def init(self, page: fitz.Page) -> Layout:
'''Initialize layout object.'''
# Layout object based on raw dict
raw_layout = page.getText('rawdict')
self._layout = Layout(raw_layout)
# get rectangle shapes from page source:
# these shapes are generally converted from docx, e.g. highlight, underline,
# which are different from PDF comments like highlight, rectangle.
if not page._isWrapped:
page._wrapContents()
# transformation matrix from PDF to PyMuPDF
M = page.transformationMatrix # PyMuPDF>=1.17.0
for xref in page.getContents():
page_content = self._doc_pdf._getXrefStream(xref).decode(
encoding="ISO-8859-1")
self._layout.rects.from_stream(page_content, M)
# get annotations(comment shapes) from PDF page: consider highlight, underline,
# strike-through-line only.
annots = page.annots()
self._layout.rects.from_annotations(annots)
# plot raw layout
if self.debug_mode:
# new section for current pdf page
new_page_section(self._doc_debug, self._layout.width,
self._layout.height, f'Page {page.number}')
# initial layout
self._layout.plot(self._doc_debug,
'Original Text Blocks',
key=PlotControl.LAYOUT)
self._layout.plot(self._doc_debug,
'Original Rectangle Shapes',
key=PlotControl.SHAPE)
return self._layout
def paths_from_annotations(page: fitz.Page):
''' Get shapes, e.g. Line, Square, Highlight, from annotations(comment shapes) in PDF page.
---
Args:
- page: fitz.Page, current page
There are stroke and fill properties for each shape, representing border and filling area respectively.
So, a square annotation with both stroke and fill will be converted to five rectangles here:
four borders and one filling area.
read more:
- https://pymupdf.readthedocs.io/en/latest/annot.html
- https://pymupdf.readthedocs.io/en/latest/vars.html#annotation-types
'''
res = []
for annot in page.annots():
# annot type, e.g. (8, 'Highlight')
key = annot.type[0]
# color, e.g. {'stroke': [1.0, 1.0, 0.0], 'fill': []}
c = annot.colors
sc = utils.RGB_value(c['stroke']) if c['stroke'] else None
fc = utils.RGB_value(c['fill']) if c['fill'] else None
# width
w = annot.border.get('width', 1.0) # width=-1 if not set
w = 1.0 if w == -1 else w # 1.0 by default
# bbox
rect = annot.rect
# considering the contributions to text format and table borders,
# only the following types are processed.
# PDF_ANNOT_LINE 3
# PDF_ANNOT_SQUARE 4
# PDF_ANNOT_HIGHLIGHT 8
# PDF_ANNOT_UNDERLINE 9
# PDF_ANNOT_STRIKEOUT 11
# Line: a space of 1.5*w around each border
#
# +----------------------------+
# | space |
# | +--------------+ |
# | | border | 1.5w |
# | +--------------+ |
# | 1.5w |
# +----------------------------+
#
if key == 3:
x0 = rect.x0 + 1.5 * w
x1 = rect.x1 - 1.5 * w
y0 = y1 = (rect.y0 + rect.y1) / 2.0
path = _add_stroke_line((x0, y0), (x1, y1), sc, w)
res.append(path)
# Square: a space of 0.5*w around eah border
# border rects and filling rects are to be extracted from original square
#
# +------------------------------------------+
# | space |
# | +----------------------------+ |
# | | border | |
# | | +--------------+ | |
# | | fill | w | 0.5w |
# | | +--------------+ | |
# | | w | |
# | +----------------------------+ |
# | 0.5w |
# +------------------------------------------+
#
elif key == 4:
# stroke rectangles
if not sc is None:
x0, y0 = rect.x0 + w, rect.y0 + w
x1, y1 = rect.x1 - w, rect.y1 - w
path = _add_stroke_rect((x0, y0), (x1, y1), sc, w)
res.append(path)
# fill rectangle
if not fc is None:
d = 1.5 * w
x0, y0 = rect.x0 + d, rect.y0 + d
x1, y1 = rect.x1 - d, rect.y1 - d
path = _add_fill_rect((x0, y0), (x1, y1), fc)
res.append(path)
# highlight, underline, strikethrough: on space
# For these shapes, `annot.rect` is a combination of all sub-highlights, especially
# the highlight is continuous in multi-lines.
# So, `annot.vertices` should be used here, i.e. vertices marked with `+` below.
# +------------------------+
# +------------------------+
# +-----------+
# +-----------+
# NOTE: Though underline and strikethrough are just lines, the affected areas are same as
# highlights, as illustrated above.
#
# https://github.com/pymupdf/PyMuPDF/issues/318
#
elif key in (8, 9, 11):
points = annot.vertices
for i in range(int(len(points) / 4.0)): # four points in a group
# highlight: whole bbox
if key == 8:
x0, y0 = points[4 * i]
x1, y1 = points[4 * i + 3]
# NOTE: this indded a stroke for PyMuPDF -> no fill color but stroke color !!
path = _add_fill_rect((x0, y0), (x1, y1), sc)
res.append(path)
else:
# underline: bottom edge
if key == 9:
start, end = points[4 * i + 2], points[4 * i + 3]
# strikethrough: average of top and bottom edge
else:
x0, x1 = points[4 * i][0], points[4 * i + 1][0]
y_ = (points[4 * i][1] + points[4 * i + 2][1]) / 2.0
start = x0, y_
end = x1, y_
path = _add_stroke_line(start, end, sc, w)
res.append(path)
return res