summaryrefslogtreecommitdiffstats
path: root/share/extensions/text_split.py
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 18:24:48 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 18:24:48 +0000
commitcca66b9ec4e494c1d919bff0f71a820d8afab1fa (patch)
tree146f39ded1c938019e1ed42d30923c2ac9e86789 /share/extensions/text_split.py
parentInitial commit. (diff)
downloadinkscape-12fc8abae6d434cac7670a59ed3a67301cc2eb10.tar.xz
inkscape-12fc8abae6d434cac7670a59ed3a67301cc2eb10.zip
Adding upstream version 1.2.2.upstream/1.2.2upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'share/extensions/text_split.py')
-rwxr-xr-xshare/extensions/text_split.py313
1 files changed, 313 insertions, 0 deletions
diff --git a/share/extensions/text_split.py b/share/extensions/text_split.py
new file mode 100755
index 0000000..dee9c88
--- /dev/null
+++ b/share/extensions/text_split.py
@@ -0,0 +1,313 @@
+#!/usr/bin/env python
+# coding=utf-8
+#
+# Copyright (C) 2009 Karlisson Bezerra, contato@nerdson.com
+# 2021 Jonathan Neuhauser, jonathan.neuhauser@outlook.com
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+#
+"""Splits a text element into lines, words, chars.
+Supports all text elements that Inkscape can create, such as normal text, shape-inside (SVG2),
+flowroot (SVG1.2), inline-size, manual kerns, and nested tspans (with possibly different kerns)
+
+The code is structured as followed. For each selected text element:
+ - preprocess_text_element duplicates the element, converts flowroots to plain text elements,
+ and simplifies manual kerning if requested (only for split_words and split_chars, for all,
+ possibly nested, children) using simplify_nested_tspans.
+ - if split lines: split_lines copies all top-level tspans from the previous step into their own
+ text element, which is otherwise a duplicate of the original text element (thus preserving
+ style and transforms), see append_splitted_element
+ - if split words or chars: split_words_or_chars: the text is recursively processed. For each tspan,
+ the content and tail is split (words: at spaces, chars: after each character) into their own
+ tspan, again using append_splitted_element. The method keeps track of the horizontal and vertical
+ coordinate, incrementing it with the number of characters and a multiple of font size.
+"""
+
+import re as regex
+from typing import Union, Callable
+
+import inkex
+from inkex import TextElement, FlowRoot, FlowPara, Tspan, Rectangle, ShapeElement
+from inkex.units import parse_unit
+from inkex.localization import inkex_gettext as _
+
+TextLike = Union[FlowRoot, TextElement]
+
+
+class TextSplit(inkex.EffectExtension):
+ """Split text up."""
+
+ def __init__(self):
+ """Initialize State machine"""
+ super().__init__()
+ self.mode: Callable
+ self.separation: float = 1
+ self.fs_multiplier: float = 0.25
+ self.current_x: float = 0
+ self.current_y: float = 0
+ self.process_kerns: bool = True
+ self.current_root: TextLike
+ self.current_fontsize: float = 0
+
+ def add_arguments(self, pars):
+ pars.add_argument("--tab", help="The selected UI tab when OK was pressed")
+ pars.add_argument(
+ "-t",
+ "--splittype",
+ default="line",
+ choices=["letter", "word", "line"],
+ help="type of split",
+ )
+ pars.add_argument(
+ "-p",
+ "--preserve",
+ type=inkex.Boolean,
+ default=True,
+ help="Preserve original",
+ )
+ pars.add_argument(
+ "-s",
+ "--separation",
+ type=float,
+ default=1,
+ help="Threshold for separating text with manual kerns in multiples of"
+ "font-size",
+ )
+
+ def effect(self):
+ """Applies the effect"""
+
+ split_type = self.options.splittype
+ preserve = self.options.preserve
+
+ # checks if the selected elements are text nodes
+ for elem in self.svg.selection.filter_nonzero(TextElement, FlowRoot):
+ try:
+ self.separation = self.options.separation
+ if split_type == "line":
+ node = self.split_lines(elem)
+ elif split_type == "word":
+ self.mode = self.process_plain_words
+ node = self.split_words_or_chars(elem)
+ else:
+ self.separation = 0
+ self.mode = self.process_plain_chars
+ node = self.split_words_or_chars(elem)
+
+ node.getparent().remove(node)
+
+ if not preserve and node is not None:
+ elem.getparent().remove(elem)
+ except TypeError as err:
+ inkex.errormsg(err) # if an element can not be processed
+
+ @staticmethod
+ def get_font_size(element):
+ """get the font size of an element"""
+ return element.specified_style()("font-size")
+
+ @staticmethod
+ def get_line_height(element: ShapeElement):
+ """get the line height of an element"""
+ return element.get_line_height_uu()
+
+ def simplify_child_tspans(self, element: TextElement):
+ """Checks all child tspans if they have manual kerns.
+ If it does, try to find words (characters with a distance > separation * font-size).
+ Then concatenate the words with spaces, set this string as a new text and"""
+ for child in list(element):
+ # process manual kerns
+ if not isinstance(child, Tspan):
+ continue
+ xvals = list(
+ map(float, filter(len, regex.split(r"[,\s]", child.get("x") or "")))
+ )
+ content = child.text
+ if content not in [None, ""] and len(xvals) >= 2:
+ fsize = self.get_font_size(child)
+ separation = self.separation * fsize
+ current_word_start = 0
+ for i in range(1, max(len(content), len(xvals))):
+ if i >= len(content) - 1 or i >= len(xvals) - 1:
+ # consume the entire remaining string
+ i = len(content)
+ if i == len(content) or abs(xvals[i] - xvals[i - 1]) > separation:
+ wordspan = Tspan(x=str(xvals[current_word_start]))
+ wordspan.text = content[current_word_start:i]
+ child.add(wordspan)
+ current_word_start = i
+ child.pop("x")
+ child.text = None
+ # process child elements
+ self.simplify_child_tspans(child)
+
+ def preprocess_text_element(self, element: TextElement):
+ """Processes a text element and returns an element containing tspans with x and y coordinate,
+ possibly nested (for Inkscape-type kerning), so that the actual splitting can work as if the
+ text was a simple text. Manual kerns (one x value per letter) are converted to spaces
+ if requested (not necessary for "split characters")"""
+
+ oldelement = element
+ if isinstance(element, FlowRoot):
+ element = TextElement()
+ oldelement.addnext(element)
+ element.style = oldelement.style
+ element.transform = oldelement.transform
+ flowref = oldelement.findone("svg:flowRegion")[0]
+ if isinstance(flowref, Rectangle):
+ flowx = element.unittouu(flowref.get("x"))
+ flowy = element.unittouu(float(flowref.get("y")))
+ first = True
+ else:
+ raise TypeError(
+ _(
+ "Element {} uses a flow region that is not a rectangle. "
+ "First unflow text."
+ ).format(element.get_id())
+ )
+ for child in oldelement:
+ if isinstance(child, FlowPara):
+ # convert the flowpara "line" (note: no automatic wrapping)
+ # to a tspan and set the y coordinate.
+ # future FlowRoot improvements could add a better conversion.
+ newchild = Tspan()
+ element.append(newchild)
+ newchild.text = child.text
+ newchild.style = child.style
+ newchild.transform = child.transform
+ newchild.set("x", flowx)
+ if first:
+ flowy += self.get_font_size(child) * 1.25
+ first = False
+ else:
+ flowy += self.get_line_height(child)
+ newchild.set("y", str(flowy))
+
+ else:
+ element = oldelement.duplicate()
+ oldelement.getparent().append(element)
+
+ element.style.pop("shape-inside", None)
+
+ # Real support for RTL text is missing, but we can emulate it by just removing the
+ # attribute. However, line breaks will be misaligned.
+ element.style.pop("direction", None)
+ for child in element:
+ child.style.pop("direction", None)
+
+ if self.process_kerns:
+ self.simplify_child_tspans(element)
+ return element
+
+ def append_splitted_element(self, text, prototype=None):
+ """Creates a new text element, sibling to self.current_root, at (self.current_x,
+ self.current_y) with content text.
+
+ text: either a Tspan that should be moved to a new text element - in this case, text is
+ a direct child of element; or a string
+ prototype: if text is a string, style and transform will be taken from prototype"""
+
+ if isinstance(text, Tspan) and text.getparent() == self.current_root:
+ # we just move the tspan to a new text element.
+ elem = self.current_root.duplicate()
+ elem.remove_all(Tspan)
+ elem.append(text)
+ elem.set("x", text.get("x"))
+ elem.set("y", text.get("y"))
+ else:
+ elem = TextElement(x=str(self.current_x), y=str(self.current_y))
+ # transfer the style from all parents, including the text element (if there's a style to
+ # the text element's parent applied, it will be duplicated, but that doesn't really
+ # matter)
+ elem.style = prototype.specified_style()
+ # the element will be appended to the parent of element, but there might be nested
+ # tspans between the prototype and the element. The next line says
+ # "compose transforms until you reach the parent of element"
+ elem.transform = (
+ -self.current_root.getparent().transform
+ ) @ prototype.composed_transform()
+ tsp = Tspan(x=str(self.current_x), y=str(self.current_y))
+ tsp.text = text
+ elem.add(tsp)
+ self.current_root.addnext(elem)
+
+ def split_lines(self, element: TextLike) -> TextElement:
+ """Splits a text into its lines"""
+ self.process_kerns = False
+ preprocessed = self.preprocess_text_element(element)
+ self.current_root = preprocessed
+ # Now we only have to copy each tspan into its own text element.
+ for child in list(preprocessed):
+ self.append_splitted_element(child)
+
+ return preprocessed
+
+ def process_plain_text(self, element, splitted):
+ """Appends new text elements to as sibling root for each element of splitted, starting at
+ self.current_x, self.current_y, incrementing those, with prototype element (that
+ styles and transforms will be taken from)"""
+ if splitted is None:
+ return
+ for word in splitted:
+ if word != "":
+ self.append_splitted_element(word, element)
+ # +1 since for words, we lost a space
+ self.current_x += (
+ self.current_fontsize * (len(word) + 1) * self.fs_multiplier
+ )
+
+ def process_plain_words(self, element, text):
+ """Calls process_plain_text for splitting words"""
+ self.fs_multiplier = 0.4
+ if text is not None:
+ self.process_plain_text(element, text.split(" "))
+
+ def process_plain_chars(self, element, text):
+ """Calls process_plain_text for splitting characters"""
+ self.fs_multiplier = 0.25
+ self.process_plain_text(element, text)
+
+ def split_words_or_chars(self, element: TextLike) -> TextElement:
+ """Splits a text into its lines"""
+ self.process_kerns = True
+ preprocessed = self.preprocess_text_element(element)
+
+ def process_element(element) -> float:
+ elem_coords = {
+ i: element.root.unittouu(element.get(i))
+ if element.get(i) is not None
+ else None
+ for i in "xy"
+ }
+ if elem_coords["x"] is not None:
+ self.current_x = elem_coords["x"]
+ if elem_coords["y"] is not None:
+ self.current_y = elem_coords["y"]
+ self.current_fontsize = self.get_font_size(element)
+ current_x = self.mode(element, element.text)
+
+ for elem in element:
+ if isinstance(elem, Tspan):
+ current_x = process_element(elem)
+ current_x = self.mode(element, elem.tail)
+ return current_x
+
+ self.current_root = preprocessed
+ process_element(preprocessed)
+ return preprocessed
+
+
+if __name__ == "__main__":
+ TextSplit().run()