infill_generator.py

#!/usr/bin/env python

# infill_generator.py
#
# Generate hatch fills for the closed paths (polygons) in the currently
# selected document elements.  If no elements are selected, then all the
# polygons throughout the document are hatched.  The fill rule is an odd/even
# rule: odd numbered intersections (1, 3, 5, etc.) are a hatch line entering
# a polygon while even numbered intersections (2, 4, 6, etc.) are the same
# hatch line exiting the polygon.
#
# This extension first decomposes the selected <path>, <rect>, <line>,
# <polyline>, <polygon>, <circle>, and <ellipse> elements into individual
# moveto and lineto coordinates. These coordinates are then used to build vertex lists.
# Only the vertex lists corresponding to polygons (closed paths) are
# kept.  Note that a single graphical element may be composed of several
# subpaths, each subpath potentially a polygon.
#
# Once the lists of all the vertices are built, potential hatch lines are
# "projected" through the bounding box containing all of the vertices.
# For each potential hatch line, all intersections with all the polygon
# edges are determined.  These intersections are stored as decimal fractions
# indicating where along the length of the hatch line the intersection
# occurs.  These values will always be in the range [0, 1].  A value of 0
# indicates that the intersection is at the start of the hatch line, a value
# of 0.5 midway, and a value of 1 at the end of the hatch line.
#
# For a given hatch line, all the fractional values are sorted and any
# duplicates removed.  Duplicates occur, for instance, when the hatch
# line passes through a polygon vertex and thus intersects two edges
# segments of the polygon: the end of one edge and the start of
# another.
#
# Once sorted and duplicates removed, an odd/even rule is applied to
# determine which segments of the potential hatch line are within
# polygons.  These segments found to be within polygons are then saved
# and become the hatch fill lines which will be drawn.
#
# With each saved hatch fill line, information about which SVG graphical
# element it is within is saved.  This way, the hatch fill lines can
# later be grouped with the element they are associated with.  This makes
# it possible to manipulate the two -- graphical element and hatch lines --
# as a single object within Inkscape.
#
# Note: we also save the transformation matrix for each graphical element.
# That way, when we group the hatch fills with the element they are
# filling, we can invert the transformation.  That is, in order to compute
# the hatch fills, we first have had apply ALL applicable transforms to
# all the graphical elements.  We need to do that so that we know where in
# the drawing  each of the graphical elements are relative to one another.
# However, this means that the hatch lines have been computed in a setting
# where no further transforms are needed.  If we then put these hatch lines
# into the same groups as the elements being hatched in the ORIGINAL
# drawing, then the hatch lines will have transforms applied again.  So,
# once we compute the hatch lines, we need to invert the transforms of
# the group they will be placed in and apply this inverse transform to the
# hatch lines.  Hence the need to save the transform matrix for every
# graphical element.

# Written by Daniel C. Newman for the Eggbot Project
# dan dot newman at mtbaldy dot us
# Last updated 28 November 2010
# 15 October 2010

# Updated by Windell H. Oskay, 6/14/2012
# Added tolerance parameter

# Update by Daniel C. Newman, 6/20/2012
# Add min span/gap width

# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

import simplepath
import simpletransform
import simplestyle
import cubicsuperpath
import cspsubdiv
import bezmisc
import math


N_PAGE_WIDTH = 3200
N_PAGE_HEIGHT = 800

'''
Geometry 101: Determing if two lines intersect

A line L is defined by two points in space P1 and P2.  Any point P on the
line L satisfies

	P = P1 + s (P2 - P1)

for some value of the real number s in the range (-infinity, infinity).
If we confine s to the range [0, 1] then we've described the line segment
with end points P1 and P2.

Consider now the line La defined by the points P1 and P2, and the line Lb
defined by the points P3 and P4.  Any points Pa and Pb on the lines La and
Lb therefore satisfy

	Pa = P1 + sa (P2 - P1)
	Pb = P3 + sb (P4 - P3)

for some values of the real numbers sa and sb.  To see if these two lines
La and Lb intersect, we wish to see if there are finite values sa and sb
for which

	Pa = Pb

Or, equivalently, we ask if there exists values of sa and sb for which
the equation

	P1 + sa (P2 - P1) = P3 + sb (P4 - P3)

holds.  If we confine ourselves to a two-dimensional plane, and take

	P1 = (x1, y1)
	P2 = (x2, y2)
	P3 = (x3, y3)
	P4 = (x4, y4)

we then find that we have two equations in two unknowns, sa and sb,

	x1 + sa ( x2 - x1 ) = x3 + sb ( x4 - x3 )
	y1 + sa ( y2 - y1 ) = y3 + sb ( y4 - y3 )

Solving these two equations for sa and sb yields,

	sa = [ ( y1 - y3 ) ( x4 - x3 ) - ( y4 - y3 ) ( x1 - x3 ) ] / d
	sb = [ ( y1 - y3 ) ( x2 - x1 ) - ( y2 - y1 ) ( x1 - x3 ) ] / d

where the denominator, d, is given by

	d = ( y4 - y3 ) ( x2 - x1 ) - ( y2 - y1 ) ( x4 - x3 )

Substituting these back for the point (x, y) of intersection gives

	x = x1 + sa ( x2 - x1 )
	y = y1 + sa ( y2 - y1 )

Note that

1. The lines are parallel when d = 0
2. The lines are coincident d = 0 and the numerators for sa & sb are zero
3. For line segments, sa and sb are in the range [0, 1]; any value outside
   that range indicates that the line segments do not intersect.
'''

def intersect( P1, P2, P3, P4 ):

	'''
	Determine if two line segments defined by the four points P1 & P2 and
	P3 & P4 intersect.  If they do intersect, then return the fractional
	point of intersection "sa" along the first line at which the
	intersection occurs.
	'''

	# Precompute these values -- note that we're basically shifting from
	#
	#		P = P1 + s (P2 - P1)
	#
	# to
	#
	# 		P = P1 + s D
	#
	# where D is a direction vector.  The solution remains the same of
	# course.  We'll just be computing D once for each line rather than
	# computing it a couple of times.

	D21x = P2[0] - P1[0]
	D21y = P2[1] - P1[1]
	D43x = P4[0] - P3[0]
	D43y = P4[1] - P3[1]

	# Denominator
	d = D21x * D43y - D21y * D43x

	# Return now if the denominator is zero
	if d == 0:
		return float( -1 )

	# For our purposes, the first line segment given
	# by P1 & P2 is the LONG hatch line running through
	# the entire drawing.  And, P3 & P4 describe the
	# usually much shorter line segment from a polygon.
	# As such, we compute sb first as it's more likely
	# to indicate "no intersection".  That is, sa is
	# more likely to indicate an intersection with a
	# much a long line containing P3 & P4.

	nb = ( P1[1] - P3[1] ) * D21x - ( P1[0] - P3[0] ) * D21y

	# Could first check if abs(nb) > abs(d) or if
	# the signs differ.
	sb = float( nb ) / float( d )
	if ( sb < 0 ) or ( sb > 1 ):
		return float( -1 )

	na = ( P1[1] - P3[1] ) * D43x -  ( P1[0] - P3[0] ) * D43y
	sa = float( na ) / float( d )
	if ( sa < 0 ) or ( sa > 1 ):
		return float( -1 )

	return sa

def interstices( P1, P2, paths, hatches, minGap=0.00001 ):

	'''
	For the line L defined by the points P1 & P2, determine the segments
	of L which lie within the polygons described by the paths stored in
	"paths"

	P1 -- (x,y) coordinate [list]
	P2 -- (x,y) coordinate [list]
	paths -- Dictionary of all the paths to check for intersections

	When an intersection of the line L is found with a polygon edge, then
	the fractional distance along the line L is saved along with the
	lxml.etree node which contained the intersecting polygon edge.  This
	fractional distance is always in the range [0, 1].

	Once all polygons have been checked, the list of fractional distances
	corresponding to intersections is sorted and any duplicates removed.
	It is then assumed that the first intersection is the line L entering
	a polygon; the second intersection the line leaving the polygon.  This
	line segment defined by the first and second intersection points is
	thus a hatch fill line we sought to generate.  In general, our hatch
	fills become the line segments described by intersection i and i+1
	with i an odd value (1, 3, 5, ...).  Since we know the lxml.etree node
	corresponding to each intersection, we can then correlate the hatch
	fill lines to the graphical elements in the original SVG document.
	This enables us to group hatch lines with the elements being hatched.

	The hatch line segments are returned by populating a dictionary.
	The dictionary is keyed off of the lxml.etree node pointer.  Each
	dictionary value is a list of 4-tuples,

		(x1, y1, x2, y2)

	where (x1, y1) and (x2, y2) are the (x,y) coordinates of the line
	segment's starting and ending points.
	'''

	sa = []

	# P1 & P2 is the hatch line
	# P3 & P4 is the polygon edge to check
	for path in paths:
		for subpath in paths[path]:
			P3 = subpath[0]
			for P4 in subpath[1:]:
				s = intersect( P1, P2, P3, P4 )
				if ( s >= 0.0 ) and ( s <= 1.0 ):
					# Save this intersection point along the hatch line
					sa.append( ( s, path ) )
				P3 = P4

	# Return now if there were no intersections
	if len( sa ) == 0:
		return None

	# Sort the intersections
	sa.sort()

	# Remove duplicates intersections.  A common case where these arise
	# is when the hatch line passes through a vertex where one line segment
	# ends and the next one begins.

	# Having had sorted the data, it's trivial to just scan through
	# removing duplicates as we go and then truncating the array
	n = len( sa )
	ilast = i = 1
	last = sa[0]
	while i < n:
		if abs( sa[i][0] - last[0] ) > 0.00001:
			sa[ilast] = last = sa[i]
			ilast += 1
		i += 1
	sa = sa[:ilast]
	if len( sa ) < 2:
		return

	if minGap <= 0.00001:
		minGap = 0.00001

	# Now, entries with even valued indices into sa[] are where we start
	# a hatch line and odd valued indices where we end the hatch line.

	last_sa = None
	i = 0
	while i < ( len( sa ) - 1 ):
	#for i in range( 0, len( sa ) - 1, 2 ):

		# Mind the gap
		if abs( sa[i][0] - sa[i+1][0] ) > minGap:

			# This segment has a (relative) length which exceeds minGap

			if ( last_sa is None ) or ( abs( sa[i][0] - last_sa[1][0] ) > minGap ):

				# And the gap between this segment and the prior segment
				# exceeds a (relative) length of minGap

				if not hatches.has_key( sa[i][1] ):
					hatches[sa[i][1]] = []

				x1 = P1[0] + sa[i][0] * ( P2[0] - P1[0] )
				y1 = P1[1] + sa[i][0] * ( P2[1] - P1[1] )
				x2 = P1[0] + sa[i+1][0] * ( P2[0] - P1[0] )
				y2 = P1[1] + sa[i+1][0] * ( P2[1] - P1[1] )

				hatches[sa[i][1]].append( [[x1, y1], [x2, y2]] )

			# Note we should not need to do the has_key() test, but let's
			# it anyway
			elif hatches.has_key( last_sa[0][1] ):

				# The gap between this segment of the hatch line and the prior
				# segment is so small that we should join the two so as to
				# prevent creating a gap of relative length <= minGap

				# Replace the endpoint of the prior hatch line with the endpoint
				# of this hatch line

				x2 = P1[0] + sa[i+1][0] * ( P2[0] - P1[0] )
				y2 = P1[1] + sa[i+1][0] * ( P2[1] - P1[1] )
				hatches[last_sa[0][1]][-1][1] = [x2, y2]

			# Remember the relative start and end of this hatch segment
			last_sa = [ sa[i], sa[i+1] ]

			i = i + 2

		else:
			i = i + 1
		# Ignore cases where the prior segment was short, this segment was
		# short, and the gap between them was short but we might just want
		# to combine them into one long segment.  That's indeed a possible
		# choice, but so may be the choice of dropping them silently in the
		# hatch segment wastebasket.

def inverseTransform ( tran ):
	'''
	An SVG transform matrix looks like

		[  a   c   e  ]
		[  b   d   f  ]
		[  0   0   1  ]

	And it's inverse is

		[  d   -c   cf - de  ]
		[ -b    a   be - af  ] * ( ad - bc ) ** -1
		[  0    0      1     ]

	And, no reasonable 2d coordinate transform will have
	the products ad and bc equal.

	SVG represents the transform matrix column by column as
	matrix(a b c d e f) while Inkscape extensions store the
	transform matrix as

		[[a, c, e], [b, d, f]]

	To invert the transform stored Inskcape style, we wish to
	produce

		[[d/D, -c/D, (cf - de)/D], [-b/D, a/D, (be-af)/D]]

	where

		D = 1 / (ad - bc)
	'''

	D = tran[0][0] * tran[1][1] - tran[1][0] * tran[0][1]
	if D == 0:
		return None

	return [[tran[1][1]/D, -tran[0][1]/D,
			(tran[0][1]*tran[1][2] - tran[1][1]*tran[0][2])/D],
			[-tran[1][0]/D, tran[0][0]/D,
			(tran[1][0]*tran[0][2] - tran[0][0]*tran[1][2])/D]]

def parseLengthWithUnits( str ):

	'''
	Parse an SVG value which may or may not have units attached
	This version is greatly simplified in that it only allows: no units,
	units of px, and units of %.  Everything else, it returns None for.
	There is a more general routine to consider in scour.py if more
	generality is ever needed.
	'''

	u = 'px'
	s = str.strip()
	if s[-2:] == 'px':
		s = s[:-2]
	elif s[-1:] == '%':
		u = '%'
		s = s[:-1]

	try:
		v = float( s )
	except:
		return None, None

	return v, u

# Lifted with impunity from eggbot.py

def subdivideCubicPath( sp, flat, i=1 ):

	"""
	Break up a bezier curve into smaller curves, each of which
	is approximately a straight line within a given tolerance
	(the "smoothness" defined by [flat]).

	This is a modified version of cspsubdiv.cspsubdiv() rewritten
	to avoid recurrence.
	"""

	while True:
		while True:
			if i >= len( sp ):
				return

			p0 = sp[i - 1][1]
			p1 = sp[i - 1][2]
			p2 = sp[i][0]
			p3 = sp[i][1]

			b = ( p0, p1, p2, p3 )

			if cspsubdiv.maxdist( b ) > flat:
				break

			i += 1

		one, two = bezmisc.beziersplitatt( b, 0.5 )
		sp[i - 1][2] = one[1]
		sp[i][0] = two[2]
		p = [one[2], one[3], two[1]]
		sp[i:1] = [p]


def distanceSquared( P1, P2 ):

	'''
	Pythagorean distance formula WITHOUT the square root.  Since
	we just want to know if the distance is less than some fixed
	fudge factor, we can just square the fudge factor once and run
	with it rather than compute square roots over and over.
	'''

	dx = P2[0] - P1[0]
	dy = P2[1] - P1[1]

	return ( dx * dx + dy * dy )

class InfillGenerator():

	def __init__( self , doc, nodeList):

		
		self.nodes = nodeList
		self.xmin, self.ymin = ( float( 0 ), float( 0 ) )
		self.xmax, self.ymax = ( float( 0 ), float( 0 ) )
		self.paths = {}
		self.grid = []
		self.hatches = {}
		self.transforms = {}
		self.minGap = float( 0 )

		# For handling an SVG viewbox attribute, we will need to know the
		# values of the document's <svg> width and height attributes as well
		# as establishing a transform from the viewbox to the display.
		self.document = doc
		self.docWidth = float( N_PAGE_WIDTH )
		self.docHeight = float( N_PAGE_HEIGHT )
		self.docTransform = [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]]


	def getLength( self, name, default ):

		'''
		Get the <svg> attribute with name "name" and default value "default"
		Parse the attribute into a value and associated units.  Then, accept
		no units (''), units of pixels ('px'), and units of percentage ('%').
		'''

		str = self.document.getroot().get( name )
		if str:
			v, u = parseLengthWithUnits( str )
			if not v:
				# Couldn't parse the value
				return None
			elif ( u == '' ) or ( u == 'px' ):
				return v
			elif u == '%':
				return float( default ) * v / 100.0
			else:
				# Unsupported units
				return None
		else:
			# No width specified; assume the default value
			return float( default )

	def getDocProps( self ):

		'''
		Get the document's height and width attributes from the <svg> tag.
		Use a default value in case the property is not present or is
		expressed in units of percentages.
		'''

		self.docHeight = self.getLength( 'height', N_PAGE_HEIGHT )
		self.docWidth = self.getLength( 'width', N_PAGE_WIDTH )
		if ( self.docHeight == None ) or ( self.docWidth == None ):
			return False
		else:
			return True

	def handleViewBox( self ):

		'''
		Set up the document-wide transform in the event that the document has an SVG viewbox
		'''

		if self.getDocProps():
			viewbox = self.document.getroot().get( 'viewBox' )
			if viewbox:
				vinfo = viewbox.strip().replace( ',', ' ' ).split( ' ' )
				if ( vinfo[2] != 0 ) and ( vinfo[3] != 0 ):
					sx = self.docWidth / float( vinfo[2] )
					sy = self.docHeight / float( vinfo[3] )
					self.docTransform = simpletransform.parseTransform( 'scale(%f,%f)' % (sx, sy) )

	def addPathVertices( self, path, node=None, transform=None ):

		'''
		Decompose the path data from an SVG element into individual
		subpaths, each starting with an absolute move-to (x, y)
		coordinate followed by one or more absolute line-to (x, y)
		coordinates.  Each subpath is stored as a list of (x, y)
		coordinates, with the first entry understood to be a
		move-to coordinate and the rest line-to coordinates.  A list
		is then made of all the subpath lists and then stored in the
		self.paths dictionary using the path's lxml.etree node pointer
		as the dictionary key.
		'''

		if ( not path ) or ( len( path ) == 0 ):
			return

		# parsePath() may raise an exception.  This is okay
		sp = simplepath.parsePath( path )
		if ( not sp ) or ( len( sp ) == 0 ):
			return

		# Get a cubic super duper path
		p = cubicsuperpath.CubicSuperPath( sp )
		if ( not p ) or ( len( p ) == 0 ):
			return

		# Apply any transformation
		if transform != None:
			simpletransform.applyTransformToPath( transform, p )

		# Now traverse the simplified path
		subpaths = []
		subpath_vertices = []
		for sp in p:
			# We've started a new subpath
			# See if there is a prior subpath and whether we should keep it
			if len( subpath_vertices ):
				if distanceSquared( subpath_vertices[0], subpath_vertices[-1] ) < 1:
					# Keep the prior subpath: it appears to be a closed path
					subpaths.append( subpath_vertices )
			subpath_vertices = []
			subdivideCubicPath( sp, float( self.options['tolerance'] /10 ) )
			for csp in sp:
				# Add this vertex to the list of vetices
				subpath_vertices.append( csp[1] )

		# Handle final subpath
		if len( subpath_vertices ):
			if distanceSquared( subpath_vertices[0], subpath_vertices[-1] ) < 1:
				# Path appears to be closed so let's keep it
				subpaths.append( subpath_vertices )

		# Empty path?
		if len( subpaths ) == 0:
			return

		# And add this path to our dictionary of paths
		self.paths[node] = subpaths

		# And save the transform for this element in a dictionary keyed
		# by the element's lxml node pointer
		self.transforms[node] = transform

	def getBoundingBox( self ):

		'''
		Determine the bounding box for our collection of polygons
		'''

		self.xmin, self.xmax = float( 1.0E70 ), float( -1.0E70 )
		self.ymin, self.ymax = float( 1.0E70 ), float( -1.0E70 )

		for path in self.paths:
			for subpath in self.paths[path]:
				for vertex in subpath:
					if vertex[0] < self.xmin:
						self.xmin = vertex[0]
					elif vertex[0] > self.xmax:
						self.xmax = vertex[0]
					if vertex[1] < self.ymin:
						self.ymin = vertex[1]
					elif vertex[1] > self.ymax:
						self.ymax = vertex[1]

	def recursivelyTraverseSvg( self, aNodeList,
			matCurrent=[[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]],
			parent_visibility='visible' ):

		'''
		Recursively walk the SVG document, building polygon vertex lists
		for each graphical element we support.

		Rendered SVG elements:
			<circle>, <ellipse>, <line>, <path>, <polygon>, <polyline>, <rect>

		Supported SVG elements:
			<group>, <use>

		Ignored SVG elements:
			<defs>, <eggbot>, <metadata>, <namedview>, <pattern>

		All other SVG elements trigger an error (including <text>)
		'''

		for node in aNodeList:

			# Ignore invisible nodes
			v = node.get( 'visibility', parent_visibility )
			if v == 'inherit':
				v = parent_visibility
			if v == 'hidden' or v == 'collapse':
				pass

			# first apply the current matrix transform to this node's tranform
			#matNew = simpletransform.composeTransform( matCurrent,
			#	simpletransform.parseTransform( node.get( "transform" ) ) )
			matNew = matCurrent

			if node.tag == addNS( 'g', 'svg' ) or node.tag == 'g':

				self.recursivelyTraverseSvg( node, matNew, parent_visibility=v )

			elif node.tag == addNS( 'use', 'svg' ) or node.tag == 'use':

				# A <use> element refers to another SVG element via an xlink:href="#blah"
				# attribute.  We will handle the element by doing an XPath search through
				# the document, looking for the element with the matching id="blah"
				# attribute.  We then recursively process that element after applying
				# any necessary (x,y) translation.
				#
				# Notes:
				#  1. We ignore the height and width attributes as they do not apply to
				#     path-like elements, and
				#  2. Even if the use element has visibility="hidden", SVG still calls
				#     for processing the referenced element.  The referenced element is
				#     hidden only if its visibility is "inherit" or "hidden".

				refid = node.get( addNS( 'href', 'xlink' ) )
				if not refid:
					pass

				# [1:] to ignore leading '#' in reference
				path = '//*[@id="%s"]' % refid[1:]
				refnode = node.xpath( path )
				if refnode:
					x = float( node.get( 'x', '0' ) )
					y = float( node.get( 'y', '0' ) )
					# Note: the transform has already been applied
					if ( x != 0 ) or ( y != 0 ):
						matNew2 = composeTransform( matNew, parseTransform( 'translate(%f,%f)' % (x,y) ) )
					else:
						matNew2 = matNew
					v = node.get( 'visibility', v )
					self.recursivelyTraverseSvg( refnode, matNew2, parent_visibility=v )

			elif node.tag == addNS( 'path', 'svg' ):

				path_data = node.get( 'd')
				if path_data:
					self.addPathVertices( path_data, node, matNew )

			elif node.tag == addNS( 'rect', 'svg' ) or node.tag == 'rect':

				# Manually transform
				#
				#    <rect x="X" y="Y" width="W" height="H"/>
				#
				# into
				#
				#    <path d="MX,Y lW,0 l0,H l-W,0 z"/>
				#
				# I.e., explicitly draw three sides of the rectangle and the
				# fourth side implicitly

				# Create a path with the outline of the rectangle
				x = float( node.get( 'x' ) )
				y = float( node.get( 'y' ) )
				if ( not x ) or ( not y ):
					pass
				w = float( node.get( 'width', '0' ) )
				h = float( node.get( 'height', '0' ) )
				a = []
				a.append( ['M ', [x, y]] )
				a.append( [' l ', [w, 0]] )
				a.append( [' l ', [0, h]] )
				a.append( [' l ', [-w, 0]] )
				a.append( [' Z', []] )
				self.addPathVertices( simplepath.formatPath( a ), node, matNew )

			elif node.tag == addNS( 'line', 'svg' ) or node.tag == 'line':

				# Convert
				#
				#   <line x1="X1" y1="Y1" x2="X2" y2="Y2/>
				#
				# to
				#
				#   <path d="MX1,Y1 LX2,Y2"/>

				x1 = float( node.get( 'x1' ) )
				y1 = float( node.get( 'y1' ) )
				x2 = float( node.get( 'x2' ) )
				y2 = float( node.get( 'y2' ) )
				if ( not x1 ) or ( not y1 ) or ( not x2 ) or ( not y2 ):
					pass
				a = []
				a.append( ['M ', [x1, y1]] )
				a.append( [' L ', [x2, y2]] )
				self.addPathVertices( simplepath.formatPath( a ), node, matNew )

			elif node.tag == addNS( 'polyline', 'svg' ) or node.tag == 'polyline':

				# Convert
				#
				#  <polyline points="x1,y1 x2,y2 x3,y3 [...]"/>
				#
				# to
				#
				#   <path d="Mx1,y1 Lx2,y2 Lx3,y3 [...]"/>
				#
				# Note: we ignore polylines with no points

				pl = node.get( 'points', '' ).strip()
				if pl == '':
					pass

				pa = pl.split()
				d = "".join( ["M " + pa[i] if i == 0 else " L " + pa[i] for i in range( 0, len( pa ) )] )
				self.addPathVertices( d, node, matNew )

			elif node.tag == addNS( 'polygon', 'svg' ) or node.tag == 'polygon':

				# Convert
				#
				#  <polygon points="x1,y1 x2,y2 x3,y3 [...]"/>
				#
				# to
				#
				#   <path d="Mx1,y1 Lx2,y2 Lx3,y3 [...] Z"/>
				#
				# Note: we ignore polygons with no points

				pl = node.get( 'points', '' ).strip()
				if pl == '':
					pass

				pa = pl.split()
				d = "".join( ["M " + pa[i] if i == 0 else " L " + pa[i] for i in range( 0, len( pa ) )] )
				d += " Z"
				self.addPathVertices( d, node, matNew )

			elif node.tag == addNS( 'ellipse', 'svg' ) or \
				node.tag == 'ellipse' or \
				node.tag == addNS( 'circle', 'svg' ) or \
				node.tag == 'circle':

					# Convert circles and ellipses to a path with two 180 degree arcs.
					# In general (an ellipse), we convert
					#
					#   <ellipse rx="RX" ry="RY" cx="X" cy="Y"/>
					#
					# to
					#
					#   <path d="MX1,CY A RX,RY 0 1 0 X2,CY A RX,RY 0 1 0 X1,CY"/>
					#
					# where
					#
					#   X1 = CX - RX
					#   X2 = CX + RX
					#
					# Note: ellipses or circles with a radius attribute of value 0 are ignored
					print "filled circle", node.tag

					if node.tag == addNS( 'ellipse', 'svg' ) or node.tag == 'ellipse':
						rx = float( node.get( 'rx', '0' ) )
						ry = float( node.get( 'ry', '0' ) )
					else:
						rx = float( node.get( 'r', '0' ) )
						ry = rx
					if rx == 0 or ry == 0:
						pass

					cx = float( node.get( 'cx', '0' ) )
					cy = float( node.get( 'cy', '0' ) )
					x1 = cx - rx
					x2 = cx + rx
					d = 'M %f,%f ' % ( x1, cy ) + \
						'A %f,%f ' % ( rx, ry ) + \
						'0 1 0 %f,%f ' % ( x2, cy ) + \
						'A %f,%f ' % ( rx, ry ) + \
						'0 1 0 %f,%f' % ( x1, cy )
						
					self.addPathVertices( d, node, matNew )

			elif node.tag == addNS( 'pattern', 'svg' ) or node.tag == 'pattern':

				pass

			elif node.tag == addNS( 'metadata', 'svg' ) or node.tag == 'metadata':

				pass

			elif node.tag == addNS( 'defs', 'svg' ) or node.tag == 'defs':

				pass

			elif node.tag == addNS( 'namedview', 'sodipodi' ) or node.tag == 'namedview':

				pass

			elif node.tag == addNS( 'eggbot', 'svg' ) or node.tag == 'eggbot':

				pass

			elif node.tag == addNS( 'text', 'svg' ) or node.tag == 'text':

				print( 'Warning: unable to draw text, please convert it to a path first.' )

				pass

			elif not isinstance( node.tag, basestring ):

				pass

			else:

				print( 'Warning: unable to draw object <%s>, please convert it to a path first.' % node.tag )
				pass

	def joinFillsWithNode ( self, node, stroke_width, color, path ):

		'''
		Generate a SVG <path> element containing the path data "path".
		Then put this new <path> element into a <group> with the supplied
		node.  This means making a new <group> element and moving node
		under it with the new <path> as a sibling element.
		'''

		if ( not path ) or ( len( path ) == 0 ):
			return

		# Make a new SVG <group> element whose parent is the parent of node
		#parent = node.getparent()
		#was: if not parent:
		#if parent is None:
		#	parent = self.document.getroot()
		#g = etree.SubElement( parent, addNS( 'g', 'svg' ) )

		# Move node to be a child of this new <g> element
		#g.append( node )

		# Now make a <path> element which contains the hatches & is a child
		# of the new <g> element
		style = { 'stroke': color, 'fill': 'none', 'stroke-width': '%f' % stroke_width}
		line_attribs = { 'style':simplestyle.formatStyle( style ), 'd': path }
		tran = node.get( 'transform' )
		if ( tran != None ) and ( tran != '' ):
			line_attribs['transform'] = tran
		return line_attribs 

	def makeHatchGrid( self, angle, spacing, init=True ):

		'''
		Build a grid of hatch lines which encompasses the entire bounding
		box of the graphical elements we are to hatch.

		1. Figure out the bounding box for all of the graphical elements
		2. Pick a rectangle larger than that bounding box so that we can
		   later rotate the rectangle and still have it cover the bounding
		   box of the graphical elements.
		3. Center the rectangle of 2 on the origin (0, 0).
		4. Build the hatch line grid in this rectangle.
		5. Rotate the rectangle by the hatch angle.
		6. Translate the center of the rotated rectangle, (0, 0), to be
		   the center of the bounding box for the graphical elements.
		7. We now have a grid of hatch lines which overlay the graphical
		   elements and can now be intersected with those graphical elements.
		'''

		# If this is the first call, do some one time initializations
		# When generating cross hatches, we may be called more than once
		if init:
			self.getBoundingBox()
			self.grid = []

		# Determine the width and height of the bounding box containing
		# all the polygons to be hatched
		w = self.xmax - self.xmin
		h = self.ymax - self.ymin

		# Nice thing about rectangles is that the diameter of the circle
		# encompassing them is the length the rectangle's diagonal...
		r = math.sqrt ( w * w + h * h ) / 2.0

		# Length of a hatch line will be 2r
		# We wish to convert the number of pixels considered insignificant
		# to a unitless, fractional length of a hatch line.  The length of
		# a hatch line is 2r pixels.  So,
		self.minGap = self.options['minGap'] / ( 2.0 * r )

		# Now generate hatch lines within the square
		# centered at (0, 0) and with side length at least d

		# While we could generate these lines running back and forth,
		# that makes for weird behavior later when applying odd/even
		# rules AND there are nested polygons.  Instead, when we
		# generate the SVG <path> elements with the hatch line
		# segments, we can do the back and forth weaving.

		# Rotation information
		ca = math.cos( math.radians( 90 - angle ) )
		sa = math.sin( math.radians( 90 - angle ) )

		# Translation information
		cx = self.xmin + ( w / 2 )
		cy = self.ymin + ( h / 2 )

		# Since the spacing may be fractional (e.g., 6.5), we
		# don't try to use range() or other integer iterator
		spacing = float( abs( spacing ) )
		i = -r
		while i <= r:
			# Line starts at (i, -r) and goes to (i, +r)
			x1 = cx + ( i * ca ) + ( r * sa ) #  i * ca - (-r) * sa
			y1 = cy + ( i * sa ) - ( r * ca ) #  i * sa + (-r) * ca
			x2 = cx + ( i * ca ) - ( r * sa ) #  i * ca - (+r) * sa
			y2 = cy + ( i * sa ) + ( r * ca ) #  i * sa + (+r) * ca
			i += spacing
			# Remove any potential hatch lines which are entirely
			# outside of the bounding box
			if (( x1 < self.xmin ) and ( x2 < self.xmin )) or \
				(( x1 > self.xmax ) and ( x2 > self.xmax )):
				continue
			if (( y1 < self.ymin ) and ( y2 < self.ymin )) or \
				(( y1 > self.ymax ) and ( y2 > self.ymax )):
				continue
			self.grid.append( ( x1, y1, x2, y2 ) )



	def effect( self , hatchSpacing = 10, crossHatch = False, hatchAngle = 0, tolerance = 20, minGap = 5, ):
		'''
			crossHatch : Generate a cross hatch pattern 
			hatchAngle : Angle of inclination for hatch lines 
			hatchSpacing : Spacing between hatch lines
			tolerance : Allowed deviation from original paths
			minGap : Minimum length of hatch segments and gaps
		'''
		self.options = {}
		self.options['crossHatch'] = crossHatch 
		self.options['hatchAngle'] = hatchAngle 
		self.options['hatchSpacing'] = hatchSpacing * 90/25.4
		self.options['tolerance'] = tolerance 
		self.options['minGap'] = minGap

		# Viewbox handling
		self.handleViewBox()

		# Traverse the selected nodes
		self.recursivelyTraverseSvg( self.nodes, self.docTransform )

		# Build a grid of possible hatch lines
		self.makeHatchGrid( float( self.options['hatchAngle'] ),
				float( self.options['hatchSpacing'] ), True )
		if self.options['crossHatch']:
			self.makeHatchGrid( float( self.options['hatchAngle'] + 90.0 ),
				float( self.options['hatchSpacing'] ), False )

		# Now loop over our hatch lines looking for intersections
		for h in self.grid:
			interstices( (h[0], h[1]), (h[2], h[3]), self.paths, self.hatches, self.minGap )

		# Target stroke width will be (doc width + doc height) / 2 / 1000
		# stroke_width_target = ( self.docHeight + self.docWidth ) / 2000
		stroke_width_target = 1

		# Each hatch line stroke will be within an SVG object which may
		# be subject to transforms.  So, on an object by object basis,
		# we need to transform our target width to a width suitable
		# for that object (so that after the object and its hatches are
		# transformed, the result has the desired width).

		# To aid in the process, we use a diagonal line segment of length
		# stroke_width_target.  We then run this segment through an object's
		# inverse transform and see what the resulting length of the inversely
		# transformed segment is.  We could, alternatively, look at the
		# x and y scaling factors in the transform and average them.
		s = stroke_width_target / math.sqrt( 2 )

		# Now, dump the hatch fills sorted by which document element
		# they correspond to.  This is made easy by the fact that we
		# saved the information and used each element's lxml.etree node
		# pointer as the dictionary key under which to save the hatch
		# fills for that node.

		fillings = []
		for key in self.hatches:
			path = ''
			direction = True
			if self.transforms.has_key( key ):
				transform = inverseTransform( self.transforms[key] )
				# Determine the scaled stroke width for a hatch line
				# We produce a line segment of unit length, transform
				# its endpoints and then determine the length of the
				# resulting line segment.
				pt1 = [0, 0]
				pt2 = [s, s]
				simpletransform.applyTransformToPoint( transform, pt1 )
				simpletransform.applyTransformToPoint( transform, pt2 )
				dx = pt2[0] - pt1[0]
				dy = pt2[1] - pt1[1]
				stroke_width = math.sqrt( dx * dx + dy * dy )
			else:
				transform = None
				stroke_width = float( 1.0 )
			for segment in self.hatches[key]:
				style = simplestyle.parseStyle(key.get("style"))
				color = style["fill"]
				if len( segment ) < 2:
					continue
				pt1 = segment[0]
				pt2 = segment[1]
				# Okay, we're going to put these hatch lines into the same
				# group as the element they hatch.  That element is down
				# some chain of SVG elements, some of which may have
				# transforms attached.  But, our hatch lines have been
				# computed assuming that those transforms have already
				# been applied (since we had to apply them so as to know
				# where this element is on the page relative to other
				# elements and their transforms).  So, we need to invert
				# the transforms for this element and then either apply
				# that inverse transform here and now or set it in a
				# transform attribute of the <path> element.  Having it
				# set in the path element seems a bit counterintuitive
				# after the fact (i.e., what's this tranform here for?).
				# So, we compute the inverse transform and apply it here.
				if transform != None:
					simpletransform.applyTransformToPoint( transform, pt1 )
					simpletransform.applyTransformToPoint( transform, pt2 )
				# Now generate the path data for the <path>
				if direction:
					# Go this direction
					path += 'M %f,%f l %f,%f ' % \
						( pt1[0], pt1[1], pt2[0] - pt1[0], pt2[1] - pt1[1] )
				else:
					# Or go this direction
					path += 'M %f,%f l %f,%f ' % \
						( pt2[0], pt2[1], pt1[0] - pt2[0], pt1[1] - pt2[1] )
				direction = not direction
			fillment = self.joinFillsWithNode( key, stroke_width, color, path[:-1] )
			fillment['id'] = key.get('id')
			fillings.append(fillment)

		#inkex.errormsg("Elapsed CPU time was %f" % (time.clock()-self.t0))
		return fillings

NSS = {
u'sodipodi' :u'http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd',
u'cc'       :u'http://creativecommons.org/ns#',
u'ccOLD'    :u'http://web.resource.org/cc/',
u'svg'      :u'http://www.w3.org/2000/svg',
u'dc'       :u'http://purl.org/dc/elements/1.1/',
u'rdf'      :u'http://www.w3.org/1999/02/22-rdf-syntax-ns#',
u'inkscape' :u'http://www.inkscape.org/namespaces/inkscape',
u'xlink'    :u'http://www.w3.org/1999/xlink',
u'xml'      :u'http://www.w3.org/XML/1998/namespace'
}

def check_inkbool(option, opt, value):
    if str(value).capitalize() == 'True':
        return True
    elif str(value).capitalize() == 'False':
        return False
    else:
        raise optparse.OptionValueError("option %s: invalid inkbool value: %s" % (opt, value))

def addNS(tag, ns=None):
    val = tag
    if ns!=None and len(ns)>0 and NSS.has_key(ns) and len(tag)>0 and tag[0]!='{':
        val = "{%s}%s" % (NSS[ns], tag)
    return val