package jalview.analysis;

import jalview.datamodel.Alignment;
import jalview.datamodel.AlignmentI;
import jalview.datamodel.Sequence;
import jalview.datamodel.SequenceI;
import jalview.io.FastaFile;

import java.util.Arrays;
import java.util.Random;

/**
 * Generates, and outputs in Fasta format, a random DNA alignment for given
 * sequence length and count. Will regenerate the same alignment each time if
 * the same random seed is used (so may be used for reproducible unit tests).
 * Not guaranteed to reproduce the same results between versions, as the rules
 * may get tweaked to produce more 'realistic' results.
 * 
 * Arguments:
 * <ul>
 * <li>length (number of bases in each sequence)</li>
 * <li>height (number of sequences)</li>
 * <li>a whole number random seed</li>
 * <li>percentage of gaps to include (0-100)</li>
 * <li>percentage chance of variation of each position (0-100)</li>
 * </ul>
 * 
 * @author gmcarstairs
 *
 */
public class DnaAlignmentGenerator
{
  private static final char GAP = '-';

  private static final char ZERO = '0';

  private static final char[] BASES = new char[] { 'G', 'T', 'C', 'A' };

  private Random random;

  /**
   * Outputs a DNA 'alignment' where each position is a random choice from
   * 'GTCA-'.
   * 
   * @param args
   */
  public static void main(String[] args)
  {
    if (args.length != 5)
    {
      usage();
      return;
    }
    int width = Integer.parseInt(args[0]);
    int height = Integer.parseInt(args[1]);
    long randomSeed = Long.valueOf(args[2]);
    int gapPercentage = Integer.valueOf(args[3]);
    int changePercentage = Integer.valueOf(args[4]);
    AlignmentI al = new DnaAlignmentGenerator().generate(width, height,
            randomSeed, gapPercentage, changePercentage);

    System.out.println("; " + height + " sequences of " + width
            + " bases with " + gapPercentage + "% gaps and "
            + changePercentage + "% mutations (random seed = " + randomSeed
            + ")");
    System.out.println(new FastaFile().print(al.getSequencesArray()));
  }

  /**
   * Print parameter help.
   */
  private static void usage()
  {
    System.out.println("Usage:");
    System.out.println("arg0: number of (non-gap) bases per sequence");
    System.out.println("arg1: number sequences");
    System.out
            .println("arg2: an integer as random seed (same seed = same results)");
    System.out.println("arg3: percentage of gaps to (randomly) generate");
    System.out
            .println("arg4: percentage of 'mutations' to (randomly) generate");
    System.out.println("Example: DnaAlignmentGenerator 12 15 387 10 5");
    System.out
            .println("- 15 sequences of 12 bases each, approx 10% gaps and 5% mutations, random seed = 387");

  }

  /**
   * Default constructor
   */
  public DnaAlignmentGenerator()
  {

  }

  /**
   * Outputs a DNA 'alignment' of given width and height, where each position is
   * a random choice from 'GTCA-'.
   * 
   * @param width
   * @param height
   * @param randomSeed
   * @param changePercentage
   * @param gapPercentage
   */
  public AlignmentI generate(int width, int height, long randomSeed,
          int gapPercentage, int changePercentage)
  {
    SequenceI[] seqs = new SequenceI[height];
    random = new Random(randomSeed);
    seqs[0] = generateSequence(1, width, gapPercentage);
    for (int seqno = 1; seqno < height; seqno++)
    {
      seqs[seqno] = generateAnotherSequence(seqs[0].getSequence(),
              seqno + 1, width, changePercentage);
    }
    AlignmentI al = new Alignment(seqs);
    return al;
  }

  /**
   * Outputs a DNA 'sequence' of given length, with some random gaps included.
   * 
   * @param seqno
   * @param length
   * @param gapPercentage
   */
  private SequenceI generateSequence(int seqno, int length,
          int gapPercentage)
  {
    StringBuilder seq = new StringBuilder(length);

    /*
     * Loop till we've added 'length' bases (excluding gaps)
     */
    for (int count = 0; count < length;)
    {
      boolean addGap = random.nextInt(100) < gapPercentage;
      char c = addGap ? GAP : BASES[random.nextInt(Integer.MAX_VALUE) % 4];
      seq.append(c);
      if (!addGap)
      {
        count++;
      }
    }
    final String seqName = "SEQ" + seqno;
    final String seqString = seq.toString();
    SequenceI sq = new Sequence(seqName, seqString);
    sq.createDatasetSequence();
    return sq;
  }

  /**
   * Generate a sequence approximately aligned to the first one.
   * 
   * @param ds
   * @param seqno
   * @param width
   *          number of bases
   * @param changePercentage
   * @return
   */
  private SequenceI generateAnotherSequence(char[] ds, int seqno,
          int width, int changePercentage)
  {
    int length = ds.length;
    char[] seq = new char[length];
    Arrays.fill(seq, ZERO);
    int gapsWanted = length - width;
    int gapsAdded = 0;

    /*
     * First 'randomly' mimic gaps in model sequence.
     */
    for (int pos = 0; pos < length; pos++)
    {
      if (ds[pos] == GAP)
      {
        /*
         * Add a gap at the same position with changePercentage likelihood
         */
        seq[pos] = randomCharacter(GAP, changePercentage);
        if (seq[pos] == GAP)
        {
          gapsAdded++;
        }
      }
    }

    /*
     * Next scatter any remaining gaps (if any) at random. This gives an even
     * distribution.
     */
    while (gapsAdded < gapsWanted)
    {
      boolean added = false;
      while (!added)
      {
        int pos = random.nextInt(length);
        if (seq[pos] != GAP)
        {
          seq[pos] = GAP;
          added = true;
          gapsAdded++;
        }
      }
    }

    /*
     * Finally fill in the rest with randomly mutated bases.
     */
    for (int pos = 0; pos < length; pos++)
    {
      if (seq[pos] == ZERO)
      {
        char c = randomCharacter(ds[pos], changePercentage);
        seq[pos] = c;
      }
    }
    final String seqName = "SEQ" + seqno;
    final String seqString = new String(seq);
    SequenceI sq = new Sequence(seqName, seqString);
    sq.createDatasetSequence();
    return sq;
  }

  /**
   * Returns a random character that is changePercentage% likely to match the
   * given type (as base or gap).
   * 
   * @param changePercentage
   * 
   * @param c
   * @return
   */
  private char randomCharacter(char c, int changePercentage)
  {
    final boolean mutation = random.nextInt(100) < changePercentage;

    if (!mutation)
    {
      return c;
    }

    char newchar = c;
    while (newchar == c)
    {
      newchar = BASES[random.nextInt(Integer.MAX_VALUE) % 4];
    }
    return newchar;
  }
}