Non-breaking spaces have different size

[ondras]

In this rendering:

First two words (o kliknutí) have a NBSP between them (czech typography rules). However, IMHO this only forbirds line breaks; there is no reason for this space to have a different size than other inter-word spaces on the same line.

[liZe]

You’re right. The current justification code adds extra space to space characters only. There are many places in the code when we assume that spaces are only "normal" spaces.

In this case, the problem is in:

WeasyPrint/weasyprint/layout/inline.py

Lines 1119 to 1125 in 1aae145

	def justify_line(context, line, extra_width):
	# TODO: We should use a better algorithm here, see
	# https://www.w3.org/TR/css-text-3/#justify-algos
	nb_spaces = count_spaces(line)
	if nb_spaces == 0:
	return
	add_word_spacing(context, line, extra_width / nb_spaces, 0)

We count spaces and use it to set justification_spacing. We should change our space detection here:

WeasyPrint/weasyprint/layout/inline.py

Line 1131 in 1aae145

return box.text.count(' ')

And fix our spacing adjustment here:

WeasyPrint/weasyprint/text/line_break.py

Lines 182 to 196 in 1aae145

	if word_spacing:
	if bytestring == b' ':
	# We need more than one space to set word spacing
	self.text = ' \u200b' # Space + zero-width space
	text, bytestring = unicode_to_char_p(self.text)
	pango.pango_layout_set_text(self.layout, text, -1)
	space_spacing = int(word_spacing * TO_UNITS + letter_spacing)
	position = bytestring.find(b' ')
	# Pango gives only half of word-spacing on boundaries
	boundary_positions = (0, len(bytestring) - 1)
	while position != -1:
	factor = 1 + (position in boundary_positions)
	add_attr(position, position + 1, factor * space_spacing)
	position = bytestring.find(b' ', position + 1)

For now we don’t have to support all justification opportunities as we don’t support text-justify, but we can at least support word separators. I’m not sure that a real list is actually defined in Unicode, as there are exceptions such as punctuation and fixed-width spaces. We can at least start with the list given by the specification.