) {
|
||||
if (!this.parent.has(x)) {
|
||||
this.parent.set(x, x);
|
||||
this.rank.set(x, 0);
|
||||
}
|
||||
}
|
||||
find(x) {
|
||||
x = String(x || "").trim();
|
||||
if (!x) return "";
|
||||
this._add(x);
|
||||
let p = this.parent.get(x);
|
||||
if (p !== x) {
|
||||
p = this.find(p);
|
||||
this.parent.set(x, p);
|
||||
}
|
||||
return p;
|
||||
}
|
||||
union(a, b) {
|
||||
a = String(a || "").trim();
|
||||
b = String(b || "").trim();
|
||||
if (!a || !b || a === b) return;
|
||||
const ra = this.find(a);
|
||||
const rb = this.find(b);
|
||||
if (!ra || !rb || ra === rb) return;
|
||||
|
||||
const rka = this.rank.get(ra) || 0;
|
||||
const rkb = this.rank.get(rb) || 0;
|
||||
|
||||
if (rka < rkb) this.parent.set(ra, rb);
|
||||
else if (rkb < rka) this.parent.set(rb, ra);
|
||||
else {
|
||||
this.parent.set(rb, ra);
|
||||
this.rank.set(ra, rka + 1);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Choose a stable representative (good enough for filtering “same-linked”)
|
||||
function compareSku(a, b) {
|
||||
a = String(a || "").trim();
|
||||
b = String(b || "").trim();
|
||||
if (a === b) return 0;
|
||||
|
||||
const aUnknown = a.startsWith("u:");
|
||||
const bUnknown = b.startsWith("u:");
|
||||
if (aUnknown !== bUnknown) return aUnknown ? 1 : -1;
|
||||
|
||||
const aNum = /^\d+$/.test(a);
|
||||
const bNum = /^\d+$/.test(b);
|
||||
if (aNum && bNum) {
|
||||
const na = Number(a), nb = Number(b);
|
||||
if (Number.isFinite(na) && Number.isFinite(nb) && na !== nb) return na < nb ? -1 : 1;
|
||||
}
|
||||
return a < b ? -1 : 1;
|
||||
}
|
||||
|
||||
function buildCanonicalSkuFnFromMeta(meta) {
|
||||
const links = Array.isArray(meta?.links) ? meta.links : [];
|
||||
if (!links.length) return (sku) => normalizeImplicitSkuKey(sku);
|
||||
|
||||
const dsu = new DSU();
|
||||
const all = new Set();
|
||||
|
||||
for (const x of links) {
|
||||
const a = normalizeImplicitSkuKey(x?.fromSku);
|
||||
const b = normalizeImplicitSkuKey(x?.toSku);
|
||||
if (!a || !b || a === b) continue;
|
||||
all.add(a);
|
||||
all.add(b);
|
||||
dsu.union(a, b);
|
||||
}
|
||||
|
||||
// root -> members
|
||||
const groupsByRoot = new Map();
|
||||
for (const s of all) {
|
||||
const r = dsu.find(s);
|
||||
if (!r) continue;
|
||||
let set = groupsByRoot.get(r);
|
||||
if (!set) groupsByRoot.set(r, (set = new Set()));
|
||||
set.add(s);
|
||||
}
|
||||
|
||||
// root -> representative
|
||||
const repByRoot = new Map();
|
||||
for (const [root, members] of groupsByRoot.entries()) {
|
||||
const arr = Array.from(members);
|
||||
arr.sort(compareSku);
|
||||
repByRoot.set(root, arr[0] || root);
|
||||
}
|
||||
|
||||
// sku -> rep
|
||||
const canonBySku = new Map();
|
||||
for (const [root, members] of groupsByRoot.entries()) {
|
||||
const rep = repByRoot.get(root) || root;
|
||||
for (const s of members) canonBySku.set(s, rep);
|
||||
canonBySku.set(rep, rep);
|
||||
}
|
||||
|
||||
return (sku) => {
|
||||
const s = normalizeImplicitSkuKey(sku);
|
||||
return canonBySku.get(s) || s;
|
||||
};
|
||||
}
|
||||
|
||||
/* ---------------- similarity (copied from viz/app) ---------------- */
|
||||
|
||||
// Normalize for search: lowercase, punctuation -> space, collapse spaces
|
||||
function normSearchText(s) {
|
||||
return String(s ?? "")
|
||||
.toLowerCase()
|
||||
.replace(/[^a-z0-9]+/g, " ")
|
||||
.replace(/\s+/g, " ")
|
||||
.trim();
|
||||
}
|
||||
|
||||
function tokenizeQuery(q) {
|
||||
const n = normSearchText(q);
|
||||
return n ? n.split(" ").filter(Boolean) : [];
|
||||
}
|
||||
|
||||
const SIM_STOP_TOKENS = new Set([
|
||||
"the","a","an","and","of","to","in","for","with",
|
||||
"year","years","yr","yrs","old",
|
||||
"whisky","whiskey","scotch","single","malt","cask","finish","edition","release","batch","strength","abv","proof",
|
||||
"anniversary",
|
||||
]);
|
||||
|
||||
const ORDINAL_RE = /^(\d+)(st|nd|rd|th)$/i;
|
||||
|
||||
function numKey(t) {
|
||||
const s = String(t || "").trim().toLowerCase();
|
||||
if (!s) return "";
|
||||
if (/^\d+$/.test(s)) return s;
|
||||
const m = s.match(ORDINAL_RE);
|
||||
return m ? m[1] : "";
|
||||
}
|
||||
|
||||
function extractAgeFromText(normName) {
|
||||
const s = String(normName || "");
|
||||
if (!s) return "";
|
||||
|
||||
const m = s.match(/\b(?:aged\s*)?(\d{1,2})\s*(?:yr|yrs|year|years)\b/i);
|
||||
if (m && m[1]) return String(parseInt(m[1], 10));
|
||||
|
||||
const m2 = s.match(/\b(\d{1,2})\s*yo\b/i);
|
||||
if (m2 && m2[1]) return String(parseInt(m2[1], 10));
|
||||
|
||||
return "";
|
||||
}
|
||||
|
||||
function filterSimTokens(tokens) {
|
||||
const out = [];
|
||||
const seen = new Set();
|
||||
|
||||
const SIM_EQUIV = new Map([
|
||||
["years", "yr"],
|
||||
["year", "yr"],
|
||||
["yrs", "yr"],
|
||||
["yr", "yr"],
|
||||
["whiskey", "whisky"],
|
||||
["whisky", "whisky"],
|
||||
["bourbon", "bourbon"],
|
||||
]);
|
||||
|
||||
const VOL_UNIT = new Set(["ml","l","cl","oz","liter","liters","litre","litres"]);
|
||||
const VOL_INLINE_RE = /^\d+(?:\.\d+)?(?:ml|l|cl|oz)$/i;
|
||||
const PCT_INLINE_RE = /^\d+(?:\.\d+)?%$/;
|
||||
|
||||
const arr = Array.isArray(tokens) ? tokens : [];
|
||||
|
||||
for (let i = 0; i < arr.length; i++) {
|
||||
const raw = arr[i];
|
||||
let t = String(raw || "").trim().toLowerCase();
|
||||
if (!t) continue;
|
||||
|
||||
if (!/[a-z0-9]/i.test(t)) continue;
|
||||
|
||||
if (VOL_INLINE_RE.test(t)) continue;
|
||||
if (PCT_INLINE_RE.test(t)) continue;
|
||||
|
||||
t = SIM_EQUIV.get(t) || t;
|
||||
|
||||
const nk = numKey(t);
|
||||
if (nk) t = nk;
|
||||
|
||||
if (VOL_UNIT.has(t) || t === "abv" || t === "proof") continue;
|
||||
|
||||
if (/^\d+(?:\.\d+)?$/.test(t)) {
|
||||
const next = String(arr[i + 1] || "").trim().toLowerCase();
|
||||
const nextNorm = SIM_EQUIV.get(next) || next;
|
||||
if (VOL_UNIT.has(nextNorm)) {
|
||||
i++;
|
||||
continue;
|
||||
}
|
||||
}
|
||||
|
||||
if (!numKey(t) && SIM_STOP_TOKENS.has(t)) continue;
|
||||
|
||||
if (seen.has(t)) continue;
|
||||
seen.add(t);
|
||||
out.push(t);
|
||||
}
|
||||
|
||||
return out;
|
||||
}
|
||||
|
||||
function tokenContainmentScore(aTokens, bTokens) {
|
||||
const A = filterSimTokens(aTokens || []);
|
||||
const B = filterSimTokens(bTokens || []);
|
||||
if (!A.length || !B.length) return 0;
|
||||
|
||||
const aSet = new Set(A);
|
||||
const bSet = new Set(B);
|
||||
|
||||
const small = aSet.size <= bSet.size ? aSet : bSet;
|
||||
const big = aSet.size <= bSet.size ? bSet : aSet;
|
||||
|
||||
let hit = 0;
|
||||
for (const t of small) if (big.has(t)) hit++;
|
||||
|
||||
const recall = hit / Math.max(1, small.size);
|
||||
const precision = hit / Math.max(1, big.size);
|
||||
const f1 = (2 * precision * recall) / Math.max(1e-9, precision + recall);
|
||||
|
||||
return f1;
|
||||
}
|
||||
|
||||
function levenshtein(a, b) {
|
||||
a = String(a || "");
|
||||
b = String(b || "");
|
||||
const n = a.length, m = b.length;
|
||||
if (!n) return m;
|
||||
if (!m) return n;
|
||||
|
||||
const dp = new Array(m + 1);
|
||||
for (let j = 0; j <= m; j++) dp[j] = j;
|
||||
|
||||
for (let i = 1; i <= n; i++) {
|
||||
let prev = dp[0];
|
||||
dp[0] = i;
|
||||
const ca = a.charCodeAt(i - 1);
|
||||
for (let j = 1; j <= m; j++) {
|
||||
const tmp = dp[j];
|
||||
const cost = ca === b.charCodeAt(j - 1) ? 0 : 1;
|
||||
dp[j] = Math.min(dp[j] + 1, dp[j - 1] + 1, prev + cost);
|
||||
prev = tmp;
|
||||
}
|
||||
}
|
||||
return dp[m];
|
||||
}
|
||||
|
||||
function numberMismatchPenalty(aTokens, bTokens) {
|
||||
const aNums = new Set((aTokens || []).map(numKey).filter(Boolean));
|
||||
const bNums = new Set((bTokens || []).map(numKey).filter(Boolean));
|
||||
if (!aNums.size || !bNums.size) return 1.0;
|
||||
for (const n of aNums) if (bNums.has(n)) return 1.0;
|
||||
return 0.28;
|
||||
}
|
||||
|
||||
// Same structure/weights as viz/app/linker/similarity.js
|
||||
function similarityScore(aName, bName) {
|
||||
const a = normSearchText(aName);
|
||||
const b = normSearchText(bName);
|
||||
if (!a || !b) return 0;
|
||||
|
||||
const aAge = extractAgeFromText(a);
|
||||
const bAge = extractAgeFromText(b);
|
||||
const ageBoth = !!(aAge && bAge);
|
||||
const ageMatch = ageBoth && aAge === bAge;
|
||||
const ageMismatch = ageBoth && aAge !== bAge;
|
||||
|
||||
const aToksRaw = tokenizeQuery(a);
|
||||
const bToksRaw = tokenizeQuery(b);
|
||||
|
||||
const aToks = filterSimTokens(aToksRaw);
|
||||
const bToks = filterSimTokens(bToksRaw);
|
||||
if (!aToks.length || !bToks.length) return 0;
|
||||
|
||||
const contain = tokenContainmentScore(aToksRaw, bToksRaw);
|
||||
|
||||
const aFirst = aToks[0] || "";
|
||||
const bFirst = bToks[0] || "";
|
||||
const firstMatch = aFirst && bFirst && aFirst === bFirst ? 1 : 0;
|
||||
|
||||
const A = new Set(aToks.slice(1));
|
||||
const B = new Set(bToks.slice(1));
|
||||
let inter = 0;
|
||||
for (const w of A) if (B.has(w)) inter++;
|
||||
const denom = Math.max(1, Math.max(A.size, B.size));
|
||||
const overlapTail = inter / denom;
|
||||
|
||||
const d = levenshtein(a, b);
|
||||
const maxLen = Math.max(1, Math.max(a.length, b.length));
|
||||
const levSim = 1 - d / maxLen;
|
||||
|
||||
let gate = firstMatch ? 1.0 : Math.min(0.80, 0.06 + 0.95 * contain);
|
||||
|
||||
const smallN = Math.min(aToks.length, bToks.length);
|
||||
if (!firstMatch && smallN <= 3 && contain < 0.78) gate *= 0.18;
|
||||
|
||||
const numGate = numberMismatchPenalty(aToks, bToks);
|
||||
|
||||
let s =
|
||||
numGate *
|
||||
(firstMatch * 3.0 +
|
||||
overlapTail * 2.2 * gate +
|
||||
levSim * (firstMatch ? 1.0 : (0.10 + 0.70 * contain)));
|
||||
|
||||
if (ageMatch) s *= 2.2;
|
||||
else if (ageMismatch) s *= 0.18;
|
||||
|
||||
s *= 1 + 0.9 * contain;
|
||||
|
||||
return s;
|
||||
}
|
||||
|
||||
/* ---------------- main logic ---------------- */
|
||||
|
||||
function main() {
|
||||
const args = parseArgs(process.argv.slice(2));
|
||||
const repoRoot = process.cwd();
|
||||
|
||||
const abPath = path.isAbsolute(args.ab) ? args.ab : path.join(repoRoot, args.ab);
|
||||
const bcPath = path.isAbsolute(args.bc) ? args.bc : path.join(repoRoot, args.bc);
|
||||
const metaPath = args.meta
|
||||
? (path.isAbsolute(args.meta) ? args.meta : path.join(repoRoot, args.meta))
|
||||
: "";
|
||||
|
||||
const ab = readJson(abPath);
|
||||
const bc = readJson(bcPath);
|
||||
|
||||
const canonicalSku = metaPath
|
||||
? buildCanonicalSkuFnFromMeta(readJson(metaPath))
|
||||
: (sku) => normalizeImplicitSkuKey(sku);
|
||||
|
||||
const abMap = buildRankMap(ab);
|
||||
const bcMap = buildRankMap(bc);
|
||||
|
||||
// Build a flat pool of candidates from AB+BC (unique by canonSku)
|
||||
const rowBySku = new Map();
|
||||
for (const m of [abMap, bcMap]) {
|
||||
for (const [canonSku, v] of m.entries()) {
|
||||
if (!rowBySku.has(canonSku)) rowBySku.set(canonSku, v.row);
|
||||
}
|
||||
}
|
||||
|
||||
const allSkus = Array.from(rowBySku.keys());
|
||||
const allNames = new Map();
|
||||
for (const sku of allSkus) allNames.set(sku, pickName(rowBySku.get(sku)));
|
||||
|
||||
const keys = new Set([...abMap.keys(), ...bcMap.keys()]);
|
||||
const diffs = [];
|
||||
|
||||
|
|
@ -88,7 +454,6 @@ function main() {
|
|||
diffs.push({
|
||||
canonSku,
|
||||
discrep,
|
||||
// tie-breakers
|
||||
sumRank: (rankAB ?? 1e9) + (rankBC ?? 1e9),
|
||||
});
|
||||
}
|
||||
|
|
@ -99,13 +464,34 @@ function main() {
|
|||
return String(x.canonSku).localeCompare(String(y.canonSku));
|
||||
});
|
||||
|
||||
const top = diffs.slice(0, args.top);
|
||||
// Keep only discrepancies that have a high-scoring "other" candidate not in same linked group
|
||||
const filtered = [];
|
||||
for (const d of diffs) {
|
||||
const skuA = String(d.canonSku);
|
||||
const nameA = allNames.get(skuA) || pickName(abMap.get(skuA)?.row) || pickName(bcMap.get(skuA)?.row);
|
||||
if (!nameA) continue;
|
||||
|
||||
for (const d of top) {
|
||||
// examples:
|
||||
// 884096 -> left=884096
|
||||
// id:1049355 -> left=id%3A1049355
|
||||
// u:bb504a62 -> left=u%3Abb504a62
|
||||
const groupA = canonicalSku(skuA);
|
||||
|
||||
let best = 0;
|
||||
for (const skuB of allSkus) {
|
||||
if (skuB === skuA) continue;
|
||||
|
||||
// not same-linked group
|
||||
if (canonicalSku(skuB) === groupA) continue;
|
||||
|
||||
const nameB = allNames.get(skuB) || "";
|
||||
if (!nameB) continue;
|
||||
|
||||
const s = similarityScore(nameA, nameB);
|
||||
if (s > best) best = s;
|
||||
}
|
||||
|
||||
if (best >= args.minScore) filtered.push(d);
|
||||
if (filtered.length >= args.top) break;
|
||||
}
|
||||
|
||||
for (const d of filtered) {
|
||||
console.log(args.base + encodeURIComponent(d.canonSku));
|
||||
}
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in a new issue