"use client";
import { useState, useEffect, useCallback } from "react";
import { motion } from "framer-motion";
import { Brain, CircleNotch as Loader2, ArrowSquareOut as ExternalLink, Dna, Stack as Layers, Target, DownloadSimple as Download } from '@phosphor-icons/react';
import { fadeUp } from "@/lib/animations";
import { useAuditTrail } from "@/hooks/useAuditTrail";
import { predictFunction, getFunctionStatus, type FunctionPredictionResult } from "@/lib/api";
import { BackButton, PageHeader, CriticalButton, FlatInput, ResultsReadyBanner } from "@/components/ui";
import { LearnPopover } from "@/components/LearnPopover";

const NAMESPACE_COLORS: Record<string, { text: string; bg: string; label: string }> = {
  MF: { text: "text-accent-cyan", bg: "bg-accent-cyan/10", label: "Molecular Function" },
  BP: { text: "text-accent-purple", bg: "bg-accent-purple/10", label: "Biological Process" },
  CC: { text: "text-accent-amber", bg: "bg-accent-amber/10", label: "Cellular Component" },
};

export default function FunctionPage() {
  useAuditTrail();

  const [pdbId, setPdbId] = useState("");
  const [jobId, setJobId] = useState<string | null>(null);
  const [status, setStatus] = useState<string>("");
  const [result, setResult] = useState<FunctionPredictionResult | null>(null);
  const [error, setError] = useState("");
  const [loading, setLoading] = useState(false);

  const poll = useCallback(async (id: string) => {
    try {
      const res = await getFunctionStatus(id);
      setStatus(res.status);
      if (res.status === "complete" && res.result) {
        setResult(res.result);
        setLoading(false);
      } else if (res.status === "failed") {
        setError(res.error || "Prediction failed");
        setLoading(false);
      }
    } catch {
      setLoading(false);
    }
  }, []);

  useEffect(() => {
    if (!jobId) return;
    const start = Date.now();
    const MAX_POLL_MS = 65 * 60 * 1000;
    const iv = setInterval(() => {
      if (Date.now() - start > MAX_POLL_MS) {
        setError("Prediction is still running on the server. Check back in a few minutes.");
        setLoading(false);
        clearInterval(iv);
        return;
      }
      poll(jobId);
    }, 2000);
    return () => clearInterval(iv);
  }, [jobId, poll]);

  useEffect(() => {
    const stored = sessionStorage.getItem('function_pdb_id');
    if (stored) {
      sessionStorage.removeItem('function_pdb_id');
      setPdbId(stored);
    }
  }, []);

  const handleSubmit = async () => {
    if (!pdbId.trim()) return;
    setLoading(true);
    setError("");
    setResult(null);
    try {
      const res = await predictFunction(pdbId.trim());
      setJobId(res.job_id);
      setStatus(res.status);
    } catch (e: any) {
      setError(typeof e?.response?.data?.detail === "string" ? e.response.data.detail : e?.response?.data?.detail?.message || e.message || "Submission failed");
      setLoading(false);
    }
  };

  const groupedTerms = result ? {
    MF: result.go_terms.filter(t => t.namespace === "MF").sort((a, b) => b.confidence - a.confidence),
    BP: result.go_terms.filter(t => t.namespace === "BP").sort((a, b) => b.confidence - a.confidence),
    CC: result.go_terms.filter(t => t.namespace === "CC").sort((a, b) => b.confidence - a.confidence),
  } : { MF: [], BP: [], CC: [] };

  const exportJson = () => {
    if (!result) return;
    const a = document.createElement("a");
    a.download = `${result.pdb_id}_function_prediction.json`;
    a.href = "data:application/json;charset=utf-8," + encodeURIComponent(JSON.stringify(result, null, 2));
    a.click();
  };

  return (
    <div className="max-w-4xl">
      <BackButton />

      <PageHeader title="Function Prediction" subtitle="Predict protein function from structure: GO terms and residue importance from a DeepFRI-inspired heuristic model." />

      <motion.div variants={fadeUp} initial={{ y: 24 }} animate="show">
        <div className="data-card p-5">
          <label className="block text-sm text-text-secondary mb-2">PDB ID</label>
          <div className="flex gap-2">
            <FlatInput value={pdbId} onChange={(e) => { setPdbId(e.target.value.toUpperCase()); setResult(null); setError(""); }}
              placeholder="e.g. 1TIM" maxLength={4}
              className="w-32 font-mono uppercase"
              onKeyDown={(e) => e.key === "Enter" && handleSubmit()} />
            <CriticalButton onClick={handleSubmit} disabled={loading || !pdbId.trim()}>
              {loading ? <Loader2 className="w-4 h-4 animate-spin" /> : <Brain className="w-4 h-4" />}
              {status === "running" || status === "queued" ? `Status: ${status}...` : "Predict Function"}
            </CriticalButton>
          </div>
        </div>
      </motion.div>

      {error && (
        <motion.div initial={{ opacity: 0 }} animate={{ opacity: 1 }} className="mt-4 glass-card p-4 border border-error/30">
          <p className="text-error text-sm">{error}</p>
        </motion.div>
      )}

      {result && (
        <motion.div id="function-results" initial={{ opacity: 0, y: 16 }} animate={{ opacity: 1, y: 0 }} className="mt-6 space-y-4">
          <ResultsReadyBanner
            title={`Prediction complete · ${result.pdb_id}`}
            subtitle={`${result.method.replace(/_/g, ' ')} · ${result.sequence_length} residues`}
          />
          <div className="flex flex-wrap items-center gap-2">
            <button onClick={exportJson}
              className="text-xs px-2.5 py-1 rounded bg-surface-1 border border-glass-border text-text-secondary hover:text-accent-cyan transition-colors flex items-center gap-1.5">
              <Download className="w-3.5 h-3.5" /> Export JSON
            </button>
            <p className="text-xs text-text-muted">
              A predictive heuristic model &mdash; this is <em>not</em> UniProt&apos;s curated function annotation (see the UniProt tool for experimentally documented function).
            </p>
          </div>
          {/* Header */}
          <div className="data-card p-4 flex flex-wrap items-center gap-4">
            <div>
              <p className="text-xs text-text-muted">PDB Entry</p>
              <a href={`https://www.rcsb.org/structure/${result.pdb_id}`} target="_blank" rel="noopener noreferrer"
                className="text-sm font-mono font-medium text-accent-cyan hover:underline flex items-center gap-1">
                {result.pdb_id} <ExternalLink className="w-3 h-3" />
              </a>
            </div>
            <div className="h-6 w-px bg-surface-3" />
            <div>
              <p className="text-xs text-text-muted">Sequence Length</p>
              <p className="text-sm font-semibold text-text-primary">{result.sequence_length} residues</p>
            </div>
            <div className="h-6 w-px bg-surface-3" />
            <div>
              <p className="text-xs text-text-muted">Prediction Method</p>
              <p className="text-sm font-medium text-text-primary">{result.method.replace(/_/g, " ")}</p>
            </div>
            <div className="h-6 w-px bg-surface-3" />
            <div>
              <p className="text-xs text-text-muted">GO Terms Predicted</p>
              <p className="text-sm font-semibold text-text-primary">{result.go_terms.length}</p>
            </div>
          </div>

          {/* GO Terms by Namespace */}
          {(["MF", "BP", "CC"] as const).map(ns => groupedTerms[ns].length > 0 && (
            <div key={ns} className="data-card p-5">
              <div className="flex items-center gap-2 mb-3">
                <Target className={`w-4 h-4 ${NAMESPACE_COLORS[ns].text}`} />
                <LearnPopover term={`GO ${NAMESPACE_COLORS[ns].label}`} topic="function"
                  explanation="A Gene Ontology category. Molecular Function = what the protein does at a molecular level. Biological Process = the larger cellular pathway it participates in. Cellular Component = where it acts.">
                  <h3 className="text-sm font-semibold text-text-primary">{NAMESPACE_COLORS[ns].label}</h3>
                </LearnPopover>
                <span className="text-xs text-text-muted">({groupedTerms[ns].length} terms)</span>
              </div>
              <div className="space-y-2">
                {groupedTerms[ns].map((go, i) => (
                  <div key={i} className={`flex items-center gap-3 ${NAMESPACE_COLORS[ns].bg} rounded-lg p-3`}>
                    <div className="flex-1 min-w-0">
                      <div className="flex items-center gap-2 flex-wrap">
                        <span className="text-sm font-medium text-text-primary">{go.name}</span>
                        <span className="text-xs font-mono text-text-muted">{go.go_id}</span>
                      </div>
                      <div className="flex items-center gap-2 mt-1">
                        <div className="w-24 h-1.5 bg-surface-3 rounded-full">
                          <div className={`h-full rounded-full ${NAMESPACE_COLORS[ns].text.replace("text-", "bg-")}`}
                            style={{ width: `${go.confidence * 100}%` }} />
                        </div>
                        <span className="text-xs font-mono text-text-muted">{(go.confidence * 100).toFixed(1)}%</span>
                        <span className={`text-xs ${go.confidence > 0.8 ? "text-good" : go.confidence > 0.6 ? "text-warn" : "text-text-muted"}`}>
                          {go.confidence > 0.8 ? "High" : go.confidence > 0.6 ? "Medium" : "Low"} confidence
                        </span>
                      </div>
                    </div>
                  </div>
                ))}
              </div>
            </div>
          ))}

          {/* EC Numbers */}
          {result.ec_numbers && result.ec_numbers.length > 0 && (
            <div className="data-card p-5">
              <h3 className="text-sm font-semibold text-text-primary mb-3 flex items-center gap-2">
                <Layers className="w-4 h-4 text-accent-amber" /> EC Number Predictions
              </h3>
              <div className="space-y-2">
                {result.ec_numbers.map((ec, i) => (
                  <div key={i} className="flex items-center justify-between bg-surface-1 rounded-lg p-3">
                    <div>
                      <span className="text-sm font-mono font-medium text-text-primary">{ec.number}</span>
                    </div>
                    <div className="flex items-center gap-2">
                      <div className="w-16 h-1.5 bg-surface-3 rounded-full">
                        <div className="h-full rounded-full bg-accent-amber" style={{ width: `${ec.confidence * 100}%` }} />
                      </div>
                      <span className="text-xs font-mono text-text-muted">{(ec.confidence * 100).toFixed(1)}%</span>
                    </div>
                  </div>
                ))}
              </div>
            </div>
          ) || (
            <div className="data-card p-4">
              <p className="text-xs text-text-muted">
                <strong className="text-text-primary">EC numbers:</strong> not predicted by the heuristic model — enzyme classification is out of scope for this approximation.
              </p>
            </div>
          )}

          {/* Saliency Map */}
          {result.saliency.length > 0 && (
            <div className="data-card p-5">
              <h3 className="text-sm font-semibold text-text-primary mb-1 flex items-center gap-2">
                <Dna className="w-4 h-4 text-good" /> Residue Importance (Saliency Map)
              </h3>
              <p className="text-xs text-text-muted mb-3">Per-residue contribution to function prediction. Higher = more important. Charged/polar residues on the surface typically dominate.</p>
              <div className="bg-surface-1 rounded-lg p-3 overflow-x-auto">
                <div className="flex gap-px min-w-max">
                  {result.saliency.map((score, i) => {
                    const r = Math.round(59 + (220 - 59) * score);
                    const g = Math.round(130 + (50 - 130) * score);
                    const b = Math.round(246 + (80 - 246) * score);
                    return (
                      <div key={i} className="w-1.5 h-10 rounded-sm cursor-pointer hover:ring-1 hover:ring-white/50 transition-all"
                        style={{ backgroundColor: `rgb(${r},${g},${b})` }}
                        title={`Residue ${i + 1}: ${score.toFixed(3)}`} />
                    );
                  })}
                </div>
              </div>
              <div className="flex justify-between text-xs text-text-muted mt-1">
                <span>N-terminus</span>
                <span>C-terminus</span>
              </div>
              <div className="flex items-center gap-4 mt-2">
                <div className="flex items-center gap-1">
                  <div className="w-3 h-3 rounded" style={{ backgroundColor: "rgb(59,130,246)" }} />
                  <span className="text-xs text-text-muted">Low importance</span>
                </div>
                <div className="flex items-center gap-1">
                  <div className="w-3 h-3 rounded" style={{ backgroundColor: "rgb(220,50,80)" }} />
                  <span className="text-xs text-text-muted">High importance</span>
                </div>
              </div>
            </div>
          )}

          {/* Sequence Composition Summary */}
          {result.composition && (
            <div className="data-card p-5">
              <h3 className="text-sm font-semibold text-text-primary mb-1 flex items-center gap-2">
                <Dna className="w-4 h-4 text-accent-cyan" /> Sequence Composition Analysis
              </h3>
              <p className="text-xs text-text-muted mb-3">Measured amino acid fractions from the actual sequence. Hydrophobic fraction and charge distribution drive the GO term assignment.</p>
              <div className="grid grid-cols-5 gap-3">
                {result.composition.aa.split("").map(aa => {
                  const frac = result.composition?.fractions[aa] ?? 0;
                  const width = Math.max(frac * 100, 1.5);
                  return (
                    <div key={aa} className="bg-surface-1 rounded-lg p-2">
                      <div className="text-sm font-mono font-semibold text-text-primary">{aa}</div>
                      <div className="w-full h-1.5 bg-surface-3 rounded-full mt-1">
                        <div className="h-full rounded-full bg-accent-cyan" style={{ width: `${width}%` }} />
                      </div>
                      <div className="text-xs text-text-muted mt-1">{(frac * 100).toFixed(1)}%</div>
                    </div>
                  );
                })}
              </div>
            </div>
          )}

          {/* Interpretation */}
          <div className="data-card p-5">
            <h3 className="text-sm font-semibold text-text-primary mb-3">Interpretation</h3>
            <div className="space-y-2 text-xs text-text-secondary">
              {result.go_terms.length > 0 && (
                <p><strong className="text-text-primary">Top Prediction:</strong> {result.go_terms.sort((a, b) => b.confidence - a.confidence)[0].name} ({result.go_terms.sort((a, b) => b.confidence - a.confidence)[0].namespace}) with {(result.go_terms.sort((a, b) => b.confidence - a.confidence)[0].confidence * 100).toFixed(1)}% confidence.</p>
              )}
              <p><strong className="text-text-primary">Methodology:</strong> {result.method === "heuristic_composition" ? "Predictions are based on amino acid composition patterns (hydrophobic fraction, charge distribution). For production use, deploy the full DeepFRI GCN model with pre-trained weights." : "Predicted using the full GCN model."}</p>
              {result.saliency.length > 0 && (() => {
                const maxIdx = result.saliency.indexOf(Math.max(...result.saliency));
                return <p><strong className="text-text-primary">Key Residue:</strong> Position {maxIdx + 1} shows highest importance (score: {result.saliency[maxIdx].toFixed(3)}). This residue likely contributes most to the predicted function.</p>;
              })()}
              <p><strong className="text-text-primary">Confidence Levels:</strong> High (&gt;80%) indicates strong compositional signal. Medium (60-80%) suggests moderate evidence. Low (&lt;60%) should be treated as tentative.</p>
            </div>
          </div>

          <p className="text-xs text-text-muted text-center">{result.note}</p>
        </motion.div>
      )}
    </div>
  );
}