#! /usr/bin/env python

# Author: Martin C. Frith 2008

# SPDX-License-Identifier: GPL-3.0-or-later

# Read pair-wise alignments in MAF or LAST tabular format: write an

# "Oxford grid", a.k.a. dotplot.

# TODO: Currently, pixels with zero aligned nt-pairs are white, and

# pixels with one or more aligned nt-pairs are black.  This can look

# too crowded for large genome alignments.  I tried shading each pixel

# according to the number of aligned nt-pairs within it, but the

# result is too faint.  How can this be done better?

import collections

import functools

import gzip

from fnmatch import fnmatchcase

import logging

from operator import itemgetter

import subprocess

import itertools, optparse, os, re, sys

# Try to make PIL/PILLOW work:

try:

    from PIL import Image, ImageDraw, ImageFont, ImageColor

except ImportError:

    import Image, ImageDraw, ImageFont, ImageColor

try:

    from future_builtins import zip

except ImportError:

    pass

def myOpen(fileName):  # faster than fileinput

    if fileName is None:

        return []

    if fileName == "-":

        return sys.stdin

    if fileName.endswith(".gz"):

        return gzip.open(fileName, "rt")  # xxx dubious for Python2

    return open(fileName)

def groupByFirstItem(things):

    for k, v in itertools.groupby(things, itemgetter(0)):

        yield k, [i[1:] for i in v]

def croppedBlocks(blocks, ranges1, ranges2):

    headBeg1, headBeg2, headSize = blocks[0]

    for r1 in ranges1:

        for r2 in ranges2:

            cropBeg1, cropEnd1 = r1

            if headBeg1 < 0:

                cropBeg1, cropEnd1 = -cropEnd1, -cropBeg1

            cropBeg2, cropEnd2 = r2

            if headBeg2 < 0:

                cropBeg2, cropEnd2 = -cropEnd2, -cropBeg2

            for beg1, beg2, size in blocks:

                b1 = max(cropBeg1, beg1)

                e1 = min(cropEnd1, beg1 + size)

                if b1 >= e1: continue

                offset = beg2 - beg1

                b2 = max(cropBeg2, b1 + offset)

                e2 = min(cropEnd2, e1 + offset)

                if b2 >= e2: continue

                yield b2 - offset, b2, e2 - b2

def tabBlocks(beg1, beg2, blocks):

    '''Get the gapless blocks of an alignment, from LAST tabular format.'''

    for i in blocks.split(","):

        if ":" in i:

            x, y = i.split(":")

            beg1 += int(x)

            beg2 += int(y)

        else:

            size = int(i)

            yield beg1, beg2, size

            beg1 += size

            beg2 += size

def mafBlocks(beg1, beg2, seq1, seq2):

    '''Get the gapless blocks of an alignment, from MAF format.'''

    size = 0

    for x, y in zip(seq1, seq2):

        if x == "-":

            if size:

                yield beg1, beg2, size

                beg1 += size

                beg2 += size

                size = 0

            beg2 += 1

        elif y == "-":

            if size:

                yield beg1, beg2, size

                beg1 += size

                beg2 += size

                size = 0

            beg1 += 1

        else:

            size += 1

    if size: yield beg1, beg2, size

def alignmentFromSegment(qrySeqName, qrySeqLen, segment):

    refSeqLen = sys.maxsize  # XXX

    refSeqName, refSeqBeg, qrySeqBeg, size = segment

    block = refSeqBeg, qrySeqBeg, size

    return refSeqName, refSeqLen, qrySeqName, qrySeqLen, [block]

def alignmentInput(lines):

    '''Get alignments and sequence lengths, from MAF or tabular format.'''

    mafCount = 0

    qrySeqName = ""

    segments = []

    for line in lines:

        w = line.split()

        if line[0] == "#":

            pass

        elif len(w) == 1:

            for i in segments:

                yield alignmentFromSegment(qrySeqName, qrySeqLen, i)

            qrySeqName = w[0]

            qrySeqLen = 0

            segments = []

        elif len(w) == 2 and qrySeqName and w[1].isdigit():

            qrySeqLen += int(w[1])

        elif len(w) == 4 and qrySeqName and w[1].isdigit() and w[3].isdigit():

            refSeqName, refSeqBeg, refSeqEnd = w[0], int(w[1]), int(w[3])

            size = abs(refSeqEnd - refSeqBeg)

            if refSeqBeg > refSeqEnd:

                refSeqBeg = -refSeqBeg

            segments.append((refSeqName, refSeqBeg, qrySeqLen, size))

            qrySeqLen += size

        elif line[0].isdigit():  # tabular format

            chr1, beg1, seqlen1 = w[1], int(w[2]), int(w[5])

            if w[4] == "-": beg1 -= seqlen1

            chr2, beg2, seqlen2 = w[6], int(w[7]), int(w[10])

            if w[9] == "-": beg2 -= seqlen2

            blocks = tabBlocks(beg1, beg2, w[11])

            yield chr1, seqlen1, chr2, seqlen2, blocks

        elif line[0] == "s":  # MAF format

            if mafCount == 0:

                chr1, beg1, seqlen1, seq1 = w[1], int(w[2]), int(w[5]), w[6]

                if w[4] == "-": beg1 -= seqlen1

                mafCount = 1

            else:

                chr2, beg2, seqlen2, seq2 = w[1], int(w[2]), int(w[5]), w[6]

                if w[4] == "-": beg2 -= seqlen2

                blocks = mafBlocks(beg1, beg2, seq1, seq2)

                yield chr1, seqlen1, chr2, seqlen2, blocks

                mafCount = 0

    for i in segments:

        yield alignmentFromSegment(qrySeqName, qrySeqLen, i)

def seqRequestFromText(text):

    if ":" in text:

        pattern, interval = text.rsplit(":", 1)

        if "-" in interval:

            beg, end = interval.rsplit("-", 1)

            return pattern, int(beg), int(end)  # beg may be negative

    return text, 0, sys.maxsize

def rangesFromSeqName(seqRequests, name, seqLen):

    if seqRequests:

        base = name.split(".")[-1]  # allow for names like hg19.chr7

        for pat, beg, end in seqRequests:

            if fnmatchcase(name, pat) or fnmatchcase(base, pat):

                yield max(beg, 0), min(end, seqLen)

    else:

        yield 0, seqLen

def updateSeqs(coverDict, seqRanges, seqName, ranges, coveredRange):

    beg, end = coveredRange

    if beg < 0:

        coveredRange = -end, -beg

    if seqName in coverDict:

        coverDict[seqName].append(coveredRange)

    else:

        coverDict[seqName] = [coveredRange]

        for beg, end in ranges:

            r = seqName, beg, end

            seqRanges.append(r)

def readAlignments(fileName, opts):

    '''Get alignments and sequence limits, from MAF or tabular format.'''

    seqRequests1 = [seqRequestFromText(i) for i in opts.seq1]

    seqRequests2 = [seqRequestFromText(i) for i in opts.seq2]

    alignments = []

    seqRanges1 = []

    seqRanges2 = []

    coverDict1 = {}

    coverDict2 = {}

    lines = myOpen(fileName)

    for seqName1, seqLen1, seqName2, seqLen2, blocks in alignmentInput(lines):

        ranges1 = sorted(rangesFromSeqName(seqRequests1, seqName1, seqLen1))

        if not ranges1: continue

        ranges2 = sorted(rangesFromSeqName(seqRequests2, seqName2, seqLen2))

        if not ranges2: continue

        b = list(croppedBlocks(list(blocks), ranges1, ranges2))

        if not b: continue

        aln = seqName1, seqName2, b

        alignments.append(aln)

        coveredRange1 = b[0][0], b[-1][0] + b[-1][2]

        updateSeqs(coverDict1, seqRanges1, seqName1, ranges1, coveredRange1)

        coveredRange2 = b[0][1], b[-1][1] + b[-1][2]

        updateSeqs(coverDict2, seqRanges2, seqName2, ranges2, coveredRange2)

    return alignments, seqRanges1, coverDict1, seqRanges2, coverDict2

def nameAndRangesFromDict(cropDict, seqName):

    if seqName in cropDict:

        return seqName, cropDict[seqName]

    n = seqName.split(".")[-1]

    if n in cropDict:

        return n, cropDict[n]

    return seqName, []

def rangesForSecondaryAlignments(primaryRanges, seqLen):

    if primaryRanges:

        return primaryRanges

    return [(0, seqLen)]

def readSecondaryAlignments(opts, cropRanges1, cropRanges2):

    cropDict1 = dict(groupByFirstItem(cropRanges1))

    cropDict2 = dict(groupByFirstItem(cropRanges2))

    alignments = []

    seqRanges1 = []

    seqRanges2 = []

    coverDict1 = {}

    coverDict2 = {}

    lines = myOpen(opts.alignments)

    for seqName1, seqLen1, seqName2, seqLen2, blocks in alignmentInput(lines):

        seqName1, ranges1 = nameAndRangesFromDict(cropDict1, seqName1)

        seqName2, ranges2 = nameAndRangesFromDict(cropDict2, seqName2)

        if not ranges1 and not ranges2:

            continue

        r1 = rangesForSecondaryAlignments(ranges1, seqLen1)

        r2 = rangesForSecondaryAlignments(ranges2, seqLen2)

        b = list(croppedBlocks(list(blocks), r1, r2))

        if not b: continue

        aln = seqName1, seqName2, b

        alignments.append(aln)

        if not ranges1:

            coveredRange1 = b[0][0], b[-1][0] + b[-1][2]

            updateSeqs(coverDict1, seqRanges1, seqName1, r1, coveredRange1)

        if not ranges2:

            coveredRange2 = b[0][1], b[-1][1] + b[-1][2]

            updateSeqs(coverDict2, seqRanges2, seqName2, r2, coveredRange2)

    return alignments, seqRanges1, coverDict1, seqRanges2, coverDict2

def twoValuesFromOption(text, separator):

    if separator in text:

        return text.split(separator)

    return text, text

def mergedRanges(ranges):

    oldBeg, maxEnd = ranges[0]

    for beg, end in ranges:

        if beg > maxEnd:

            yield oldBeg, maxEnd

            oldBeg = beg

            maxEnd = end

        elif end > maxEnd:

            maxEnd = end

    yield oldBeg, maxEnd

def mergedRangesPerSeq(coverDict):

    for k, v in coverDict.items():

        v.sort()

        yield k, list(mergedRanges(v))

def coveredLength(mergedCoverDict):

    return sum(sum(e - b for b, e in v) for v in mergedCoverDict.values())

def trimmed(seqRanges, coverDict, minAlignedBases, maxGapFrac, endPad, midPad):

    maxEndGapFrac, maxMidGapFrac = twoValuesFromOption(maxGapFrac, ",")

    maxEndGap = max(float(maxEndGapFrac) * minAlignedBases, endPad * 1.0)

    maxMidGap = max(float(maxMidGapFrac) * minAlignedBases, midPad * 2.0)

    for seqName, rangeBeg, rangeEnd in seqRanges:

        seqBlocks = coverDict[seqName]

        blocks = [i for i in seqBlocks if i[0] < rangeEnd and i[1] > rangeBeg]

        if blocks[0][0] - rangeBeg > maxEndGap:

            rangeBeg = blocks[0][0] - endPad

        for j, y in enumerate(blocks):

            if j:

                x = blocks[j - 1]

                if y[0] - x[1] > maxMidGap:

                    yield seqName, rangeBeg, x[1] + midPad

                    rangeBeg = y[0] - midPad

        if rangeEnd - blocks[-1][1] > maxEndGap:

            rangeEnd = blocks[-1][1] + endPad

        yield seqName, rangeBeg, rangeEnd

def rangesWithStrandInfo(seqRanges, strandOpt, alignments, seqIndex):

    if strandOpt == "1":

        forwardMinusReverse = collections.defaultdict(int)

        for i in alignments:

            blocks = i[2]

            beg1, beg2, size = blocks[0]

            numOfAlignedLetterPairs = sum(i[2] for i in blocks)

            if (beg1 < 0) != (beg2 < 0):  # opposite-strand alignment

                numOfAlignedLetterPairs *= -1

            forwardMinusReverse[i[seqIndex]] += numOfAlignedLetterPairs

    strandNum = 0

    for seqName, beg, end in seqRanges:

        if strandOpt == "1":

            strandNum = 1 if forwardMinusReverse[seqName] >= 0 else 2

        yield seqName, beg, end, strandNum

def natural_sort_key(my_string):

    '''Return a sort key for "natural" ordering, e.g. chr9 < chr10.'''

    parts = re.split(r'(\d+)', my_string)

    parts[1::2] = map(int, parts[1::2])

    return parts

def nameKey(oneSeqRanges):

    return natural_sort_key(oneSeqRanges[0][0])

def sizeKey(oneSeqRanges):

    return sum(b - e for n, b, e, s in oneSeqRanges), nameKey(oneSeqRanges)

def alignmentKey(seqNamesToLists, oneSeqRanges):

    seqName = oneSeqRanges[0][0]

    alignmentsOfThisSequence = seqNamesToLists[seqName]

    numOfAlignedLetterPairs = sum(i[3] for i in alignmentsOfThisSequence)

    toMiddle = numOfAlignedLetterPairs // 2

    for i in alignmentsOfThisSequence:

        toMiddle -= i[3]

        if toMiddle < 0:

            return i[1:3]  # sequence-rank and "position" of this alignment

def rankAndFlipPerSeq(seqRanges):

    rangesGroupedBySeqName = itertools.groupby(seqRanges, itemgetter(0))

    for rank, group in enumerate(rangesGroupedBySeqName):

        seqName, ranges = group

        strandNum = next(ranges)[3]

        flip = 1 if strandNum < 2 else -1

        yield seqName, (rank, flip)

def alignmentSortData(alignments, seqIndex, otherNamesToRanksAndFlips):

    otherIndex = 1 - seqIndex

    for i in alignments:

        blocks = i[2]

        otherRank, otherFlip = otherNamesToRanksAndFlips[i[otherIndex]]

        otherPos = otherFlip * abs(blocks[0][otherIndex] +

                                   blocks[-1][otherIndex] + blocks[-1][2])

        numOfAlignedLetterPairs = sum(i[2] for i in blocks)

        yield i[seqIndex], otherRank, otherPos, numOfAlignedLetterPairs

def mySortedRanges(seqRanges, sortOpt, seqIndex, alignments, otherRanges):

    rangesGroupedBySeqName = itertools.groupby(seqRanges, itemgetter(0))

    g = [list(ranges) for seqName, ranges in rangesGroupedBySeqName]

    for i in g:

        if i[0][3] > 1:

            i.reverse()

    if sortOpt == "1":

        g.sort(key=nameKey)

    if sortOpt == "2":

        g.sort(key=sizeKey)

    if sortOpt == "3":

        otherNamesToRanksAndFlips = dict(rankAndFlipPerSeq(otherRanges))

        alns = sorted(alignmentSortData(alignments, seqIndex,

                                        otherNamesToRanksAndFlips))

        alnsGroupedBySeqName = itertools.groupby(alns, itemgetter(0))

        seqNamesToLists = dict((k, list(v)) for k, v in alnsGroupedBySeqName)

        g.sort(key=functools.partial(alignmentKey, seqNamesToLists))

    return [j for i in g for j in i]

def allSortedRanges(opts, alignments, alignmentsB,

                    seqRanges1, seqRangesB1, seqRanges2, seqRangesB2):

    o1, oB1 = twoValuesFromOption(opts.strands1, ":")

    o2, oB2 = twoValuesFromOption(opts.strands2, ":")

    if o1 == "1" and o2 == "1":

        raise Exception("the strand options have circular dependency")

    seqRanges1 = list(rangesWithStrandInfo(seqRanges1, o1, alignments, 0))

    seqRanges2 = list(rangesWithStrandInfo(seqRanges2, o2, alignments, 1))

    seqRangesB1 = list(rangesWithStrandInfo(seqRangesB1, oB1, alignmentsB, 0))

    seqRangesB2 = list(rangesWithStrandInfo(seqRangesB2, oB2, alignmentsB, 1))

    o1, oB1 = twoValuesFromOption(opts.sort1, ":")

    o2, oB2 = twoValuesFromOption(opts.sort2, ":")

    if o1 == "3" and o2 == "3":

        raise Exception("the sort options have circular dependency")

    if o1 != "3":

        s1 = mySortedRanges(seqRanges1, o1, None, None, None)

    if o2 != "3":

        s2 = mySortedRanges(seqRanges2, o2, None, None, None)

    if o1 == "3":

        s1 = mySortedRanges(seqRanges1, o1, 0, alignments, s2)

    if o2 == "3":

        s2 = mySortedRanges(seqRanges2, o2, 1, alignments, s1)

    t1 = mySortedRanges(seqRangesB1, oB1, 0, alignmentsB, s2)

    t2 = mySortedRanges(seqRangesB2, oB2, 1, alignmentsB, s1)

    return s1 + t1, s2 + t2

def prettyNum(n):

    t = str(n)

    groups = []

    while t:

        groups.append(t[-3:])

        t = t[:-3]

    return ",".join(reversed(groups))

def sizeText(size):

    suffixes = "bp", "kb", "Mb", "Gb"

    for i, x in enumerate(suffixes):

        j = 10 ** (i * 3)

        if size < j * 10:

            return "%.2g" % (1.0 * size / j) + x

        if size < j * 1000 or i == len(suffixes) - 1:

            return "%.0f" % (1.0 * size / j) + x

def labelText(seqRange, labelOpt):

    seqName, beg, end, strandNum = seqRange

    if labelOpt == 1:

        return seqName + ": " + sizeText(end - beg)

    if labelOpt == 2:

        return seqName + ":" + prettyNum(beg) + ": " + sizeText(end - beg)

    if labelOpt == 3:

        return seqName + ":" + prettyNum(beg) + "-" + prettyNum(end)

    return seqName

def rangeLabels(seqRanges, labelOpt, font, fontsize, image_mode, textRot):

    if fontsize:

        image_size = 1, 1

        im = Image.new(image_mode, image_size)

        draw = ImageDraw.Draw(im)

    x = y = 0

    for r in seqRanges:

        text = labelText(r, labelOpt)

        if fontsize:

            x, y = draw.textsize(text, font=font)

            if textRot:

                x, y = y, x

        yield text, x, y, r[3]

def dataFromRanges(sortedRanges, font, fontSize, imageMode, labelOpt, textRot):

    for seqName, rangeBeg, rangeEnd, strandNum in sortedRanges:

        out = [seqName, str(rangeBeg), str(rangeEnd)]

        if strandNum > 0:

            out.append(".+-"[strandNum])

        logging.info("\t".join(out))

    logging.info("")

    rangeSizes = [e - b for n, b, e, s in sortedRanges]

    labs = list(rangeLabels(sortedRanges, labelOpt, font, fontSize,

                            imageMode, textRot))

    margin = max(i[2] for i in labs)

    # xxx the margin may be too big, because some labels may get omitted

    return rangeSizes, labs, margin

def div_ceil(x, y):

    '''Return x / y rounded up.'''

    q, r = divmod(x, y)

    return q + (r != 0)

def get_bp_per_pix(rangeSizes, pixTweenRanges, maxPixels):

    '''Get the minimum bp-per-pixel that fits in the size limit.'''

    logging.info("choosing bp per pixel...")

    numOfRanges = len(rangeSizes)

    maxPixelsInRanges = maxPixels - pixTweenRanges * (numOfRanges - 1)

    if maxPixelsInRanges < numOfRanges:

        raise Exception("can't fit the image: too many sequences?")

    negLimit = -maxPixelsInRanges

    negBpPerPix = sum(rangeSizes) // negLimit

    while True:

        if sum(i // negBpPerPix for i in rangeSizes) >= negLimit:

            return -negBpPerPix

        negBpPerPix -= 1

def getRangePixBegs(rangePixLens, pixTweenRanges, margin):

    '''Get the start pixel for each range.'''

    rangePixBegs = []

    pix_tot = margin - pixTweenRanges

    for i in rangePixLens:

        pix_tot += pixTweenRanges

        rangePixBegs.append(pix_tot)

        pix_tot += i

    return rangePixBegs

def pixelData(rangeSizes, bp_per_pix, pixTweenRanges, margin):

    '''Return pixel information about the ranges.'''

    rangePixLens = [div_ceil(i, bp_per_pix) for i in rangeSizes]

    rangePixBegs = getRangePixBegs(rangePixLens, pixTweenRanges, margin)

    tot_pix = rangePixBegs[-1] + rangePixLens[-1]

    return rangePixBegs, rangePixLens, tot_pix

def drawLineForward(hits, width, bp_per_pix, beg1, beg2, size):

    while True:

        q1, r1 = divmod(beg1, bp_per_pix)

        q2, r2 = divmod(beg2, bp_per_pix)

        hits[q2 * width + q1] |= 1

        next_pix = min(bp_per_pix - r1, bp_per_pix - r2)

        if next_pix >= size: break

        beg1 += next_pix

        beg2 += next_pix

        size -= next_pix

def drawLineReverse(hits, width, bp_per_pix, beg1, beg2, size):

    while True:

        q1, r1 = divmod(beg1, bp_per_pix)

        q2, r2 = divmod(beg2, bp_per_pix)

        hits[q2 * width + q1] |= 2

        next_pix = min(bp_per_pix - r1, r2 + 1)

        if next_pix >= size: break

        beg1 += next_pix

        beg2 -= next_pix

        size -= next_pix

def strandAndOrigin(ranges, beg, size):

    isReverseStrand = (beg < 0)

    if isReverseStrand:

        beg = -(beg + size)

    for rangeBeg, rangeEnd, isReverseRange, origin in ranges:

        if rangeEnd > beg:  # assumes the ranges are sorted

            return (isReverseStrand != isReverseRange), origin

def alignmentPixels(width, height, alignments, bp_per_pix,

                    rangeDict1, rangeDict2):

    hits = [0] * (width * height)  # the image data

    for seq1, seq2, blocks in alignments:

        beg1, beg2, size = blocks[0]

        isReverse1, ori1 = strandAndOrigin(rangeDict1[seq1], beg1, size)

        isReverse2, ori2 = strandAndOrigin(rangeDict2[seq2], beg2, size)

        for beg1, beg2, size in blocks:

            if isReverse1:

                beg1 = -(beg1 + size)

                beg2 = -(beg2 + size)

            if isReverse1 == isReverse2:

                drawLineForward(hits, width, bp_per_pix,

                                ori1 + beg1, ori2 + beg2, size)

            else:

                drawLineReverse(hits, width, bp_per_pix,

                                ori1 + beg1, ori2 - beg2 - 1, size)

    return hits

def orientedBlocks(alignments, seqIndex):

    otherIndex = 1 - seqIndex

    for a in alignments:

        seq1, seq2, blocks = a

        for b in blocks:

            beg1, beg2, size = b

            if b[seqIndex] < 0:

                b = -(beg1 + size), -(beg2 + size), size

            yield a[seqIndex], b[seqIndex], a[otherIndex], b[otherIndex], size

def drawJoins(im, alignments, bpPerPix, seqIndex, rangeDict1, rangeDict2):

    blocks = orientedBlocks(alignments, seqIndex)

    oldSeq1 = ""

    for seq1, beg1, seq2, beg2, size in sorted(blocks):

        isReverse1, ori1 = strandAndOrigin(rangeDict1[seq1], beg1, size)

        isReverse2, ori2 = strandAndOrigin(rangeDict2[seq2], beg2, size)

        end1 = beg1 + size - 1

        end2 = beg2 + size - 1

        if isReverse1:

            beg1 = -(beg1 + 1)

            end1 = -(end1 + 1)

        if isReverse2:

            beg2 = -(beg2 + 1)

            end2 = -(end2 + 1)

        newPix1 = (ori1 + beg1) // bpPerPix

        newPix2 = (ori2 + beg2) // bpPerPix

        if seq1 == oldSeq1:

            lowerPix2 = min(oldPix2, newPix2)

            upperPix2 = max(oldPix2, newPix2)

            midPix1 = (oldPix1 + newPix1) // 2

            if isReverse1:

                midPix1 = (oldPix1 + newPix1 + 1) // 2

                oldPix1, newPix1 = newPix1, oldPix1

            if upperPix2 - lowerPix2 > 1 and oldPix1 <= newPix1 <= oldPix1 + 1:

                if seqIndex == 0:

                    box = midPix1, lowerPix2, midPix1 + 1, upperPix2 + 1

                else:

                    box = lowerPix2, midPix1, upperPix2 + 1, midPix1 + 1

                im.paste("lightgray", box)

        oldPix1 = (ori1 + end1) // bpPerPix

        oldPix2 = (ori2 + end2) // bpPerPix

        oldSeq1 = seq1

def expandedSeqDict(seqDict):

    '''Allow lookup by short sequence names, e.g. chr7 as well as hg19.chr7.'''

    newDict = seqDict.copy()

    for name, x in seqDict.items():

        if "." in name:

            base = name.split(".")[-1]

            if base in newDict:  # an ambiguous case was found:

                return seqDict   # so give up completely

            newDict[base] = x

    return newDict

def readBed(fileName, rangeDict):

    for line in myOpen(fileName):

        w = line.split()

        if not w: continue

        seqName = w[0]

        if seqName not in rangeDict: continue

        beg = int(w[1])

        end = int(w[2])

        layer = 900

        color = "#fbf"

        if len(w) > 4:

            if w[4] != ".":

                layer = float(w[4])

            if len(w) > 5:

                if len(w) > 8 and w[8].count(",") == 2:

                    color = "rgb(" + w[8] + ")"

                else:

                    strand = w[5]

                    isRev = rangeDict[seqName][0][2]

                    if strand == "+" and not isRev or strand == "-" and isRev:

                        color = "#ffe8e8"

                    if strand == "-" and not isRev or strand == "+" and isRev:

                        color = "#e8e8ff"

        yield layer, color, seqName, beg, end

def commaSeparatedInts(text):

    return map(int, text.rstrip(",").split(","))

def readGenePred(opts, fileName, rangeDict):

    for line in myOpen(fileName):

        fields = line.split()

        if not fields: continue

        if fields[2] not in "+-": fields = fields[1:]

        seqName = fields[1]

        if seqName not in rangeDict: continue

        #strand = fields[2]

        cdsBeg = int(fields[5])

        cdsEnd = int(fields[6])

        exonBegs = commaSeparatedInts(fields[8])

        exonEnds = commaSeparatedInts(fields[9])

        for beg, end in zip(exonBegs, exonEnds):

            yield 300, opts.exon_color, seqName, beg, end

            b = max(beg, cdsBeg)

            e = min(end, cdsEnd)

            if b < e: yield 400, opts.cds_color, seqName, b, e

def readRmsk(fileName, rangeDict):

    for line in myOpen(fileName):

        fields = line.split()

        if len(fields) == 17:  # rmsk.txt

            seqName = fields[5]

            if seqName not in rangeDict: continue  # do this ASAP for speed

            beg = int(fields[6])

            end = int(fields[7])

            strand = fields[9]

            repeatClass = fields[11]

        elif len(fields) == 15:  # .out

            seqName = fields[4]

            if seqName not in rangeDict: continue

            beg = int(fields[5]) - 1

            end = int(fields[6])

            strand = fields[8]

            repeatClass = fields[10].split("/")[0]

        else:

            continue

        if repeatClass in ("Low_complexity", "Simple_repeat", "Satellite"):

            yield 200, "#fbf", seqName, beg, end

        elif (strand == "+") != rangeDict[seqName][0][2]:

            yield 100, "#ffe8e8", seqName, beg, end

        else:

            yield 100, "#e8e8ff", seqName, beg, end

def isExtraFirstGapField(fields):

    return fields[4].isdigit()

def readGaps(opts, fileName, rangeDict):

    '''Read locations of unsequenced gaps, from an agp or gap file.'''

    for line in myOpen(fileName):

        w = line.split()

        if not w or w[0][0] == "#": continue

        if isExtraFirstGapField(w): w = w[1:]

        if w[4] not in "NU": continue

        seqName = w[0]

        if seqName not in rangeDict: continue

        end = int(w[2])

        beg = end - int(w[5])  # zero-based coordinate

        if w[7] == "yes":

            yield 3000, opts.bridged_color, seqName, beg, end

        else:

            yield 2000, opts.unbridged_color, seqName, beg, end

def bedBoxes(beds, rangeDict, margin, edge, isTop, bpPerPix):

    for layer, color, seqName, bedBeg, bedEnd in beds:

        for rangeBeg, rangeEnd, isReverseRange, origin in rangeDict[seqName]:

            beg = max(bedBeg, rangeBeg)

            end = min(bedEnd, rangeEnd)

            if beg >= end: continue

            if isReverseRange:

                beg, end = -end, -beg

            if layer <= 1000:

                # include partly-covered pixels

                b = (origin + beg) // bpPerPix

                e = div_ceil(origin + end, bpPerPix)

            else:

                # exclude partly-covered pixels

                b = div_ceil(origin + beg, bpPerPix)

                e = (origin + end) // bpPerPix

                if e <= b: continue

                if bedEnd >= rangeEnd:  # include partly-covered end pixels

                    if isReverseRange:

                        b = (origin + beg) // bpPerPix

                    else:

                        e = div_ceil(origin + end, bpPerPix)

            if isTop:

                box = b, margin, e, edge

            else:

                box = margin, b, edge, e

            yield layer, color, box

def drawAnnotations(im, boxes):

    # xxx use partial transparency for different-color overlaps?

    for layer, color, box in boxes:

        im.paste(color, box)

def placedLabels(labels, rangePixBegs, rangePixLens, beg, end):

    '''Return axis labels with endpoint & sort-order information.'''

    maxWidth = end - beg

    for i, j, k in zip(labels, rangePixBegs, rangePixLens):

        text, textWidth, textHeight, strandNum = i

        if textWidth > maxWidth:

            continue

        labelBeg = j + (k - textWidth) // 2

        labelEnd = labelBeg + textWidth

        sortKey = textWidth - k

        if labelBeg < beg:

            sortKey += maxWidth * (beg - labelBeg)

            labelBeg = beg

            labelEnd = beg + textWidth

        if labelEnd > end:

            sortKey += maxWidth * (labelEnd - end)

            labelEnd = end

            labelBeg = end - textWidth

        yield sortKey, labelBeg, labelEnd, text, textHeight, strandNum

def nonoverlappingLabels(labels, minPixTweenLabels):

    '''Get a subset of non-overlapping axis labels, greedily.'''

    out = []

    for i in labels:

        beg = i[1] - minPixTweenLabels

        end = i[2] + minPixTweenLabels

        if all(j[2] <= beg or j[1] >= end for j in out):

            out.append(i)

    return out

def axisImage(labels, rangePixBegs, rangePixLens, textRot,

              textAln, font, image_mode, opts):

    '''Make an image of axis labels.'''

    beg = rangePixBegs[0]

    end = rangePixBegs[-1] + rangePixLens[-1]

    margin = max(i[2] for i in labels)

    labels = sorted(placedLabels(labels, rangePixBegs, rangePixLens, beg, end))

    minPixTweenLabels = 0 if textRot else opts.label_space

    labels = nonoverlappingLabels(labels, minPixTweenLabels)

    image_size = (margin, end) if textRot else (end, margin)

    im = Image.new(image_mode, image_size, opts.margin_color)

    draw = ImageDraw.Draw(im)

    for sortKey, labelBeg, labelEnd, text, textHeight, strandNum in labels:

        base = margin - textHeight if textAln else 0

        position = (base, labelBeg) if textRot else (labelBeg, base)

        fill = ("black", opts.forwardcolor, opts.reversecolor)[strandNum]

        draw.text(position, text, font=font, fill=fill)

    return im

def rangesWithOrigins(sortedRanges, rangePixBegs, rangePixLens, bpPerPix):

    for i, j, k in zip(sortedRanges, rangePixBegs, rangePixLens):

        seqName, rangeBeg, rangeEnd, strandNum = i

        isReverseRange = (strandNum > 1)

        if isReverseRange:

            origin = bpPerPix * (j + k) + rangeBeg

        else:

            origin = bpPerPix * j - rangeBeg

        yield seqName, (rangeBeg, rangeEnd, isReverseRange, origin)

def rangesPerSeq(sortedRanges, rangePixBegs, rangePixLens, bpPerPix):

    a = rangesWithOrigins(sortedRanges, rangePixBegs, rangePixLens, bpPerPix)

    for k, v in itertools.groupby(a, itemgetter(0)):

        yield k, sorted(i[1] for i in v)

def getFont(opts):

    if opts.fontfile:

        return ImageFont.truetype(opts.fontfile, opts.fontsize)

    fileNames = []

    try:

        x = ["fc-match", "-f%{file}", "arial"]

        p = subprocess.Popen(x, stdout=subprocess.PIPE, stderr=subprocess.PIPE,

                             universal_newlines=True)

        out, err = p.communicate()

        fileNames.append(out)

    except OSError as e:

        logging.info("fc-match error: " + str(e))

    fileNames.append("/Library/Fonts/Arial.ttf")  # for Mac

    for i in fileNames:

        try:

            font = ImageFont.truetype(i, opts.fontsize)

            logging.info("font: " + i)

            return font

        except IOError as e:

            logging.info("font load error: " + str(e))

    return ImageFont.load_default()

def sequenceSizesAndNames(seqRanges):

    for seqName, ranges in itertools.groupby(seqRanges, itemgetter(0)):

        size = sum(e - b for n, b, e in ranges)

        yield size, seqName

def biggestSequences(seqRanges, maxNumOfSequences):

    s = sorted(sequenceSizesAndNames(seqRanges), reverse=True)

    if len(s) > maxNumOfSequences:

        logging.warning("too many sequences - discarding the smallest ones")

        s = s[:maxNumOfSequences]

    return set(i[1] for i in s)

def remainingSequenceRanges(seqRanges, alignments, seqIndex):

    remainingSequences = set(i[seqIndex] for i in alignments)

    return [i for i in seqRanges if i[0] in remainingSequences]

def lastDotplot(opts, args):

    logLevel = logging.INFO if opts.verbose else logging.WARNING

    logging.basicConfig(format="%(filename)s: %(message)s", level=logLevel)

    font = getFont(opts)

    image_mode = 'RGB'

    forward_color = ImageColor.getcolor(opts.forwardcolor, image_mode)

    reverse_color = ImageColor.getcolor(opts.reversecolor, image_mode)

    zipped_colors = zip(forward_color, reverse_color)

    overlap_color = tuple([(i + j) // 2 for i, j in zipped_colors])

    maxGap1, maxGapB1 = twoValuesFromOption(opts.max_gap1, ":")

    maxGap2, maxGapB2 = twoValuesFromOption(opts.max_gap2, ":")

    logging.info("reading alignments...")

    alnData = readAlignments(args[0], opts)

    alignments, seqRanges1, coverDict1, seqRanges2, coverDict2 = alnData

    if not alignments: raise Exception("there are no alignments")

    logging.info("cutting...")

    coverDict1 = dict(mergedRangesPerSeq(coverDict1))

    coverDict2 = dict(mergedRangesPerSeq(coverDict2))

    minAlignedBases = min(coveredLength(coverDict1), coveredLength(coverDict2))

    pad = int(opts.pad * minAlignedBases)

    cutRanges1 = list(trimmed(seqRanges1, coverDict1, minAlignedBases,

                              maxGap1, pad, pad))

    cutRanges2 = list(trimmed(seqRanges2, coverDict2, minAlignedBases,

                              maxGap2, pad, pad))

    biggestSeqs1 = biggestSequences(cutRanges1, opts.maxseqs)

    cutRanges1 = [i for i in cutRanges1 if i[0] in biggestSeqs1]

    alignments = [i for i in alignments if i[0] in biggestSeqs1]

    cutRanges2 = remainingSequenceRanges(cutRanges2, alignments, 1)

    biggestSeqs2 = biggestSequences(cutRanges2, opts.maxseqs)

    cutRanges2 = [i for i in cutRanges2 if i[0] in biggestSeqs2]

    alignments = [i for i in alignments if i[1] in biggestSeqs2]

    cutRanges1 = remainingSequenceRanges(cutRanges1, alignments, 0)

    logging.info("reading secondary alignments...")

    alnDataB = readSecondaryAlignments(opts, cutRanges1, cutRanges2)

    alignmentsB, seqRangesB1, coverDictB1, seqRangesB2, coverDictB2 = alnDataB

    logging.info("cutting...")

    coverDictB1 = dict(mergedRangesPerSeq(coverDictB1))

    coverDictB2 = dict(mergedRangesPerSeq(coverDictB2))

    cutRangesB1 = trimmed(seqRangesB1, coverDictB1, minAlignedBases,

                          maxGapB1, 0, 0)

    cutRangesB2 = trimmed(seqRangesB2, coverDictB2, minAlignedBases,

                          maxGapB2, 0, 0)

    logging.info("sorting...")

    sortOut = allSortedRanges(opts, alignments, alignmentsB,

                              cutRanges1, cutRangesB1, cutRanges2, cutRangesB2)

    sortedRanges1, sortedRanges2 = sortOut

    textRot1 = "vertical".startswith(opts.rot1)

    i1 = dataFromRanges(sortedRanges1, font,

                        opts.fontsize, image_mode, opts.labels1, textRot1)

    rangeSizes1, labelData1, tMargin = i1

    textRot2 = "horizontal".startswith(opts.rot2)

    i2 = dataFromRanges(sortedRanges2, font,

                        opts.fontsize, image_mode, opts.labels2, textRot2)

    rangeSizes2, labelData2, lMargin = i2

    maxPixels1 = opts.width  - lMargin

    maxPixels2 = opts.height - tMargin

    bpPerPix1 = get_bp_per_pix(rangeSizes1, opts.border_pixels, maxPixels1)

    bpPerPix2 = get_bp_per_pix(rangeSizes2, opts.border_pixels, maxPixels2)

    bpPerPix = max(bpPerPix1, bpPerPix2)

    logging.info("bp per pixel = " + str(bpPerPix))

    p1 = pixelData(rangeSizes1, bpPerPix, opts.border_pixels, lMargin)

    rangePixBegs1, rangePixLens1, width = p1

    rangeDict1 = dict(rangesPerSeq(sortedRanges1, rangePixBegs1,

                                   rangePixLens1, bpPerPix))

    p2 = pixelData(rangeSizes2, bpPerPix, opts.border_pixels, tMargin)

    rangePixBegs2, rangePixLens2, height = p2

    rangeDict2 = dict(rangesPerSeq(sortedRanges2, rangePixBegs2,

                                   rangePixLens2, bpPerPix))

    logging.info("width:  " + str(width))

    logging.info("height: " + str(height))

    logging.info("processing alignments...")

    allAlignments = alignments + alignmentsB

    hits = alignmentPixels(width, height, allAlignments, bpPerPix,

                           rangeDict1, rangeDict2)

    logging.info("reading annotations...")

    rangeDict1 = expandedSeqDict(rangeDict1)

    beds1 = itertools.chain(readBed(opts.bed1, rangeDict1),

                            readRmsk(opts.rmsk1, rangeDict1),

                            readGenePred(opts, opts.genePred1, rangeDict1),

                            readGaps(opts, opts.gap1, rangeDict1))

    b1 = bedBoxes(beds1, rangeDict1, tMargin, height, True, bpPerPix)

    rangeDict2 = expandedSeqDict(rangeDict2)

    beds2 = itertools.chain(readBed(opts.bed2, rangeDict2),

                            readRmsk(opts.rmsk2, rangeDict2),

                            readGenePred(opts, opts.genePred2, rangeDict2),

                            readGaps(opts, opts.gap2, rangeDict2))

    b2 = bedBoxes(beds2, rangeDict2, lMargin, width, False, bpPerPix)

    boxes = sorted(itertools.chain(b1, b2))

    logging.info("drawing...")

    image_size = width, height

    im = Image.new(image_mode, image_size, opts.background_color)

    drawAnnotations(im, boxes)

    joinA, joinB = twoValuesFromOption(opts.join, ":")

    if joinA in "13":

        drawJoins(im, alignments, bpPerPix, 0, rangeDict1, rangeDict2)

    if joinB in "13":

        drawJoins(im, alignmentsB, bpPerPix, 0, rangeDict1, rangeDict2)

    if joinA in "23":

        drawJoins(im, alignments, bpPerPix, 1, rangeDict2, rangeDict1)

    if joinB in "23":

        drawJoins(im, alignmentsB, bpPerPix, 1, rangeDict2, rangeDict1)

    for i in range(height):

        for j in range(width):

            store_value = hits[i * width + j]

            xy = j, i

            if   store_value == 1: im.putpixel(xy, forward_color)

            elif store_value == 2: im.putpixel(xy, reverse_color)

            elif store_value == 3: im.putpixel(xy, overlap_color)

    if opts.fontsize != 0:

        axis1 = axisImage(labelData1, rangePixBegs1, rangePixLens1,

                          textRot1, False, font, image_mode, opts)

        if textRot1:

            axis1 = axis1.transpose(Image.ROTATE_90)

        axis2 = axisImage(labelData2, rangePixBegs2, rangePixLens2,

                          textRot2, textRot2, font, image_mode, opts)

        if not textRot2:

            axis2 = axis2.transpose(Image.ROTATE_270)

        im.paste(axis1, (0, 0))

        im.paste(axis2, (0, 0))

    for i in rangePixBegs1[1:]:

        box = i - opts.border_pixels, tMargin, i, height

        im.paste(opts.border_color, box)

    for i in rangePixBegs2[1:]:

        box = lMargin, i - opts.border_pixels, width, i

        im.paste(opts.border_color, box)

    im.save(args[1])

if __name__ == "__main__":

    usage = """%prog --help

   or: %prog [options] maf-or-tab-alignments dotplot.png

   or: %prog [options] maf-or-tab-alignments dotplot.gif

   or: ..."""

    description = "Draw a dotplot of pair-wise sequence alignments in MAF or tabular format."

    op = optparse.OptionParser(usage=usage, description=description)

    op.add_option("-v", "--verbose", action="count",

                  help="show progress messages & data about the plot")

    # Replace "width" & "height" with a single "length" option?

    op.add_option("-x", "--width", metavar="INT", type="int", default=1000,

                  help="maximum width in pixels (default: %default)")

    op.add_option("-y", "--height", metavar="INT", type="int", default=1000,

                  help="maximum height in pixels (default: %default)")

    op.add_option("-m", "--maxseqs", type="int", default=100, metavar="M",

                  help="maximum number of horizontal or vertical sequences "

                  "(default=%default)")

    op.add_option("-1", "--seq1", metavar="PATTERN", action="append",

                  default=[],

                  help="which sequences to show from the 1st genome")

    op.add_option("-2", "--seq2", metavar="PATTERN", action="append",

                  default=[],

                  help="which sequences to show from the 2nd genome")

    op.add_option("-c", "--forwardcolor", metavar="COLOR", default="red",

                  help="color for forward alignments (default: %default)")