annotate src/utils.py @ 45:b74429b0f5c4 v1.26.5+1

+++++ v1.26.5+1
author Franz Glasner <fzglas.hg@dom66.de>
date Sat, 11 Oct 2025 17:17:13 +0200
parents a6bc019ac0b2
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1 # ------------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
2 # Copyright 2020-2022, Harald Lieder, mailto:harald.lieder@outlook.com
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
3 # License: GNU AFFERO GPL 3.0, https://www.gnu.org/licenses/agpl-3.0.html
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
4 #
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
5 # Part of "PyMuPDF", a Python binding for "MuPDF" (http://mupdf.com), a
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
6 # lightweight PDF, XPS, and E-book viewer, renderer and toolkit which is
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
7 # maintained and developed by Artifex Software, Inc. https://artifex.com.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
8 # ------------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
9 import math
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
10 import typing
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
11 import weakref
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
12
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
13 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
14 from . import pymupdf
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
15 except Exception:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
16 import pymupdf
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
17 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
18 from . import mupdf
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
19 except Exception:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
20 import mupdf
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
21
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
22 _format_g = pymupdf.format_g
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
23
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
24 g_exceptions_verbose = pymupdf.g_exceptions_verbose
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
25
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
26 point_like = "point_like"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
27 rect_like = "rect_like"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
28 matrix_like = "matrix_like"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
29 quad_like = "quad_like"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
30
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
31 # ByteString is gone from typing in 3.14.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
32 # collections.abc.Buffer available from 3.12 only
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
33 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
34 ByteString = typing.ByteString
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
35 except AttributeError:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
36 # pylint: disable=unsupported-binary-operation
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
37 ByteString = bytes | bytearray | memoryview
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
38
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
39 AnyType = typing.Any
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
40 OptInt = typing.Union[int, None]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
41 OptFloat = typing.Optional[float]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
42 OptStr = typing.Optional[str]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
43 OptDict = typing.Optional[dict]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
44 OptBytes = typing.Optional[ByteString]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
45 OptSeq = typing.Optional[typing.Sequence]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
46
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
47 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
48 This is a collection of functions to extend PyMupdf.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
49 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
50
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
51
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
52 def get_text_blocks(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
53 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
54 clip: rect_like = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
55 flags: OptInt = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
56 textpage: pymupdf.TextPage = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
57 sort: bool = False,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
58 ) -> list:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
59 """Return the text blocks on a page.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
60
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
61 Notes:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
62 Lines in a block are concatenated with line breaks.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
63 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
64 flags: (int) control the amount of data parsed into the textpage.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
65 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
66 A list of the blocks. Each item contains the containing rectangle
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
67 coordinates, text lines, running block number and block type.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
68 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
69 pymupdf.CheckParent(page)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
70 if flags is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
71 flags = pymupdf.TEXTFLAGS_BLOCKS
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
72 tp = textpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
73 if tp is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
74 tp = page.get_textpage(clip=clip, flags=flags)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
75 elif getattr(tp, "parent") != page:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
76 raise ValueError("not a textpage of this page")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
77
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
78 blocks = tp.extractBLOCKS()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
79 if textpage is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
80 del tp
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
81 if sort:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
82 blocks.sort(key=lambda b: (b[3], b[0]))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
83 return blocks
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
84
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
85
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
86 def get_text_words(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
87 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
88 clip: rect_like = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
89 flags: OptInt = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
90 textpage: pymupdf.TextPage = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
91 sort: bool = False,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
92 delimiters=None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
93 tolerance=3,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
94 ) -> list:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
95 """Return the text words as a list with the bbox for each word.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
96
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
97 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
98 page: pymupdf.Page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
99 clip: (rect-like) area on page to consider
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
100 flags: (int) control the amount of data parsed into the textpage.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
101 textpage: (pymupdf.TextPage) either passed-in or None.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
102 sort: (bool) sort the words in reading sequence.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
103 delimiters: (str,list) characters to use as word delimiters.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
104 tolerance: (float) consider words to be part of the same line if
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
105 top or bottom coordinate are not larger than this. Relevant
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
106 only if sort=True.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
107
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
108 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
109 Word tuples (x0, y0, x1, y1, "word", bno, lno, wno).
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
110 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
111
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
112 def sort_words(words):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
113 """Sort words line-wise, forgiving small deviations."""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
114 words.sort(key=lambda w: (w[3], w[0]))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
115 nwords = [] # final word list
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
116 line = [words[0]] # collects words roughly in same line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
117 lrect = pymupdf.Rect(words[0][:4]) # start the line rectangle
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
118 for w in words[1:]:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
119 wrect = pymupdf.Rect(w[:4])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
120 if (
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
121 abs(wrect.y0 - lrect.y0) <= tolerance
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
122 or abs(wrect.y1 - lrect.y1) <= tolerance
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
123 ):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
124 line.append(w)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
125 lrect |= wrect
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
126 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
127 line.sort(key=lambda w: w[0]) # sort words in line l-t-r
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
128 nwords.extend(line) # append to final words list
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
129 line = [w] # start next line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
130 lrect = wrect # start next line rect
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
131
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
132 line.sort(key=lambda w: w[0]) # sort words in line l-t-r
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
133 nwords.extend(line) # append to final words list
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
134
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
135 return nwords
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
136
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
137 pymupdf.CheckParent(page)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
138 if flags is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
139 flags = pymupdf.TEXTFLAGS_WORDS
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
140 tp = textpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
141 if tp is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
142 tp = page.get_textpage(clip=clip, flags=flags)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
143 elif getattr(tp, "parent") != page:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
144 raise ValueError("not a textpage of this page")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
145
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
146 words = tp.extractWORDS(delimiters)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
147
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
148 # if textpage was given, we subselect the words in clip
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
149 if textpage is not None and clip is not None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
150 # sub-select words contained in clip
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
151 clip = pymupdf.Rect(clip)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
152 words = [
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
153 w for w in words if abs(clip & w[:4]) >= 0.5 * abs(pymupdf.Rect(w[:4]))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
154 ]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
155
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
156 if textpage is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
157 del tp
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
158 if words and sort:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
159 # advanced sort if any words found
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
160 words = sort_words(words)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
161
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
162 return words
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
163
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
164
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
165 def get_sorted_text(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
166 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
167 clip: rect_like = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
168 flags: OptInt = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
169 textpage: pymupdf.TextPage = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
170 tolerance=3,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
171 ) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
172 """Extract plain text avoiding unacceptable line breaks.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
173
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
174 Text contained in clip will be sorted in reading sequence. Some effort
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
175 is also spent to simulate layout vertically and horizontally.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
176
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
177 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
178 page: pymupdf.Page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
179 clip: (rect-like) only consider text inside
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
180 flags: (int) text extraction flags
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
181 textpage: pymupdf.TextPage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
182 tolerance: (float) consider words to be on the same line if their top
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
183 or bottom coordinates do not differ more than this.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
184
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
185 Notes:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
186 If a TextPage is provided, all text is checked for being inside clip
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
187 with at least 50% of its bbox.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
188 This allows to use some "global" TextPage in conjunction with sub-
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
189 selecting words in parts of the defined TextPage rectangle.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
190
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
191 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
192 A text string in reading sequence. Left indentation of each line,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
193 inter-line and inter-word distances strive to reflect the layout.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
194 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
195
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
196 def line_text(clip, line):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
197 """Create the string of one text line.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
198
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
199 We are trying to simulate some horizontal layout here, too.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
200
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
201 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
202 clip: (pymupdf.Rect) the area from which all text is being read.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
203 line: (list) word tuples (rect, text) contained in the line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
204 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
205 Text in this line. Generated from words in 'line'. Distance from
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
206 predecessor is translated to multiple spaces, thus simulating
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
207 text indentations and large horizontal distances.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
208 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
209 line.sort(key=lambda w: w[0].x0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
210 ltext = "" # text in the line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
211 x1 = clip.x0 # end coordinate of ltext
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
212 lrect = pymupdf.EMPTY_RECT() # bbox of this line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
213 for r, t in line:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
214 lrect |= r # update line bbox
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
215 # convert distance to previous word to multiple spaces
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
216 dist = max(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
217 int(round((r.x0 - x1) / r.width * len(t))),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
218 0 if (x1 == clip.x0 or r.x0 <= x1) else 1,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
219 ) # number of space characters
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
220
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
221 ltext += " " * dist + t # append word string
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
222 x1 = r.x1 # update new end position
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
223 return ltext
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
224
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
225 # Extract words in correct sequence first.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
226 words = [
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
227 (pymupdf.Rect(w[:4]), w[4])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
228 for w in get_text_words(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
229 page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
230 clip=clip,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
231 flags=flags,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
232 textpage=textpage,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
233 sort=True,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
234 tolerance=tolerance,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
235 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
236 ]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
237
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
238 if not words: # no text present
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
239 return ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
240 totalbox = pymupdf.EMPTY_RECT() # area covering all text
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
241 for wr, text in words:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
242 totalbox |= wr
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
243
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
244 lines = [] # list of reconstituted lines
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
245 line = [words[0]] # current line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
246 lrect = words[0][0] # the line's rectangle
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
247
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
248 # walk through the words
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
249 for wr, text in words[1:]: # start with second word
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
250 w0r, _ = line[-1] # read previous word in current line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
251
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
252 # if this word matches top or bottom of the line, append it
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
253 if abs(lrect.y0 - wr.y0) <= tolerance or abs(lrect.y1 - wr.y1) <= tolerance:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
254 line.append((wr, text))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
255 lrect |= wr
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
256 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
257 # output current line and re-initialize
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
258 ltext = line_text(totalbox, line)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
259 lines.append((lrect, ltext))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
260 line = [(wr, text)]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
261 lrect = wr
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
262
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
263 # also append unfinished last line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
264 ltext = line_text(totalbox, line)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
265 lines.append((lrect, ltext))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
266
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
267 # sort all lines vertically
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
268 lines.sort(key=lambda l: (l[0].y1))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
269
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
270 text = lines[0][1] # text of first line
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
271 y1 = lines[0][0].y1 # its bottom coordinate
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
272 for lrect, ltext in lines[1:]:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
273 distance = min(int(round((lrect.y0 - y1) / lrect.height)), 5)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
274 breaks = "\n" * (distance + 1)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
275 text += breaks + ltext
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
276 y1 = lrect.y1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
277
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
278 # return text in clip
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
279 return text
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
280
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
281
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
282 def get_textbox(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
283 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
284 rect: rect_like,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
285 textpage: pymupdf.TextPage = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
286 ) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
287 tp = textpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
288 if tp is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
289 tp = page.get_textpage()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
290 elif getattr(tp, "parent") != page:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
291 raise ValueError("not a textpage of this page")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
292 rc = tp.extractTextbox(rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
293 if textpage is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
294 del tp
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
295 return rc
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
296
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
297
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
298 def get_text_selection(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
299 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
300 p1: point_like,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
301 p2: point_like,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
302 clip: rect_like = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
303 textpage: pymupdf.TextPage = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
304 ):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
305 pymupdf.CheckParent(page)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
306 tp = textpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
307 if tp is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
308 tp = page.get_textpage(clip=clip, flags=pymupdf.TEXT_DEHYPHENATE)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
309 elif getattr(tp, "parent") != page:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
310 raise ValueError("not a textpage of this page")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
311 rc = tp.extractSelection(p1, p2)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
312 if textpage is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
313 del tp
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
314 return rc
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
315
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
316
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
317 def get_textpage_ocr(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
318 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
319 flags: int = 0,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
320 language: str = "eng",
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
321 dpi: int = 72,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
322 full: bool = False,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
323 tessdata: str = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
324 ) -> pymupdf.TextPage:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
325 """Create a Textpage from combined results of normal and OCR text parsing.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
326
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
327 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
328 flags: (int) control content becoming part of the result.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
329 language: (str) specify expected language(s). Default is "eng" (English).
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
330 dpi: (int) resolution in dpi, default 72.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
331 full: (bool) whether to OCR the full page image, or only its images (default)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
332 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
333 pymupdf.CheckParent(page)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
334 tessdata = pymupdf.get_tessdata(tessdata)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
335
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
336 def full_ocr(page, dpi, language, flags):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
337 zoom = dpi / 72
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
338 mat = pymupdf.Matrix(zoom, zoom)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
339 pix = page.get_pixmap(matrix=mat)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
340 ocr_pdf = pymupdf.Document(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
341 "pdf",
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
342 pix.pdfocr_tobytes(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
343 compress=False,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
344 language=language,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
345 tessdata=tessdata,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
346 ),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
347 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
348 ocr_page = ocr_pdf.load_page(0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
349 unzoom = page.rect.width / ocr_page.rect.width
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
350 ctm = pymupdf.Matrix(unzoom, unzoom) * page.derotation_matrix
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
351 tpage = ocr_page.get_textpage(flags=flags, matrix=ctm)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
352 ocr_pdf.close()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
353 pix = None
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
354 tpage.parent = weakref.proxy(page)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
355 return tpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
356
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
357 # if OCR for the full page, OCR its pixmap @ desired dpi
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
358 if full:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
359 return full_ocr(page, dpi, language, flags)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
360
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
361 # For partial OCR, make a normal textpage, then extend it with text that
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
362 # is OCRed from each image.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
363 # Because of this, we need the images flag bit set ON.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
364 tpage = page.get_textpage(flags=flags)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
365 for block in page.get_text("dict", flags=pymupdf.TEXT_PRESERVE_IMAGES)["blocks"]:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
366 if block["type"] != 1: # only look at images
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
367 continue
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
368 bbox = pymupdf.Rect(block["bbox"])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
369 if bbox.width <= 3 or bbox.height <= 3: # ignore tiny stuff
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
370 continue
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
371 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
372 pix = pymupdf.Pixmap(block["image"]) # get image pixmap
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
373 if pix.n - pix.alpha != 3: # we need to convert this to RGB!
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
374 pix = pymupdf.Pixmap(pymupdf.csRGB, pix)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
375 if pix.alpha: # must remove alpha channel
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
376 pix = pymupdf.Pixmap(pix, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
377 imgdoc = pymupdf.Document(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
378 "pdf",
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
379 pix.pdfocr_tobytes(language=language, tessdata=tessdata),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
380 ) # pdf with OCRed page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
381 imgpage = imgdoc.load_page(0) # read image as a page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
382 pix = None
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
383 # compute matrix to transform coordinates back to that of 'page'
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
384 imgrect = imgpage.rect # page size of image PDF
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
385 shrink = pymupdf.Matrix(1 / imgrect.width, 1 / imgrect.height)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
386 mat = shrink * block["transform"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
387 imgpage.extend_textpage(tpage, flags=0, matrix=mat)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
388 imgdoc.close()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
389 except (RuntimeError, mupdf.FzErrorBase):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
390 if 0 and g_exceptions_verbose:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
391 # Don't show exception info here because it can happen in
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
392 # normal operation (see test_3842b).
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
393 pymupdf.exception_info()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
394 tpage = None
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
395 pymupdf.message("Falling back to full page OCR")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
396 return full_ocr(page, dpi, language, flags)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
397
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
398 return tpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
399
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
400
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
401 def get_text(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
402 page: pymupdf.Page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
403 option: str = "text",
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
404 *,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
405 clip: rect_like = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
406 flags: OptInt = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
407 textpage: pymupdf.TextPage = None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
408 sort: bool = False,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
409 delimiters=None,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
410 tolerance=3,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
411 ):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
412 """Extract text from a page or an annotation.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
413
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
414 This is a unifying wrapper for various methods of the pymupdf.TextPage class.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
415
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
416 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
417 option: (str) text, words, blocks, html, dict, json, rawdict, xhtml or xml.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
418 clip: (rect-like) restrict output to this area.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
419 flags: bit switches to e.g. exclude images or decompose ligatures.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
420 textpage: reuse this pymupdf.TextPage and make no new one. If specified,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
421 'flags' and 'clip' are ignored.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
422
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
423 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
424 the output of methods get_text_words / get_text_blocks or pymupdf.TextPage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
425 methods extractText, extractHTML, extractDICT, extractJSON, extractRAWDICT,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
426 extractXHTML or etractXML respectively.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
427 Default and misspelling choice is "text".
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
428 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
429 formats = {
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
430 "text": pymupdf.TEXTFLAGS_TEXT,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
431 "html": pymupdf.TEXTFLAGS_HTML,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
432 "json": pymupdf.TEXTFLAGS_DICT,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
433 "rawjson": pymupdf.TEXTFLAGS_RAWDICT,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
434 "xml": pymupdf.TEXTFLAGS_XML,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
435 "xhtml": pymupdf.TEXTFLAGS_XHTML,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
436 "dict": pymupdf.TEXTFLAGS_DICT,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
437 "rawdict": pymupdf.TEXTFLAGS_RAWDICT,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
438 "words": pymupdf.TEXTFLAGS_WORDS,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
439 "blocks": pymupdf.TEXTFLAGS_BLOCKS,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
440 }
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
441 option = option.lower()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
442 assert option in formats
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
443 if option not in formats:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
444 option = "text"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
445 if flags is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
446 flags = formats[option]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
447
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
448 if option == "words":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
449 return get_text_words(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
450 page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
451 clip=clip,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
452 flags=flags,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
453 textpage=textpage,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
454 sort=sort,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
455 delimiters=delimiters,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
456 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
457 if option == "blocks":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
458 return get_text_blocks(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
459 page, clip=clip, flags=flags, textpage=textpage, sort=sort
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
460 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
461
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
462 if option == "text" and sort:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
463 return get_sorted_text(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
464 page,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
465 clip=clip,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
466 flags=flags,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
467 textpage=textpage,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
468 tolerance=tolerance,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
469 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
470
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
471 pymupdf.CheckParent(page)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
472 cb = None
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
473 if option in ("html", "xml", "xhtml"): # no clipping for MuPDF functions
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
474 clip = page.cropbox
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
475 if clip is not None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
476 clip = pymupdf.Rect(clip)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
477 cb = None
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
478 elif type(page) is pymupdf.Page:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
479 cb = page.cropbox
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
480 # pymupdf.TextPage with or without images
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
481 tp = textpage
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
482 #pymupdf.exception_info()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
483 if tp is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
484 tp = page.get_textpage(clip=clip, flags=flags)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
485 elif getattr(tp, "parent") != page:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
486 raise ValueError("not a textpage of this page")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
487 #pymupdf.log( '{option=}')
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
488 if option == "json":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
489 t = tp.extractJSON(cb=cb, sort=sort)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
490 elif option == "rawjson":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
491 t = tp.extractRAWJSON(cb=cb, sort=sort)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
492 elif option == "dict":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
493 t = tp.extractDICT(cb=cb, sort=sort)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
494 elif option == "rawdict":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
495 t = tp.extractRAWDICT(cb=cb, sort=sort)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
496 elif option == "html":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
497 t = tp.extractHTML()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
498 elif option == "xml":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
499 t = tp.extractXML()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
500 elif option == "xhtml":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
501 t = tp.extractXHTML()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
502 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
503 t = tp.extractText(sort=sort)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
504
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
505 if textpage is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
506 del tp
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
507 return t
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
508
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
509
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
510 def getLinkDict(ln, document=None) -> dict:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
511 if isinstance(ln, pymupdf.Outline):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
512 dest = ln.destination(document)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
513 elif isinstance(ln, pymupdf.Link):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
514 dest = ln.dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
515 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
516 assert 0, f'Unexpected {type(ln)=}.'
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
517 nl = {"kind": dest.kind, "xref": 0}
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
518 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
519 if hasattr(ln, 'rect'):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
520 nl["from"] = ln.rect
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
521 except Exception:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
522 # This seems to happen quite often in PyMuPDF/tests.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
523 if g_exceptions_verbose >= 2: pymupdf.exception_info()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
524 pass
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
525 pnt = pymupdf.Point(0, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
526 if dest.flags & pymupdf.LINK_FLAG_L_VALID:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
527 pnt.x = dest.lt.x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
528 if dest.flags & pymupdf.LINK_FLAG_T_VALID:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
529 pnt.y = dest.lt.y
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
530
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
531 if dest.kind == pymupdf.LINK_URI:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
532 nl["uri"] = dest.uri
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
533
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
534 elif dest.kind == pymupdf.LINK_GOTO:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
535 nl["page"] = dest.page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
536 nl["to"] = pnt
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
537 if dest.flags & pymupdf.LINK_FLAG_R_IS_ZOOM:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
538 nl["zoom"] = dest.rb.x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
539 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
540 nl["zoom"] = 0.0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
541
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
542 elif dest.kind == pymupdf.LINK_GOTOR:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
543 nl["file"] = dest.file_spec.replace("\\", "/")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
544 nl["page"] = dest.page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
545 if dest.page < 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
546 nl["to"] = dest.dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
547 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
548 nl["to"] = pnt
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
549 if dest.flags & pymupdf.LINK_FLAG_R_IS_ZOOM:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
550 nl["zoom"] = dest.rb.x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
551 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
552 nl["zoom"] = 0.0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
553
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
554 elif dest.kind == pymupdf.LINK_LAUNCH:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
555 nl["file"] = dest.file_spec.replace("\\", "/")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
556
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
557 elif dest.kind == pymupdf.LINK_NAMED:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
558 # The dicts should not have same key(s).
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
559 assert not (dest.named.keys() & nl.keys())
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
560 nl.update(dest.named)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
561 if 'to' in nl:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
562 nl['to'] = pymupdf.Point(nl['to'])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
563
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
564 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
565 nl["page"] = dest.page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
566 return nl
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
567
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
568
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
569 def getDestStr(xref: int, ddict: dict) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
570 """Calculate the PDF action string.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
571
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
572 Notes:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
573 Supports Link annotations and outline items (bookmarks).
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
574 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
575 if not ddict:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
576 return ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
577 str_goto = lambda a, b, c, d: f"/A<</S/GoTo/D[{a} 0 R/XYZ {_format_g((b, c, d))}]>>"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
578 str_gotor1 = lambda a, b, c, d, e, f: f"/A<</S/GoToR/D[{a} /XYZ {_format_g((b, c, d))}]/F<</F{e}/UF{f}/Type/Filespec>>>>"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
579 str_gotor2 = lambda a, b, c: f"/A<</S/GoToR/D{a}/F<</F{b}/UF{c}/Type/Filespec>>>>"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
580 str_launch = lambda a, b: f"/A<</S/Launch/F<</F{a}/UF{b}/Type/Filespec>>>>"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
581 str_uri = lambda a: f"/A<</S/URI/URI{a}>>"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
582
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
583 if type(ddict) in (int, float):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
584 dest = str_goto(xref, 0, ddict, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
585 return dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
586 d_kind = ddict.get("kind", pymupdf.LINK_NONE)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
587
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
588 if d_kind == pymupdf.LINK_NONE:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
589 return ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
590
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
591 if ddict["kind"] == pymupdf.LINK_GOTO:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
592 d_zoom = ddict.get("zoom", 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
593 to = ddict.get("to", pymupdf.Point(0, 0))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
594 d_left, d_top = to
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
595 dest = str_goto(xref, d_left, d_top, d_zoom)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
596 return dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
597
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
598 if ddict["kind"] == pymupdf.LINK_URI:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
599 dest = str_uri(pymupdf.get_pdf_str(ddict["uri"]),)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
600 return dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
601
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
602 if ddict["kind"] == pymupdf.LINK_LAUNCH:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
603 fspec = pymupdf.get_pdf_str(ddict["file"])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
604 dest = str_launch(fspec, fspec)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
605 return dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
606
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
607 if ddict["kind"] == pymupdf.LINK_GOTOR and ddict["page"] < 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
608 fspec = pymupdf.get_pdf_str(ddict["file"])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
609 dest = str_gotor2(pymupdf.get_pdf_str(ddict["to"]), fspec, fspec)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
610 return dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
611
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
612 if ddict["kind"] == pymupdf.LINK_GOTOR and ddict["page"] >= 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
613 fspec = pymupdf.get_pdf_str(ddict["file"])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
614 dest = str_gotor1(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
615 ddict["page"],
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
616 ddict["to"].x,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
617 ddict["to"].y,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
618 ddict["zoom"],
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
619 fspec,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
620 fspec,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
621 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
622 return dest
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
623
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
624 return ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
625
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
626
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
627 def getLinkText(page: pymupdf.Page, lnk: dict) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
628 # --------------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
629 # define skeletons for /Annots object texts
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
630 # --------------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
631 ctm = page.transformation_matrix
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
632 ictm = ~ctm
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
633 r = lnk["from"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
634 rect = _format_g(tuple(r * ictm))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
635
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
636 annot = ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
637 if lnk["kind"] == pymupdf.LINK_GOTO:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
638 if lnk["page"] >= 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
639 txt = pymupdf.annot_skel["goto1"] # annot_goto
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
640 pno = lnk["page"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
641 xref = page.parent.page_xref(pno)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
642 pnt = lnk.get("to", pymupdf.Point(0, 0)) # destination point
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
643 dest_page = page.parent[pno]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
644 dest_ctm = dest_page.transformation_matrix
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
645 dest_ictm = ~dest_ctm
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
646 ipnt = pnt * dest_ictm
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
647 annot = txt(xref, ipnt.x, ipnt.y, lnk.get("zoom", 0), rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
648 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
649 txt = pymupdf.annot_skel["goto2"] # annot_goto_n
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
650 annot = txt(pymupdf.get_pdf_str(lnk["to"]), rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
651
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
652 elif lnk["kind"] == pymupdf.LINK_GOTOR:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
653 if lnk["page"] >= 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
654 txt = pymupdf.annot_skel["gotor1"] # annot_gotor
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
655 pnt = lnk.get("to", pymupdf.Point(0, 0)) # destination point
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
656 if type(pnt) is not pymupdf.Point:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
657 pnt = pymupdf.Point(0, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
658 annot = txt(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
659 lnk["page"],
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
660 pnt.x,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
661 pnt.y,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
662 lnk.get("zoom", 0),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
663 lnk["file"],
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
664 lnk["file"],
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
665 rect,
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
666 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
667 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
668 txt = pymupdf.annot_skel["gotor2"] # annot_gotor_n
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
669 annot = txt(pymupdf.get_pdf_str(lnk["to"]), lnk["file"], rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
670
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
671 elif lnk["kind"] == pymupdf.LINK_LAUNCH:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
672 txt = pymupdf.annot_skel["launch"] # annot_launch
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
673 annot = txt(lnk["file"], lnk["file"], rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
674
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
675 elif lnk["kind"] == pymupdf.LINK_URI:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
676 txt = pymupdf.annot_skel["uri"] # txt = annot_uri
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
677 annot = txt(lnk["uri"], rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
678
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
679 elif lnk["kind"] == pymupdf.LINK_NAMED:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
680 txt = pymupdf.annot_skel["named"] # annot_named
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
681 lname = lnk.get("name") # check presence of key
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
682 if lname is None: # if missing, fall back to alternative
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
683 lname = lnk["nameddest"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
684 annot = txt(lname, rect)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
685 if not annot:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
686 return annot
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
687
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
688 # add a /NM PDF key to the object definition
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
689 link_names = dict( # existing ids and their xref
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
690 [(x[0], x[2]) for x in page.annot_xrefs() if x[1] == pymupdf.PDF_ANNOT_LINK] # pylint: disable=no-member
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
691 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
692
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
693 old_name = lnk.get("id", "") # id value in the argument
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
694
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
695 if old_name and (lnk["xref"], old_name) in link_names.items():
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
696 name = old_name # no new name if this is an update only
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
697 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
698 i = 0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
699 stem = pymupdf.TOOLS.set_annot_stem() + "-L%i"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
700 while True:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
701 name = stem % i
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
702 if name not in link_names.values():
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
703 break
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
704 i += 1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
705 # add /NM key to object definition
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
706 annot = annot.replace("/Link", "/Link/NM(%s)" % name)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
707 return annot
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
708
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
709
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
710 # ----------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
711 # Name: wx.lib.colourdb.py
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
712 # Purpose: Adds a bunch of colour names and RGB values to the
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
713 # colour database so they can be found by name
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
714 #
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
715 # Author: Robin Dunn
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
716 #
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
717 # Created: 13-March-2001
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
718 # Copyright: (c) 2001-2017 by Total Control Software
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
719 # Licence: wxWindows license
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
720 # Tags: phoenix-port, unittest, documented
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
721 # ----------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
722
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
723
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
724 def getColorList() -> list:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
725 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
726 Returns a list of upper-case colour names.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
727 :rtype: list of strings
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
728 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
729 return [name for name, r, g, b in pymupdf.colors_wx_list()]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
730
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
731
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
732 def getColorInfoList() -> list:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
733 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
734 Returns list of (name, red, gree, blue) tuples, where:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
735 name: upper-case color name.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
736 read, green, blue: integers in range 0..255.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
737 :rtype: list of tuples
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
738 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
739 return pymupdf.colors_wx_list()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
740
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
741
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
742 def getColor(name: str) -> tuple:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
743 """Retrieve RGB color in PDF format by name.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
744
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
745 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
746 a triple of floats in range 0 to 1. In case of name-not-found, "white" is returned.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
747 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
748 return pymupdf.colors_pdf_dict().get(name.lower(), (1, 1, 1))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
749
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
750
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
751 def getColorHSV(name: str) -> tuple:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
752 """Retrieve the hue, saturation, value triple of a color name.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
753
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
754 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
755 a triple (degree, percent, percent). If not found (-1, -1, -1) is returned.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
756 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
757 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
758 x = getColorInfoList()[getColorList().index(name.upper())]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
759 except Exception:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
760 if g_exceptions_verbose: pymupdf.exception_info()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
761 return (-1, -1, -1)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
762
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
763 r = x[1] / 255.0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
764 g = x[2] / 255.0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
765 b = x[3] / 255.0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
766 cmax = max(r, g, b)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
767 V = round(cmax * 100, 1)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
768 cmin = min(r, g, b)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
769 delta = cmax - cmin
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
770 if delta == 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
771 hue = 0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
772 elif cmax == r:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
773 hue = 60.0 * (((g - b) / delta) % 6)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
774 elif cmax == g:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
775 hue = 60.0 * (((b - r) / delta) + 2)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
776 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
777 hue = 60.0 * (((r - g) / delta) + 4)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
778
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
779 H = int(round(hue))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
780
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
781 if cmax == 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
782 sat = 0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
783 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
784 sat = delta / cmax
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
785 S = int(round(sat * 100))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
786
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
787 return (H, S, V)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
788
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
789
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
790 def _get_font_properties(doc: pymupdf.Document, xref: int) -> tuple:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
791 fontname, ext, stype, buffer = doc.extract_font(xref)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
792 asc = 0.8
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
793 dsc = -0.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
794 if ext == "":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
795 return fontname, ext, stype, asc, dsc
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
796
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
797 if buffer:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
798 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
799 font = pymupdf.Font(fontbuffer=buffer)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
800 asc = font.ascender
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
801 dsc = font.descender
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
802 bbox = font.bbox
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
803 if asc - dsc < 1:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
804 if bbox.y0 < dsc:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
805 dsc = bbox.y0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
806 asc = 1 - dsc
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
807 except Exception:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
808 pymupdf.exception_info()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
809 asc *= 1.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
810 dsc *= 1.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
811 return fontname, ext, stype, asc, dsc
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
812 if ext != "n/a":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
813 try:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
814 font = pymupdf.Font(fontname)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
815 asc = font.ascender
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
816 dsc = font.descender
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
817 except Exception:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
818 pymupdf.exception_info()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
819 asc *= 1.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
820 dsc *= 1.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
821 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
822 asc *= 1.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
823 dsc *= 1.2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
824 return fontname, ext, stype, asc, dsc
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
825
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
826
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
827 def _show_fz_text( text):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
828 #if mupdf_cppyy:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
829 # assert isinstance( text, cppyy.gbl.mupdf.Text)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
830 #else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
831 # assert isinstance( text, mupdf.Text)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
832 num_spans = 0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
833 num_chars = 0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
834 span = text.m_internal.head
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
835 while 1:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
836 if not span:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
837 break
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
838 num_spans += 1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
839 num_chars += span.len
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
840 span = span.next
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
841 return f'num_spans={num_spans} num_chars={num_chars}'
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
842
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
843
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
844 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
845 Handle page labels for PDF documents.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
846
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
847 Reading
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
848 -------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
849 * compute the label of a page
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
850 * find page number(s) having the given label.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
851
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
852 Writing
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
853 -------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
854 Supports setting (defining) page labels for PDF documents.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
855
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
856 A big Thank You goes to WILLIAM CHAPMAN who contributed the idea and
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
857 significant parts of the following code during late December 2020
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
858 through early January 2021.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
859 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
860
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
861
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
862 def rule_dict(item):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
863 """Make a Python dict from a PDF page label rule.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
864
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
865 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
866 item -- a tuple (pno, rule) with the start page number and the rule
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
867 string like <</S/D...>>.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
868 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
869 A dict like
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
870 {'startpage': int, 'prefix': str, 'style': str, 'firstpagenum': int}.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
871 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
872 # Jorj McKie, 2021-01-06
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
873
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
874 pno, rule = item
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
875 rule = rule[2:-2].split("/")[1:] # strip "<<" and ">>"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
876 d = {"startpage": pno, "prefix": "", "firstpagenum": 1}
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
877 skip = False
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
878 for i, item in enumerate(rule): # pylint: disable=redefined-argument-from-local
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
879 if skip: # this item has already been processed
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
880 skip = False # deactivate skipping again
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
881 continue
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
882 if item == "S": # style specification
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
883 d["style"] = rule[i + 1] # next item has the style
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
884 skip = True # do not process next item again
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
885 continue
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
886 if item.startswith("P"): # prefix specification: extract the string
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
887 x = item[1:].replace("(", "").replace(")", "")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
888 d["prefix"] = x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
889 continue
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
890 if item.startswith("St"): # start page number specification
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
891 x = int(item[2:])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
892 d["firstpagenum"] = x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
893 return d
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
894
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
895
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
896 def get_label_pno(pgNo, labels):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
897 """Return the label for this page number.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
898
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
899 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
900 pgNo: page number, 0-based.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
901 labels: result of doc._get_page_labels().
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
902 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
903 The label (str) of the page number. Errors return an empty string.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
904 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
905 # Jorj McKie, 2021-01-06
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
906
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
907 item = [x for x in labels if x[0] <= pgNo][-1]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
908 rule = rule_dict(item)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
909 prefix = rule.get("prefix", "")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
910 style = rule.get("style", "")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
911 # make sure we start at 0 when enumerating the alphabet
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
912 delta = -1 if style in ("a", "A") else 0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
913 pagenumber = pgNo - rule["startpage"] + rule["firstpagenum"] + delta
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
914 return construct_label(style, prefix, pagenumber)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
915
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
916
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
917 def construct_label(style, prefix, pno) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
918 """Construct a label based on style, prefix and page number."""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
919 # William Chapman, 2021-01-06
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
920
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
921 n_str = ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
922 if style == "D":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
923 n_str = str(pno)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
924 elif style == "r":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
925 n_str = integerToRoman(pno).lower()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
926 elif style == "R":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
927 n_str = integerToRoman(pno).upper()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
928 elif style == "a":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
929 n_str = integerToLetter(pno).lower()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
930 elif style == "A":
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
931 n_str = integerToLetter(pno).upper()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
932 result = prefix + n_str
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
933 return result
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
934
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
935
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
936 def integerToLetter(i) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
937 """Returns letter sequence string for integer i."""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
938 # William Chapman, Jorj McKie, 2021-01-06
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
939 import string
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
940 ls = string.ascii_uppercase
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
941 n, a = 1, i
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
942 while pow(26, n) <= a:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
943 a -= int(math.pow(26, n))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
944 n += 1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
945
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
946 str_t = ""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
947 for j in reversed(range(n)):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
948 f, g = divmod(a, int(math.pow(26, j)))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
949 str_t += ls[f]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
950 a = g
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
951 return str_t
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
952
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
953
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
954 def integerToRoman(num: int) -> str:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
955 """Return roman numeral for an integer."""
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
956 # William Chapman, Jorj McKie, 2021-01-06
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
957
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
958 roman = (
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
959 (1000, "M"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
960 (900, "CM"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
961 (500, "D"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
962 (400, "CD"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
963 (100, "C"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
964 (90, "XC"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
965 (50, "L"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
966 (40, "XL"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
967 (10, "X"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
968 (9, "IX"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
969 (5, "V"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
970 (4, "IV"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
971 (1, "I"),
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
972 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
973
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
974 def roman_num(num):
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
975 for r, ltr in roman:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
976 x, _ = divmod(num, r)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
977 yield ltr * x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
978 num -= r * x
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
979 if num <= 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
980 break
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
981
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
982 return "".join([a for a in roman_num(num)])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
983
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
984
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
985 # -------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
986 # Functions to recover the quad contained in a text extraction bbox
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
987 # -------------------------------------------------------------------
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
988 def recover_bbox_quad(line_dir: tuple, span: dict, bbox: tuple) -> pymupdf.Quad:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
989 """Compute the quad located inside the bbox.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
990
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
991 The bbox may be any of the resp. tuples occurring inside the given span.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
992
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
993 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
994 line_dir: (tuple) 'line["dir"]' of the owning line or None.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
995 span: (dict) the span. May be from get_texttrace() method.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
996 bbox: (tuple) the bbox of the span or any of its characters.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
997 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
998 The quad which is wrapped by the bbox.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
999 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1000 if line_dir is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1001 line_dir = span["dir"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1002 cos, sin = line_dir
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1003 bbox = pymupdf.Rect(bbox) # make it a rect
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1004 if pymupdf.TOOLS.set_small_glyph_heights(): # ==> just fontsize as height
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1005 d = 1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1006 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1007 d = span["ascender"] - span["descender"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1008
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1009 height = d * span["size"] # the quad's rectangle height
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1010 # The following are distances from the bbox corners, at which we find the
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1011 # respective quad points. The computation depends on in which quadrant the
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1012 # text writing angle is located.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1013 hs = height * sin
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1014 hc = height * cos
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1015 if hc >= 0 and hs <= 0: # quadrant 1
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1016 ul = bbox.bl - (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1017 ur = bbox.tr + (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1018 ll = bbox.bl - (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1019 lr = bbox.tr + (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1020 elif hc <= 0 and hs <= 0: # quadrant 2
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1021 ul = bbox.br + (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1022 ur = bbox.tl - (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1023 ll = bbox.br + (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1024 lr = bbox.tl - (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1025 elif hc <= 0 and hs >= 0: # quadrant 3
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1026 ul = bbox.tr - (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1027 ur = bbox.bl + (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1028 ll = bbox.tr - (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1029 lr = bbox.bl + (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1030 else: # quadrant 4
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1031 ul = bbox.tl + (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1032 ur = bbox.br - (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1033 ll = bbox.tl + (0, hc)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1034 lr = bbox.br - (hs, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1035 return pymupdf.Quad(ul, ur, ll, lr)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1036
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1037
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1038 def recover_quad(line_dir: tuple, span: dict) -> pymupdf.Quad:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1039 """Recover the quadrilateral of a text span.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1040
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1041 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1042 line_dir: (tuple) 'line["dir"]' of the owning line.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1043 span: the span.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1044 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1045 The quadrilateral enveloping the span's text.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1046 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1047 if type(line_dir) is not tuple or len(line_dir) != 2:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1048 raise ValueError("bad line dir argument")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1049 if type(span) is not dict:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1050 raise ValueError("bad span argument")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1051 return recover_bbox_quad(line_dir, span, span["bbox"])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1052
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1053
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1054 def recover_line_quad(line: dict, spans: list = None) -> pymupdf.Quad:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1055 """Calculate the line quad for 'dict' / 'rawdict' text extractions.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1056
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1057 The lower quad points are those of the first, resp. last span quad.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1058 The upper points are determined by the maximum span quad height.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1059 From this, compute a rect with bottom-left in (0, 0), convert this to a
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1060 quad and rotate and shift back to cover the text of the spans.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1061
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1062 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1063 spans: (list, optional) sub-list of spans to consider.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1064 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1065 pymupdf.Quad covering selected spans.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1066 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1067 if spans is None: # no sub-selection
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1068 spans = line["spans"] # all spans
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1069 if len(spans) == 0:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1070 raise ValueError("bad span list")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1071 line_dir = line["dir"] # text direction
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1072 cos, sin = line_dir
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1073 q0 = recover_quad(line_dir, spans[0]) # quad of first span
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1074 if len(spans) > 1: # get quad of last span
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1075 q1 = recover_quad(line_dir, spans[-1])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1076 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1077 q1 = q0 # last = first
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1078
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1079 line_ll = q0.ll # lower-left of line quad
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1080 line_lr = q1.lr # lower-right of line quad
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1081
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1082 mat0 = pymupdf.planish_line(line_ll, line_lr)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1083
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1084 # map base line to x-axis such that line_ll goes to (0, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1085 x_lr = line_lr * mat0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1086
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1087 small = pymupdf.TOOLS.set_small_glyph_heights() # small glyph heights?
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1088
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1089 h = max(
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1090 [s["size"] * (1 if small else (s["ascender"] - s["descender"])) for s in spans]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1091 )
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1092
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1093 line_rect = pymupdf.Rect(0, -h, x_lr.x, 0) # line rectangle
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1094 line_quad = line_rect.quad # make it a quad and:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1095 line_quad *= ~mat0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1096 return line_quad
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1097
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1098
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1099 def recover_span_quad(line_dir: tuple, span: dict, chars: list = None) -> pymupdf.Quad:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1100 """Calculate the span quad for 'dict' / 'rawdict' text extractions.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1101
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1102 Notes:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1103 There are two execution paths:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1104 1. For the full span quad, the result of 'recover_quad' is returned.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1105 2. For the quad of a sub-list of characters, the char quads are
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1106 computed and joined. This is only supported for the "rawdict"
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1107 extraction option.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1108
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1109 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1110 line_dir: (tuple) 'line["dir"]' of the owning line.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1111 span: (dict) the span.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1112 chars: (list, optional) sub-list of characters to consider.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1113 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1114 pymupdf.Quad covering selected characters.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1115 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1116 if line_dir is None: # must be a span from get_texttrace()
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1117 line_dir = span["dir"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1118 if chars is None: # no sub-selection
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1119 return recover_quad(line_dir, span)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1120 if "chars" not in span.keys():
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1121 raise ValueError("need 'rawdict' option to sub-select chars")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1122
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1123 q0 = recover_char_quad(line_dir, span, chars[0]) # quad of first char
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1124 if len(chars) > 1: # get quad of last char
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1125 q1 = recover_char_quad(line_dir, span, chars[-1])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1126 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1127 q1 = q0 # last = first
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1128
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1129 span_ll = q0.ll # lower-left of span quad
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1130 span_lr = q1.lr # lower-right of span quad
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1131 mat0 = pymupdf.planish_line(span_ll, span_lr)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1132 # map base line to x-axis such that span_ll goes to (0, 0)
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1133 x_lr = span_lr * mat0
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1134
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1135 small = pymupdf.TOOLS.set_small_glyph_heights() # small glyph heights?
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1136 h = span["size"] * (1 if small else (span["ascender"] - span["descender"]))
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1137
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1138 span_rect = pymupdf.Rect(0, -h, x_lr.x, 0) # line rectangle
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1139 span_quad = span_rect.quad # make it a quad and:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1140 span_quad *= ~mat0 # rotate back and shift back
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1141 return span_quad
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1142
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1143
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1144 def recover_char_quad(line_dir: tuple, span: dict, char: dict) -> pymupdf.Quad:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1145 """Recover the quadrilateral of a text character.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1146
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1147 This requires the "rawdict" option of text extraction.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1148
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1149 Args:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1150 line_dir: (tuple) 'line["dir"]' of the span's line.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1151 span: (dict) the span dict.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1152 char: (dict) the character dict.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1153 Returns:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1154 The quadrilateral enveloping the character.
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1155 """
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1156 if line_dir is None:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1157 line_dir = span["dir"]
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1158 if type(line_dir) is not tuple or len(line_dir) != 2:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1159 raise ValueError("bad line dir argument")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1160 if type(span) is not dict:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1161 raise ValueError("bad span argument")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1162 if type(char) is dict:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1163 bbox = pymupdf.Rect(char["bbox"])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1164 elif type(char) is tuple:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1165 bbox = pymupdf.Rect(char[3])
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1166 else:
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1167 raise ValueError("bad span argument")
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1168
1d09e1dec1d9 ADD: PyMuPDF v1.26.4: the original sdist.
Franz Glasner <fzglas.hg@dom66.de>
parents:
diff changeset
1169 return recover_bbox_quad(line_dir, span, bbox)