Re: [help-texinfo] Unicode line drawing chars in pdf output

[James Cloos]

>>>>> "TN" == Thien-Thi Nguyen <address@hidden> writes:

TN> The word "trivial" is tantalizing, but i have no experience whatsoever.
TN> I suppose "FF" stands for "Font Forge".  Is that right?  Any suggestions
TN> for a font newbie to ramp-up on these terms / algorithms very welcome.

Sorry that I took so long to reply.

FontForge kept crashing (SEGV) when I tried to work out a recipe.

I've recently recompiled it w/ less optimization and that seems to avoid
the crashes.

The basic idea it to:

  go the the Element -> Font Info dialog

  choose the General tab

  change Em size to 1000 (make sure Scale outlines is selected)

  hit the OK button (this will take a while)

  save your work so far

  then back to Element -> Font Info

  choose the Layers tab

  select all layers cubic

  hit OK.  Another dialog will pop up requiring confirmation

  save again

  Edit -> Select -> Glyphs worth outputting

  Element -> Correct Direction

  And be happy you saved your work :(

  restart ff :)

  Edit -> Select -> glyphs in need of hinting

  Hints -> Autohint

  save

  select glyphs (perhaps autohintable is enough)

  Element -> Style -> Condense

  90% is close to the transformation needed to match the width of CMTT10

  save

  File -> Generate Fonts

  and discover that ff isn't quite smart enough to keep a fixed width
  font fixed in the face of its consense/expand algorithm.

Which shows that this'll need more though.

Perhaps removing every non-line-drawing glyph will make it easier?

That certainly would make the per-glyph steps faster.

I did this once before for someone who needed a narrower font for
setting code in a book (the same reason cmtt is narrower) and got
a fixed-width font out of it.  

But that might have been before George made the scaling treat stems,
counters and side-bearings separately.  The transform dialog might
be the answer here.  Or perphaps a python script can determine the
optimal transforms for each glyph.

It seems it is not as trivial as I recalled.

-JimC
-- 
James Cloos <address@hidden>         OpenPGP: 1024D/ED7DAEA6