src/lib/anonymizer.ts

export function normalizeInputText(rawText: string): string {
  return String(rawText || "")
    .replace(/\r/g, "\n")
    .replace(/\bblanco\b/gi, " ")
    .replace(/\bletras blanco\b/gi, " ")
    .replace(/\bblancobl\b/gi, " ")
    .replace(/[ \t]+/g, " ")
    .replace(/\n{2,}/g, "\n")
    .trim();
}

export function removeAdministrativePatterns(text: string): string {
  const patternsToRemove = [
    /Nombre:.*$/gim,
    /Número:.*$/gim,
    /Historia:.*$/gim,
    /T\.\s*Sanitaria:.*$/gim,
    /Solicitante:.*$/gim,
    /Servicio:.*$/gim,
    /Destino[s]?:.*$/gim,
    /Destinos:.*$/gim,
    /Centro:.*$/gim,
    /Sexo:.*$/gim,
    /Edad:.*$/gim,
    /Habitación:.*$/gim,
    /Cama:.*$/gim,
    /Recep\.?Muestra.*$/gim,
    /Fch\.?Informe.*$/gim,
    /Fecha de análisis:.*$/gim,
    /Resultados validados por:.*$/gim,
    /Tipo de Muestra:.*$/gim,
    /Tipo de informe:.*$/gim,
    /Página\s*\d+\/\d+/gim,
    /Hospital Ernest Lluch.*$/gim,
    /H\.\s*ERNEST LLUCH.*$/gim,
    /INFORME DE RESULTADOS.*$/gim,
    /Calatayud,.*$/gim,
    /Tecnica:.*$/gim,
    /T[eé]cnica:.*$/gim,
    /_{5,}/g
  ];

  let cleaned = text;
  patternsToRemove.forEach((rx) => {
    cleaned = cleaned.replace(rx, "");
  });

  return cleaned;
}

export function removeDirectIdentifiers(text: string): string {
  return text
    .replace(/\b[A-Z]{2}\d{9,}[A-Z]?\b/g, " ")
    .replace(/\b\d{6,}\b/g, " ")
    .replace(/\b\d{1,2}\/\d{1,2}\/\d{2,4}\b/g, " ")
    .replace(/\b\d{1,2}:\d{2}:\d{2}\b/g, " ")
    .replace(/\b[A-Z]{2,}-[A-Z]{2,}\b/g, " ")
    .replace(/[A-ZÁÉÍÓÚÑ]{2,},\s*[A-ZÁÉÍÓÚÑ\s]{2,}/g, " ");
}

export function keepUsefulLabLines(lines: string[]): string[] {
  const allowedSectionPatterns = [
    /^BIOQUIMICA$/i,
    /^BIOQUIMICA GENERAL$/i,
    /^HEMATOLOGIA$/i,
    /^HEMATIMETRIA$/i,
    /^HEMOSTASIA$/i,
    /^GASOMETRIA$/i,
    /^ORINA$/i,
    /^COAGULACION$/i,
    /^OTRAS PRUEBAS:?$/i
  ];

  const obviousNoisePatterns = [
    /^Magnitud Resultado Unidades Intervalo de Referencia Biológico$/i,
    /^Tipo de informe:.*$/i,
    /^Resultados validados por:.*$/i,
    /^Tipo de Muestra:.*$/i,
    /^\.+$/,
    /^-$/,
    /^Hospital .*$/i,
    /^INFORME DE RESULTADOS$/i,
    /^Nombre:.*$/i,
    /^Historia:.*$/i,
    /^Sexo:.*$/i,
    /^Edad:.*$/i,
    /^Recep\.Muestra.*$/i,
    /^Fch\.Informe.*$/i
  ];

  return lines.filter((line) => {
    if (!line || line.length < 2) return false;
    if (obviousNoisePatterns.some((rx) => rx.test(line))) return false;
    if (allowedSectionPatterns.some((rx) => rx.test(line))) return true;

    const hasPend = /\bPEND\b/i.test(line);
    const hasValue = /\b\d+[.,]?\d*\b/.test(line);
    const hasRange = /\b\d+[.,]?\d*\s*-\s*\d+[.,]?\d*\b/.test(line);
    const hasUnits =
      /\b(mg\/dL|g\/dL|mmol\/L|mEq\/L|mil\/mm3|mill\/mm3|fl|pg|%|u\/L|ui\/L|ng\/mL|mL\/min\/1\.73m\^?2|mL\/min|seg|s|mill|mil|pg|fl)\b/i.test(
        line
      );
    
    // If it has a value and units, or a value and a range, or is a section header, keep it.
    // Also keep lines that look like parameters if they are reasonably long.
    const looksLikeParameter = /^[A-ZÁÉÍÓÚÑ][A-ZÁÉÍÓÚÑ0-9\s\.\-\(\)\/]{3,}$/i.test(line);

    return hasPend || hasRange || hasUnits || hasValue || looksLikeParameter;
  });
}

export function finalizeCleanLines(text: string): string[] {
  return text
    .split("\n")
    .map((line) => line.trim())
    .map((line) => line.replace(/[ \t]+/g, " ").trim())
    .filter(Boolean);
}

export function anonymizeLabsText(rawText: string): string {
  let text = normalizeInputText(rawText);
  text = removeAdministrativePatterns(text);
  text = removeDirectIdentifiers(text);

  const cleanedLines = finalizeCleanLines(text);
  const usefulLines = keepUsefulLabLines(cleanedLines);

  const result = usefulLines.join("\n").trim();
  console.log("Anonymized Labs Text length:", result.length);
  return result;
}

export function anonymizeTreatmentsText(rawText: string): string {
  let text = normalizeInputText(rawText);
  text = removeAdministrativePatterns(text);
  text = removeDirectIdentifiers(text);

  const cleanedLines = finalizeCleanLines(text).filter((line) => {
    const hasDose =
      /\b\d+[.,]?\d*\s?(mg|mcg|g|ui|ml|ug)\b/i.test(line);
    const hasFrequency =
      /(cada\s+\d+\s*h|cada\s+\d+\s*horas|al d[ií]a|si precisa|por la noche|por la ma[ñn]ana|semanal|q\d+h)/i.test(
        line
      );
    const looksLikeDrug =
      /^[A-ZÁÉÍÓÚÑa-záéíóúñ][A-ZÁÉÍÓÚÑa-záéíóúñ0-9\s\/\.\-]+$/i.test(line);

    return looksLikeDrug || hasDose || hasFrequency;
  });

  const result = cleanedLines.join("\n").trim();
  console.log("Anonymized Treatments Text length:", result.length);
  return result;
}