Non-breaking space

[TychoVrahe]

Introduce non-breaking space character, which will be used to prevent breaking lines where non appropriate, like after the a article.

[andrewkozlik]

I was thinking about this just recently, when I didn't like breaking "[...] account #1" after "account".

I was looking briefly at the algorithm at the time and one thing I wanted to delve deeper into is how it breaks words. It introduces a dash at the end of the line, when it wants to break, but what if the word already has a dash at that position? For a word like "non-private" you could get:

[...] non--
private [...]

or

[...] non-
-private [...]

This is untested, just my hypothesis. The correct breaking is

[...] non-
private [...]

Finally, I was thinking about introducing something like \- in LaTeX. It would encourage the algorithm to break at that point if the whole word does not fit on the line. So for example, let's say you have a sentence with words "{W1} {W2} {W3} ..." and W2 is every\-where, then:

• if "{W1} everywhere" fits on a line, then ignore the \- and continue as usual.
• if "{W1} everywhere" doesn't fit on a line, but "{W1} every-" does fit, then break it as ["{W1} every-", "where {W3} ..."].
• if "{W1} every-" doesn't fit on a line, but "{W1}" does, then break it as ["{W1}", "everywhere {W3} ..."].

For practical reasons, we'd probably want to reuse some of the standard escape-sequences for the nbsp and the \-. The options seem to be \t, \b, \f, \v, \a.

[mmilata]

Looks like we started receiving translated strings containing non-breakable spaces in the right places, yet can't handle them.

• we could add the char to core/tools/codegen/foreign_chars.py but we'd need to introduce similar mechanism for english which probably doesn't make sense, or
• we could add special handling for 0xA0 codepoint in the text renderer.

After it works the commit 1b31b4033069fc0d4d789e5e3a004b5c8cfc9199 needs to be reverted.

[matejcik]

we could add special handling for 0xA0 codepoint in the text renderer.

I'd just hack get_glyph to return the 0x20 (ascii space) glyph for 0xA0 also. That way the space is counted for width but doesn't count as a breaking point (those are hard-coded to ascii space and newlines i think?)