Asciidoctor-pdf: Index term gets double space or extra padding-left?

[peterjaap]

I have the following text;

From Laravel's perspective, a model is an _Eloquent_ model. ((Eloquent)) is an Object Relational Mapper (ORM) (((Object Relational Mapper,ORM))) built specifically for ((Laravel)) but is useful in any PHP application. It ties a model to a corresponding table using a simple ((ActiveRecord)) pattern and offers a ((fluent API)).

In my theme, I use the MinionPro font, which I defined like this;

font:
  catalog:
    merge: true
    MinionPro:
      regular: MinionPro/MinionProRegular.ttf
      italic: MinionPro/MinionProItalic.ttf
      bold: MinionPro/MinionProBold.ttf
      bold_italic: MinionPro/MinionProBoldItalic.ttf

base:
  font:
    family: MinionPro
    size: 12
    color: #303030

When I generate my PDF, this is my output. You can see that the index terms have extra whitespace preceding them.

As you can see, this also happens when using concealed index terms.

Here is the copy/pasted text;

From Laravel’s perspective, a model is an Eloquent model. Eloquent is
an Object Relational Mapper (ORM) built specifically for Laravel but is
useful in any PHP application. It ties a model to a corresponding table
using a simple ActiveRecord pattern and offers a fluent API.

I also ran this copy/pasted text through a tool to check whether the space is a weird unicode character but it appears to be a normal space (U+0020 : SPACE [SP]).

I also can't find anything controlling this behavior in the theming/styling docs. It doesn't happen when using other fonts so I think it might be font-related but I don't know how or why. This is what it looks like with the default font;

[mojavelinux]

You're observation is correct. Non-visible index terms cause spaces on either side not to be collapsed. (We have to use a null character to force an anchor to be created by the PDF generator). Therefore, you need to put the terms directly adjacent to existing text so that they don't stand alone.

We don't see this in the HTML output even though there is extra space since HTML is much more aggressive about collapsing spaces.

[mojavelinux]

Hmm, though I see the same behavior is happening with visible index terms...that is a problem.

[mojavelinux]

Oh, I know what it is. It means the font you are using is missing one of the required glyphs (in this case the null character) (or it doesn't have a zero width in the font). When the glyph is missing (or has the incorrect width), the PDF generator inserts a space (since it is interpreting the font).

This is detailed here: https://github.com/asciidoctor/asciidoctor-pdf/blob/v1.5.x/docs/theming-guide.adoc#modifying-the-font

(We don't see this behavior when using the provided Noto Serif font).

[mojavelinux]

It looks like we could make this more portable by switching from the null character to a zero-width space. The reason is, the PDF generator already removes zero-width spaces from the text during processing without consulting the font. So you don't need that glyph in your font, it won't show up in the output, it always resolves to a 0 width (so space won't be reserved for it). Seems like reason enough to it. I'll do some testing.

[mojavelinux]

Yep, that's definitely the right solution.

[peterjaap]

Great! For now I'll just remove any space before (( in my preprocessor.

[mojavelinux]

It turns out that advance won't work, at least not for visible index terms. The correct workaround is to open up the font (using fontforge) and copy the U+200B value to the U+0000 value, then save the font.

(This could also be done with a script, which is what we used for preparing the bundled fonts).

[peterjaap]

@mojavelinux I was just trying that out with FontForge :) Thanks

[mojavelinux]

I just gave it a try and it worked. Here's what I did:

1. Open font in fontforge
2. Navigate to U+200B (using Ctrl+Shift+>)
3. Right click and Copy
4. Navigate to U+0000
5. Right click and Paste
6. File > Generate Fonts...
7. Generate

[mojavelinux]

Once I sort out why one of the tests is failing, I'll create a PR to make the switch from null to zero-width space to avoid this whole exercise.

[peterjaap]

My font also doesn't have the U+200B character. I tried copying the exclamation mark to U+0000 and U+200B to see if that would show up, but it doesn't. When I copy the exclamation mark to the space character, it does show up.

[peterjaap]

I've also tried adding Noto Serif as a fallback font so it will pick the missing glyphs from there but that doesn't help either.

[mojavelinux]

Find any other glyph in the font that is empty to copy. Then, edit the
metrics and set the width to 0. You just need the glyph to be empty with 0
width.

[peterjaap]

@mojavelinux thanks, setting the width to 0 worked! :+1:

Edit; I could have read that in the docs....

[peterjaap]

The issue with the concealed index item persists using this workaround but I guess you already knew that.

[mojavelinux]

The issue with the concealed index item persists using this workaround but I guess you already knew that.

Yes, concealed terms in the middle of text should be affixed to visible text. That's just a general best practice because the spaces on either side can be misinterpreted (as there are many layers of processing to convince they are not significant).

[mojavelinux]

Good news! I found the source of the bug. When the null character is concealed, its width is not set to 0. If the font incorrectly reports the null character as having a width, then you end up with extra space (even though no character is detected at that location). I'll submit a PR shortly.

[peterjaap]

Nice!

On Thu, 30 Apr 2020 at 23:33, Dan Allen [e-mail] wrote:

Good news! I found the source of the bug. When the null character is
concealed, its width is not set to 0. If the font incorrectly reports the
null character as having a width, then you end up with extra space (even
though no character is detected at that location). I'll submit a PR shortly.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/asciidoctor/asciidoctor-pdf/issues/1672#issuecomment-622126202,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AADJKAFODWRYQLMI4FWMQX3RPHVCXANCNFSM4MTQSCKA
.