/usr/lib/python2.7/dist-packages/pdfrw/pagemerge.py is in python-pdfrw 0.4-1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 | # A part of pdfrw (https://github.com/pmaupin/pdfrw)
# Copyright (C) 2015 Patrick Maupin, Austin, Texas
# MIT license -- See LICENSE.txt for details
'''
This module contains code to edit pages. Sort of a canvas, I
suppose, but I wouldn't want to call it that and get people all
excited or anything.
No, this is just for doing basic things like merging/splitting
apart pages, watermarking, etc. All it does is allow converting
pages (or parts of pages) into Form XObject rectangles, and then
plopping those down on new or pre-existing pages.
'''
from .objects import PdfDict, PdfArray, PdfName
from .buildxobj import pagexobj, ViewInfo
NullInfo = ViewInfo()
class RectXObj(PdfDict):
''' This class facilitates doing positioning (moving and scaling)
of Form XObjects within their containing page, by modifying
the Form XObject's transformation matrix.
By default, this class keeps the aspect ratio locked. For
example, if your object is foo, you can write 'foo.w = 200',
and it will scale in both the x and y directions.
To unlock the aspect ration, you have to do a tiny bit of math
and call the scale function.
'''
def __init__(self, page, viewinfo=NullInfo, **kw):
''' The page is a page returned by PdfReader. It will be
turned into a cached Form XObject (so that multiple
rectangles can be extracted from it if desired), and then
another Form XObject will be built using it and the viewinfo
(which should be a ViewInfo class). The viewinfo includes
source coordinates (from the top/left) and rotation information.
Once the object has been built, its destination coordinates
may be examined and manipulated by using x, y, w, h, and
scale. The destination coordinates are in the normal
PDF programmatic system (starting at bottom left).
'''
if kw:
if viewinfo is not NullInfo:
raise ValueError("Cannot modify preexisting ViewInfo")
viewinfo = ViewInfo(**kw)
viewinfo.cacheable = False
base = pagexobj(page, viewinfo)
self.update(base)
self.indirect = True
self.stream = base.stream
private = self.private
private._rect = [base.x, base.y, base.w, base.h]
matrix = self.Matrix
if matrix is None:
matrix = self.Matrix = PdfArray((1, 0, 0, 1, 0, 0))
private._matrix = matrix # Lookup optimization
# Default to lower-left corner
self.x = 0
self.y = 0
@property
def x(self):
''' X location (from left) of object in points
'''
return self._rect[0]
@property
def y(self):
''' Y location (from bottom) of object in points
'''
return self._rect[1]
@property
def w(self):
''' Width of object in points
'''
return self._rect[2]
@property
def h(self):
''' Height of object in points
'''
return self._rect[3]
def __setattr__(self, name, value, next=PdfDict.__setattr__,
mine=set('x y w h'.split())):
''' The underlying __setitem__ won't let us use a property
setter, so we have to fake one.
'''
if name not in mine:
return next(self, name, value)
if name in 'xy':
r_index, m_index = (0, 4) if name == 'x' else (1, 5)
self._rect[r_index], old = value, self._rect[r_index]
self._matrix[m_index] += value - old
else:
index = 2 + (value == 'h')
self.scale(value / self._rect[index])
def scale(self, x_scale, y_scale=None):
''' Current scaling deals properly with things that
have been rotated in 90 degree increments
(via the ViewMerge object given when instantiating).
'''
if y_scale is None:
y_scale = x_scale
x, y, w, h = rect = self._rect
ao, bo, co, do, eo, fo = matrix = self._matrix
an = ao * x_scale
bn = bo * y_scale
cn = co * x_scale
dn = do * y_scale
en = x + (eo - x) * 1.0 * (an + cn) / (ao + co)
fn = y + (fo - y) * 1.0 * (bn + dn) / (bo + do)
matrix[:] = an, bn, cn, dn, en, fn
rect[:] = x, y, w * x_scale, h * y_scale
@property
def box(self):
''' Return the bounding box for the object
'''
x, y, w, h = self._rect
return PdfArray([x, y, x + w, y + h])
class PageMerge(list):
''' A PageMerge object can have 0 or 1 underlying pages
(that get edited with the results of the merge)
and 0-n RectXObjs that can be applied before or
after the underlying page.
'''
page = None
mbox = None
cbox = None
resources = None
rotate = None
contents = None
def __init__(self, page=None):
if page is not None:
self.setpage(page)
def setpage(self, page):
if page.Type != PdfName.Page:
raise TypeError("Expected page")
self.append(None) # Placeholder
self.page = page
inheritable = page.inheritable
self.mbox = inheritable.MediaBox
self.cbox = inheritable.CropBox
self.resources = inheritable.Resources
self.rotate = inheritable.Rotate
self.contents = page.Contents
def __add__(self, other):
if isinstance(other, dict):
other = [other]
for other in other:
self.add(other)
return self
def add(self, obj, prepend=False, **kw):
if kw:
obj = RectXObj(obj, **kw)
elif obj.Type == PdfName.Page:
obj = RectXObj(obj)
if prepend:
self.insert(0, obj)
else:
self.append(obj)
return self
def render(self):
def do_xobjs(xobj_list, restore_first=False):
content = ['Q'] if restore_first else []
for obj in xobj_list:
index = PdfName('pdfrw_%d' % (key_offset + len(xobjs)))
if xobjs.setdefault(index, obj) is not obj:
raise KeyError("XObj key %s already in use" % index)
content.append('%s Do' % index)
return PdfDict(indirect=True, stream='\n'.join(content))
mbox = self.mbox
cbox = self.cbox
page = self.page
old_contents = self.contents
resources = self.resources or PdfDict()
key_offset = 0
xobjs = resources.XObject
if xobjs is None:
xobjs = resources.XObject = PdfDict()
else:
allkeys = xobjs.keys()
if allkeys:
keys = (x for x in allkeys if x.startswith('/pdfrw_'))
keys = (x for x in keys if x[7:].isdigit())
keys = sorted(keys, key=lambda x: int(x[7:]))
key_offset = (int(keys[-1][7:]) + 1) if keys else 0
key_offset -= len(allkeys)
if old_contents is None:
new_contents = do_xobjs(self)
else:
isdict = isinstance(old_contents, PdfDict)
old_contents = [old_contents] if isdict else old_contents
new_contents = PdfArray()
index = self.index(None)
if index:
new_contents.append(do_xobjs(self[:index]))
index += 1
if index < len(self):
# There are elements to add after the original page contents,
# so push the graphics state to the stack. Restored below.
new_contents.append(PdfDict(indirect=True, stream='q'))
new_contents.extend(old_contents)
if index < len(self):
# Restore graphics state and add other elements.
new_contents.append(do_xobjs(self[index:], restore_first=True))
if mbox is None:
cbox = None
mbox = self.xobj_box
mbox[0] = min(0, mbox[0])
mbox[1] = min(0, mbox[1])
page = PdfDict(indirect=True) if page is None else page
page.Type = PdfName.Page
page.Resources = resources
page.MediaBox = mbox
page.CropBox = cbox
page.Rotate = self.rotate
page.Contents = new_contents
return page
@property
def xobj_box(self):
''' Return the smallest box that encloses every object
in the list.
'''
a, b, c, d = zip(*(xobj.box for xobj in self))
return PdfArray((min(a), min(b), max(c), max(d)))
|