white-space: nowrap still wraps on hyphens

[Smylers]

With white-space: nowrap, WeasyPrint will still wrap text on a hyphen (technically U+2D hyphen-minus) even when it won't wrap on a space at that point.

See: http://www.stripey.com/demo/weasyprint/nowrap_hyphen.html, which WeasyPrint (version 0.21) renders as: http://www.stripey.com/demo/weasyprint/nowrap_hyphen.pdf

Obviously hyphens aren't white space, but it doesn't seem helpful to wrap at this point. Unfortunately the stable CSS2.1 spec makes no mention of what the white-space property should do with wrapping characters that aren't white space: http://www.w3.org/TR/CSS2/text.html#white-space-prop

However the CSS3 editors' draft makes clear that white-space governs all ‘soft wrap opportunities’, which includes punctuation characters where wrapping can occur: http://dev.w3.org/csswg/css-text/#white-space-property

[SimonSapin]

I believe this is Pango’s doing. IMO the "proper" fix for this and other related bugs would be to do line breaking ourselves in WeasyPrint rather than leave it to Pango. But for this specifically maybe we can special-case white-space: nowrap and give infinite available width to Pango, rather than just replace spaces with no-break spaces as (IIRC) we do now.

[Smylers]

rather than just replace spaces with no-break spaces as (IIRC) we do now.

In that case, another possible way of getting this working would be to additionally massage the input text to Pango by inserting U+2060 word joiners just before and after each hyphen or slash character.

That is actually what I'm now doing in the HTML template for affected text, though obviously that's a layering violation, since it means that the HTML has to know which CSS rules will be applied where.

[SirNeuman]

So is there any simple workaround for this problem currently? In my css i have white-space: nowrap, which works for spaces but not for hyphens.

[Smylers]

So is there any simple workaround for this problem currently?

Its simplicity is subjective, but the workaround I mention above is to process text that's going to be put in nowrap containers with:

$text =~ s!([/-])!\N{WORD JOINER}$1\N{WORD JOINER}!g;

I happen to be using Perl, but the underlying idea should translate straightforwardly to other text-processing languages.

[SirNeuman]

Ahh ok. Sorry, by simple i meant either html or css way of doing it. I'll have to find the python equivalent of this then. I'm not too good with regex stuff, but it's always satisfying figuring it out.

[liZe]

Grouped in #301.

[liZe]

Even without rewriting the line breaking algorithm, this bug is fixed in a quite clean way.