PolyWeather/frontend/lib/model-summary.ts

import type { ScanOpportunityRow } from "@/lib/dashboard-types";
import {
  REGIONS,
  getCityRegion,
} from "@/components/dashboard/scan-terminal/continent-grouping";

export const MODEL_SUMMARY_MODEL_COLUMNS = [
  { key: "ECMWF", label: "ECMWF" },
  { key: "ECMWF AIFS", label: "ECMWF AIFS" },
  { key: "GFS", label: "GFS" },
  { key: "ICON", label: "ICON" },
  { key: "ICON-EU", label: "ICON-EU" },
  { key: "GEM", label: "GEM" },
  { key: "GDPS", label: "GDPS" },
  { key: "JMA", label: "JMA" },
  { key: "AROME HD", label: "AROME HD" },
  { key: "HRRR", label: "HRRR" },
  { key: "NAM", label: "NAM" },
  { key: "WeatherNext 2", label: "WeatherNext 2" },
] as const;

export type ModelSummaryColumnKey = (typeof MODEL_SUMMARY_MODEL_COLUMNS)[number]["key"];

export type ModelSummaryProbabilityBucket = {
  key: string;
  label: string;
  value: number;
  lower: number;
  upper: number;
  probability: number;
};

export type ModelSummaryMarketMatch = {
  key: string;
  label: string;
  modelProbability: number | null;
  marketUrl: string | null;
};

export type ModelSummaryRow = {
  cityKey: string;
  cityName: string;
  regionLabel: string;
  regionLabelZh: string;
  regionSort: number;
  tempSymbol: string;
  localTime: string;
  timezoneOffsetSeconds: number | null;
  debPrediction: number | null;
  models: Record<ModelSummaryColumnKey, number | null>;
  modelMedian: number | null;
  modelSpread: number | null;
  probabilityBuckets: ModelSummaryProbabilityBucket[];
  probabilityBucketMap: Record<string, ModelSummaryProbabilityBucket>;
  gaussianMu: number | null;
  probabilityEngine: string | null;
  topProbabilityBucketKey: string | null;
  marketMatches: ModelSummaryMarketMatch[];
  searchText: string;
};

export type ModelSummaryFilters = {
  query: string;
  debOnly: boolean;
  wideSpreadOnly: boolean;
};

const WIDE_SPREAD_THRESHOLD = 2;

function finiteNumber(value: unknown): number | null {
  if (value == null) return null;
  if (typeof value === "string" && value.trim() === "") return null;
  const numericValue = Number(value);
  return Number.isFinite(numericValue) ? numericValue : null;
}

function roundToOneDecimal(value: number) {
  return Math.round(value * 10) / 10;
}

function median(values: number[]) {
  if (!values.length) return null;
  const sorted = [...values].sort((a, b) => a - b);
  const mid = Math.floor(sorted.length / 2);
  if (sorted.length % 2 === 1) return roundToOneDecimal(sorted[mid]);
  return roundToOneDecimal((sorted[mid - 1] + sorted[mid]) / 2);
}

function spread(values: number[]) {
  if (!values.length) return null;
  return roundToOneDecimal(Math.max(...values) - Math.min(...values));
}

function probabilityFromBucket(bucket: Record<string, unknown>) {
  const raw = finiteNumber(bucket.probability ?? bucket.model_probability);
  if (raw == null) return null;
  return raw > 1 ? raw / 100 : raw;
}

function formatBucketBound(value: number) {
  return Number(value.toFixed(1)).toString();
}

function probabilityBucketKey(lower: number, upper: number, unit: string) {
  return `${formatBucketBound(lower)}-${formatBucketBound(upper)}${unit || "°C"}`;
}

function probabilityBucketLabel(lower: number, upper: number, unit: string) {
  return probabilityBucketKey(lower, upper, unit);
}

function isFahrenheitUnit(unit: string) {
  return unit.toUpperCase().includes("F");
}

function marketOptionBucketForValue(value: number, unit: string) {
  const settledValue = Math.round(value);
  if (isFahrenheitUnit(unit)) {
    const lowerValue = settledValue % 2 === 0 ? settledValue : settledValue - 1;
    const upperValue = lowerValue + 1;
    return {
      key: `${lowerValue}-${upperValue}${unit || "°F"}`,
      label: `${lowerValue}-${upperValue}${unit || "°F"}`,
      lower: lowerValue - 0.5,
      upper: upperValue + 0.5,
    };
  }
  return {
    key: `${settledValue}${unit || "°C"}`,
    label: `${settledValue}${unit || "°C"}`,
    lower: settledValue - 0.5,
    upper: settledValue + 0.5,
  };
}

function roundProbability(value: number) {
  return Math.round(value * 1000) / 1000;
}

function sourceMarketBuckets(row: ScanOpportunityRow) {
  const marketRow = row as ScanOpportunityRow & {
    all_buckets?: Array<Record<string, unknown>> | null;
    top_buckets?: Array<Record<string, unknown>> | null;
  };
  return (
    Array.isArray(marketRow.all_buckets) && marketRow.all_buckets.length
      ? marketRow.all_buckets
      : Array.isArray(marketRow.top_buckets)
        ? marketRow.top_buckets
        : []
  ) as Array<Record<string, unknown>>;
}

function parseMarketOptionBucket(bucket: Record<string, unknown>, unit: string) {
  const rawLabel = String(bucket.label || bucket.bucket || bucket.range || "").trim();
  const labelNumbers = rawLabel.match(/-?\d+(?:\.\d+)?/g)?.map(Number) || [];
  const lower = finiteNumber(bucket.lower);
  const upper = finiteNumber(bucket.upper);
  let lowerValue: number | null = null;
  let upperValue: number | null = null;

  if (/below/i.test(rawLabel) && labelNumbers.length) {
    upperValue = Math.round(labelNumbers[0]);
  } else if (/higher/i.test(rawLabel) && labelNumbers.length) {
    lowerValue = Math.round(labelNumbers[0]);
  } else if (labelNumbers.length >= 2) {
    lowerValue = Math.round(labelNumbers[0]);
    upperValue = Math.round(labelNumbers[1]);
  } else if (labelNumbers.length === 1) {
    lowerValue = Math.round(labelNumbers[0]);
    upperValue = Math.round(labelNumbers[0]);
  } else if (lower != null && upper != null && upper > lower) {
    lowerValue = Math.ceil(lower);
    upperValue = Math.ceil(upper) - 1;
  } else {
    const value = finiteNumber(bucket.value ?? bucket.temp ?? bucket.temperature);
    if (value == null) return null;
    const option = marketOptionBucketForValue(value, unit);
    lowerValue = Math.ceil(option.lower);
    upperValue = Math.ceil(option.upper) - 1;
  }

  const finiteLower = lowerValue ?? Number.NEGATIVE_INFINITY;
  const finiteUpper = upperValue ?? Number.POSITIVE_INFINITY;
  if (finiteUpper < finiteLower) return null;

  let label = rawLabel;
  if (!label || /\.5\b/.test(label)) {
    const representative =
      Number.isFinite(finiteLower) && Number.isFinite(finiteUpper)
        ? (finiteLower + finiteUpper) / 2
        : Number.isFinite(finiteLower)
          ? finiteLower
          : finiteUpper;
    label = marketOptionBucketForValue(representative, unit).label;
  }

  return {
    key: label,
    label,
    lowerValue: finiteLower,
    upperValue: finiteUpper,
    sortValue: Number.isFinite(finiteLower) ? finiteLower : finiteUpper,
  };
}

function buildProbabilityBuckets(row: ScanOpportunityRow): ModelSummaryProbabilityBucket[] {
  const rawBuckets = (
    Array.isArray(row.distribution_full) && row.distribution_full.length
      ? row.distribution_full
      : Array.isArray(row.distribution_preview)
        ? row.distribution_preview
        : []
  ) as Array<Record<string, unknown>>;
  const unit = row.temp_symbol || "°C";
  const rawProbabilityPoints = rawBuckets
    .map((bucket) => {
      const value = finiteNumber(bucket.value ?? bucket.temp ?? bucket.temperature);
      const probability = probabilityFromBucket(bucket);
      if (value == null || probability == null || probability <= 0) return null;
      return { value, settledValue: Math.round(value), probability };
    })
    .filter((point): point is { value: number; settledValue: number; probability: number } => point !== null);

  const marketBuckets = sourceMarketBuckets(row)
    .map((bucket) => parseMarketOptionBucket(bucket, unit))
    .filter((bucket): bucket is NonNullable<typeof bucket> => bucket !== null);

  if (marketBuckets.length && rawProbabilityPoints.length) {
    const fromMarketBuckets = marketBuckets
      .map((bucket) => {
        const matchingPoints = rawProbabilityPoints.filter(
          (point) =>
            point.settledValue >= bucket.lowerValue &&
            point.settledValue <= bucket.upperValue,
        );
        const probability = matchingPoints.reduce((sum, point) => sum + point.probability, 0);
        if (probability <= 0) return null;
        const weightedValue = matchingPoints.reduce(
          (sum, point) => sum + point.value * point.probability,
          0,
        );
        return {
          key: bucket.key,
          label: bucket.label,
          value: roundToOneDecimal(weightedValue / probability),
          lower: bucket.sortValue,
          upper: bucket.sortValue,
          probability: roundProbability(probability),
        };
      })
      .filter((bucket): bucket is ModelSummaryProbabilityBucket => bucket !== null)
      .sort((a, b) => a.lower - b.lower || a.upper - b.upper);

    if (fromMarketBuckets.length) return fromMarketBuckets;
  }

  const grouped = new Map<
    string,
    {
      label: string;
      lower: number;
      upper: number;
      probability: number;
      weightedValue: number;
    }
  >();

  rawProbabilityPoints.forEach(({ value, probability }) => {
    const option = marketOptionBucketForValue(value, unit);
    const existing = grouped.get(option.key);
    if (existing) {
      existing.probability += probability;
      existing.weightedValue += value * probability;
      return;
    }
    grouped.set(option.key, {
      label: option.label,
      lower: option.lower,
      upper: option.upper,
      probability,
      weightedValue: value * probability,
    });
  });

  return [...grouped.entries()]
    .map(([key, bucket]) => ({
      key,
      label: bucket.label,
      value:
        bucket.probability > 0
          ? roundToOneDecimal(bucket.weightedValue / bucket.probability)
          : roundToOneDecimal((bucket.lower + bucket.upper) / 2),
      lower: bucket.lower,
      upper: bucket.upper,
      probability: roundProbability(bucket.probability),
    }))
    .sort((a, b) => a.lower - b.lower || a.upper - b.upper);
}

function weightedProbabilityMu(buckets: ModelSummaryProbabilityBucket[]) {
  const totalProbability = buckets.reduce((sum, bucket) => sum + bucket.probability, 0);
  if (totalProbability <= 0) return null;
  const weightedValue = buckets.reduce(
    (sum, bucket) => sum + bucket.value * bucket.probability,
    0,
  );
  return roundToOneDecimal(weightedValue / totalProbability);
}

function normalizeProbability(value: unknown) {
  const numericValue = finiteNumber(value);
  if (numericValue == null) return null;
  return numericValue > 1 ? numericValue / 100 : numericValue;
}

function marketBucketLabel(bucket: Record<string, unknown>, tempSymbol: string) {
  const textLabel = String(bucket.label || bucket.bucket || bucket.range || "").trim();
  if (textLabel) return textLabel;
  const lower = finiteNumber(bucket.lower);
  const upper = finiteNumber(bucket.upper);
  if (lower != null && upper != null && upper > lower) {
    return probabilityBucketLabel(lower, upper, String(bucket.unit || tempSymbol || "°C"));
  }
  const value = finiteNumber(bucket.value ?? bucket.temp ?? bucket.temperature);
  return value == null ? "—" : `${formatBucketBound(value)}${bucket.unit || tempSymbol || "°C"}`;
}

function buildMarketMatches(row: ScanOpportunityRow): ModelSummaryMarketMatch[] {
  const marketRow = row as ScanOpportunityRow & {
    all_buckets?: Array<Record<string, unknown>> | null;
    top_buckets?: Array<Record<string, unknown>> | null;
  };
  const sourceBuckets = (
    Array.isArray(marketRow.all_buckets) && marketRow.all_buckets.length
      ? marketRow.all_buckets
      : Array.isArray(marketRow.top_buckets)
        ? marketRow.top_buckets
        : []
  ) as Array<Record<string, unknown>>;
  const tempSymbol = row.temp_symbol || "°C";

  return sourceBuckets
    .map((bucket, index) => {
      const label = marketBucketLabel(bucket, tempSymbol);
      const modelProbability = normalizeProbability(bucket.model_probability ?? bucket.probability);
      return {
        key: `${label}-${index}`,
        label,
        modelProbability,
        marketUrl: typeof bucket.market_url === "string" ? bucket.market_url : null,
      };
    })
    .sort((a, b) => {
      return (b.modelProbability ?? -1) - (a.modelProbability ?? -1);
    });
}

function normalizeCityKey(row: ScanOpportunityRow, index: number) {
  const rawKey = row.city || row.city_display_name || row.display_name || `row-${index}`;
  return String(rawKey).trim().toLowerCase();
}

function normalizeLocalTime(value: unknown) {
  const text = String(value || "").trim();
  if (!text) return "";
  const match = text.match(/(\d{1,2}):(\d{2})/);
  if (!match) return text;
  return `${match[1].padStart(2, "0")}:${match[2]}`;
}

function resolveRegion(row: ScanOpportunityRow, isEn: boolean) {
  const configuredRegionKey = getCityRegion(row);
  const configuredRegion = configuredRegionKey
    ? REGIONS.find((region) => region.key === configuredRegionKey)
    : null;
  if (configuredRegion) {
    return {
      label: isEn ? configuredRegion.labelEn : configuredRegion.labelZh,
      labelEn: configuredRegion.labelEn,
      labelZh: configuredRegion.labelZh,
      sort: configuredRegion.sort,
    };
  }

  const labelEn = row.trading_region_label || row.trading_region_label_zh || "—";
  const labelZh = row.trading_region_label_zh || row.trading_region_label || "—";
  return {
    label: isEn ? labelEn : labelZh,
    labelEn,
    labelZh,
    sort: finiteNumber(row.trading_region_sort) ?? 999,
  };
}

export function formatModelSummaryTemp(value: number | null | undefined, symbol = "°C") {
  const numericValue = finiteNumber(value);
  if (numericValue == null) return "—";
  return `${numericValue.toFixed(1)}${symbol || "°C"}`;
}

export function formatModelSummaryProbability(value: number | null | undefined) {
  const numericValue = finiteNumber(value);
  if (numericValue == null) return "—";
  return `${Math.round((numericValue > 1 ? numericValue / 100 : numericValue) * 100)}%`;
}

export function formatModelSummaryLocalTime(
  row: Pick<ModelSummaryRow, "localTime" | "timezoneOffsetSeconds">,
  nowMs: number | null | undefined = Date.now(),
) {
  const offsetSeconds = finiteNumber(row.timezoneOffsetSeconds);
  const timestampMs = finiteNumber(nowMs);
  if (offsetSeconds == null || timestampMs == null) return row.localTime || "—";
  const localDate = new Date(timestampMs + offsetSeconds * 1000);
  const hours = String(localDate.getUTCHours()).padStart(2, "0");
  const minutes = String(localDate.getUTCMinutes()).padStart(2, "0");
  return `${hours}:${minutes}`;
}

export function buildModelSummaryRows(
  rows: ScanOpportunityRow[],
  isEn: boolean,
): ModelSummaryRow[] {
  const byCity = new Map<string, ModelSummaryRow>();

  rows.forEach((row, index) => {
    const cityKey = normalizeCityKey(row, index);
    if (byCity.has(cityKey)) return;

    const cityName = row.city_display_name || row.display_name || row.city || "—";
    const region = resolveRegion(row, isEn);
    const rawModelSources = row.model_cluster_sources || {};
    const models = MODEL_SUMMARY_MODEL_COLUMNS.reduce(
      (acc, column) => {
        acc[column.key] = finiteNumber(rawModelSources[column.key]);
        return acc;
      },
      {} as Record<ModelSummaryColumnKey, number | null>,
    );
    const modelValues = MODEL_SUMMARY_MODEL_COLUMNS.map((column) => models[column.key]).filter(
      (value): value is number => value != null,
    );
    const modelSearchText = MODEL_SUMMARY_MODEL_COLUMNS.filter(
      (column) => models[column.key] != null,
    )
      .map((column) => column.label)
      .join(" ");
    const probabilityBuckets = buildProbabilityBuckets(row);
    const probabilityBucketMap = Object.fromEntries(
      probabilityBuckets.map((bucket) => [bucket.key, bucket]),
    );
    const topProbabilityBucket =
      probabilityBuckets.length > 0
        ? probabilityBuckets.reduce((best, bucket) =>
            bucket.probability > best.probability ? bucket : best,
          )
        : null;
    const probabilitySearchText = probabilityBuckets
      .map((bucket) => `${bucket.label} ${formatModelSummaryProbability(bucket.probability)}`)
      .join(" ");
    const marketMatches = buildMarketMatches(row);
    const marketSearchText = marketMatches
      .map(
        (match) =>
          `${match.label} ${formatModelSummaryProbability(match.modelProbability)}`,
      )
      .join(" ");

    byCity.set(cityKey, {
      cityKey,
      cityName,
      regionLabel: region.label,
      regionLabelZh: region.labelZh,
      regionSort: region.sort,
      tempSymbol: row.temp_symbol || "°C",
      localTime: normalizeLocalTime(row.local_time),
      timezoneOffsetSeconds: finiteNumber(row.tz_offset_seconds),
      debPrediction: finiteNumber(row.deb_prediction),
      models,
      modelMedian: median(modelValues),
      modelSpread: spread(modelValues),
      probabilityBuckets,
      probabilityBucketMap,
      gaussianMu: weightedProbabilityMu(probabilityBuckets),
      probabilityEngine: row.probability_engine || (probabilityBuckets.length ? "legacy" : null),
      topProbabilityBucketKey: topProbabilityBucket?.key || null,
      marketMatches,
      searchText:
        `${cityName} ${row.city || ""} ${region.labelEn} ${region.labelZh} ${modelSearchText} ${probabilitySearchText} ${marketSearchText}`.toLowerCase(),
    });
  });

  return [...byCity.values()].sort((a, b) => {
    if (a.regionSort !== b.regionSort) return a.regionSort - b.regionSort;
    return a.cityName.localeCompare(b.cityName, isEn ? "en" : "zh-CN", {
      sensitivity: "base",
    });
  });
}

export function filterModelSummaryRows(
  rows: ModelSummaryRow[],
  filters: ModelSummaryFilters,
): ModelSummaryRow[] {
  const query = filters.query.trim().toLowerCase();

  return rows.filter((row) => {
    if (query && !row.searchText.includes(query)) return false;
    if (filters.debOnly && row.debPrediction == null) return false;
    if (
      filters.wideSpreadOnly &&
      (row.modelSpread == null || row.modelSpread < WIDE_SPREAD_THRESHOLD)
    ) {
      return false;
    }
    return true;
  });
}

export function hasModelSummaryForecastData(rows: ModelSummaryRow[]) {
  return rows.some((row) => {
    if (row.debPrediction != null) return true;
    return MODEL_SUMMARY_MODEL_COLUMNS.some((column) => row.models[column.key] != null);
  });
}