last-dotplot

#! /usr/bin/env python

# 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
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

def myOpen(fileName):  # faster than fileinput
    if fileName is None:
        return []
    if fileName == "-":
        return sys.stdin
    if fileName.endswith(".gz"):
        return gzip.open(fileName)
    return open(fileName)

def warn(message):
    if opts.verbose:
        prog = os.path.basename(sys.argv[0])
        sys.stderr.write(prog + ": " + message + "\n")

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 itertools.izip(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 alignmentInput(lines):
    '''Get alignments and sequence lengths, from MAF or tabular format.'''
    mafCount = 0
    for line in lines:
        w = line.split()
        if 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

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 = map(seqRequestFromText, opts.seq1)
    seqRequests2 = map(seqRequestFromText, 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.iteritems():
        v.sort()
        yield k, list(mergedRanges(v))

def coveredLength(mergedCoverDict):
    return sum(sum(e - b for b, e in v) for v in mergedCoverDict.itervalues())

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])
        warn("\t".join(out))
    warn("")
    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.'''
    warn("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 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]
        else:
            continue
        if repeatClass in ("Low_complexity", "Simple_repeat"):
            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)
        out, err = p.communicate()
        fileNames.append(out)
    except OSError as e:
        warn("fc-match error: " + str(e))
    fileNames.append("/Library/Fonts/Arial.ttf")  # for Mac
    for i in fileNames:
        try:
            font = ImageFont.truetype(i, opts.fontsize)
            warn("font: " + i)
            return font
        except IOError as e:
            warn("font load error: " + str(e))
    return ImageFont.load_default()

def lastDotplot(opts, args):
    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, ":")

    warn("reading alignments...")
    alnData = readAlignments(args[0], opts)
    alignments, seqRanges1, coverDict1, seqRanges2, coverDict2 = alnData
    if not alignments: raise Exception("there are no alignments")
    warn("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))

    warn("reading secondary alignments...")
    alnDataB = readSecondaryAlignments(opts, cutRanges1, cutRanges2)
    alignmentsB, seqRangesB1, coverDictB1, seqRangesB2, coverDictB2 = alnDataB
    warn("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)

    warn("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)
    warn("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))

    warn("width:  " + str(width))
    warn("height: " + str(height))

    warn("processing alignments...")
    hits = alignmentPixels(width, height, alignments + alignmentsB, bpPerPix,
                           rangeDict1, rangeDict2)

    warn("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))

    warn("drawing...")

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

    drawAnnotations(im, boxes)

    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")
    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")
    # Replace "width" & "height" with a single "length" option?
    op.add_option("-x", "--width", type="int", default=1000,
                  help="maximum width in pixels (default: %default)")
    op.add_option("-y", "--height", type="int", default=1000,
                  help="maximum height in pixels (default: %default)")
    op.add_option("-c", "--forwardcolor", metavar="COLOR", default="red",
                  help="color for forward alignments (default: %default)")
    op.add_option("-r", "--reversecolor", metavar="COLOR", default="blue",
                  help="color for reverse alignments (default: %default)")
    op.add_option("--alignments", metavar="FILE", help="secondary alignments")
    op.add_option("--sort1", default="1", metavar="N",
                  help="genome1 sequence order: 0=input order, 1=name order, "
                  "2=length order, 3=alignment order (default=%default)")
    op.add_option("--sort2", default="1", metavar="N",
                  help="genome2 sequence order: 0=input order, 1=name order, "
                  "2=length order, 3=alignment order (default=%default)")
    op.add_option("--strands1", default="0", metavar="N", help=
                  "genome1 sequence orientation: 0=forward orientation, "
                  "1=alignment orientation (default=%default)")
    op.add_option("--strands2", default="0", metavar="N", help=
                  "genome2 sequence orientation: 0=forward orientation, "
                  "1=alignment orientation (default=%default)")
    op.add_option("--max-gap1", metavar="FRAC", default="0.5,2", help=
                  "maximum unaligned (end,mid) gap in genome1: "
                  "fraction of aligned length (default=%default)")
    op.add_option("--max-gap2", metavar="FRAC", default="0.5,2", help=
                  "maximum unaligned (end,mid) gap in genome2: "
                  "fraction of aligned length (default=%default)")
    op.add_option("--pad", metavar="FRAC", type="float", default=0.04, help=
                  "pad length when cutting unaligned gaps: "
                  "fraction of aligned length (default=%default)")
    op.add_option("--border-pixels", metavar="INT", type="int", default=1,
                  help="number of pixels between sequences (default=%default)")
    op.add_option("--border-color", metavar="COLOR", default="black",
                  help="color for pixels between sequences (default=%default)")
    # --break-color and/or --break-pixels for intra-sequence breaks?
    op.add_option("--margin-color", metavar="COLOR", default="#dcdcdc",
                  help="margin color")

    og = optparse.OptionGroup(op, "Text options")
    og.add_option("-f", "--fontfile", metavar="FILE",
                  help="TrueType or OpenType font file")
    og.add_option("-s", "--fontsize", metavar="SIZE", type="int", default=14,
                  help="TrueType or OpenType font size (default: %default)")
    og.add_option("--labels1", type="int", default=0, metavar="N", help=
                  "genome1 labels: 0=name, 1=name:length, "
                  "2=name:start:length, 3=name:start-end (default=%default)")
    og.add_option("--labels2", type="int", default=0, metavar="N", help=
                  "genome2 labels: 0=name, 1=name:length, "
                  "2=name:start:length, 3=name:start-end (default=%default)")
    og.add_option("--rot1", metavar="ROT", default="h",
                  help="text rotation for the 1st genome (default=%default)")
    og.add_option("--rot2", metavar="ROT", default="v",
                  help="text rotation for the 2nd genome (default=%default)")
    op.add_option_group(og)

    og = optparse.OptionGroup(op, "Annotation options")
    og.add_option("--bed1", metavar="FILE",
                  help="read genome1 annotations from BED file")
    og.add_option("--bed2", metavar="FILE",
                  help="read genome2 annotations from BED file")
    og.add_option("--rmsk1", metavar="FILE", help="read genome1 repeats from "
                  "RepeatMasker .out or rmsk.txt file")
    og.add_option("--rmsk2", metavar="FILE", help="read genome2 repeats from "
                  "RepeatMasker .out or rmsk.txt file")
    op.add_option_group(og)

    og = optparse.OptionGroup(op, "Gene options")
    og.add_option("--genePred1", metavar="FILE",
                  help="read genome1 genes from genePred file")
    og.add_option("--genePred2", metavar="FILE",
                  help="read genome2 genes from genePred file")
    og.add_option("--exon-color", metavar="COLOR", default="PaleGreen",
                  help="color for exons (default=%default)")
    og.add_option("--cds-color", metavar="COLOR", default="LimeGreen",
                  help="color for protein-coding regions (default=%default)")
    op.add_option_group(og)

    og = optparse.OptionGroup(op, "Unsequenced gap options")
    og.add_option("--gap1", metavar="FILE",
                  help="read genome1 unsequenced gaps from agp or gap file")
    og.add_option("--gap2", metavar="FILE",
                  help="read genome2 unsequenced gaps from agp or gap file")
    og.add_option("--bridged-color", metavar="COLOR", default="yellow",
                  help="color for bridged gaps (default: %default)")
    og.add_option("--unbridged-color", metavar="COLOR", default="orange",
                  help="color for unbridged gaps (default: %default)")
    op.add_option_group(og)
    (opts, args) = op.parse_args()
    if len(args) != 2: op.error("2 arguments needed")

    opts.background_color = "white"
    opts.label_space = 5     # minimum number of pixels between axis labels

    try: lastDotplot(opts, args)
    except KeyboardInterrupt: pass  # avoid silly error message
    except Exception as e:
        prog = os.path.basename(sys.argv[0])
        sys.exit(prog + ": error: " + str(e))