How can one specify in PL-file the kerning for a glyph and a ligature (say, digit and en-dash) except for setting it for the first part of the ligature (which I would like to avoid)?

  • 2
    I don't think it's possible to have separate kerning for 0/hyphen and 0/en-dash unless you input the en-dash directly as \char... (masked by a macro, of course).
    – egreg
    Sep 15, 2014 at 13:28
  • @egreg Are you sure? The reason I ask is that if you construct a ligature from existing glyphs, you copy the kerning from the originals to ensure correct kerning on each side. So if I'm constructing fi, I'd copy the left kerning pairs from f and the right kerning pairs from i. If it wasn't possible for the kerning to be different, there couldn't be any need to do this, could there? For example, latin.mtx includes \setleftkerning{ffi}{f}{1000} and \setrightkerning{ffi}{i}{1000} to get the kerning of ffi with other characters right.
    – cfr
    Sep 15, 2014 at 14:43
  • Why is this tagged dvips, by the way? It is in no way specific to that but is relevant to all cases where standard TeX fonts are used, including pdfTeX etc.
    – cfr
    Sep 15, 2014 at 14:52
  • I tried adding (LABEL C 0)(KRN O 174 R 2)(STOP) to the ligatures of cmr12; then 0-, 0-- and 0--- show no kerning; if I change 174 in 55, the kern is added to all three instances. Note: I used cmr12 because cmr10 is preloaded, so a new TFM is not read in.
    – egreg
    Sep 15, 2014 at 15:02
  • 1
    @cfr The TeXbook, of course; cmr10 is on page 427. But also pdflatex nfssfont can be handy: you interactively set the font you want the table of. Type cmr10 at the first prompt and \table\bye at the second one. The commands are the same as described in the METAFONTbook for testfont.tex, but with nfssfont you can also call a font by LaTeX attributes: just hit return at the first prompt instead of supplying a TFM name and you'll be presented with questions for encoding, family, weight, shape and size.
    – egreg
    Sep 15, 2014 at 21:57

1 Answer 1


My experiment, described below, seems to prove that this is impossible. When TeX finds the cluster of characters


(where abc represents any three characters) it looks whether a kerning pair or ligature for ab is defined and, in this case it inserts the kern or makes the ligature. If the ligaturing mechanism has substituted d to ab, a kern or ligature between d and c is tried; otherwise the pair bc is considered.

Here's the description of the experiment.

  1. I converted cmr12.tfm to cmr12.pl with tftopl.

  2. I added the lines

    (LABEL C 0)
    (KRN O 174 R 2)

    to the LIGTABLE property, which should define a kerning pair between 0 and the em-dash and ran pltotf.

  3. I ran TeX on the test file

    \font\x=cmr12 \x % load the modified font metric file

    and no kern showed up.

  4. I changed the lines above in the cmr12.pl file into

    (LABEL C 0)
    (KRN O 55 R 2)

    (octal 55 is the slot for the hyphen), converted again to a .tfm file and ran again TeX on the test file above; in this case the kern showed up on all three lines.

The situation after a ligature is different and, indeed, cmr12.pl shows the same kerning pairs for f and octal 13 (ligature ff), to wit

   (LABEL C f)
   (LIG C i O 14)
   (LIG C f O 13)
   (LIG C l O 15)
   (KRN O 47 R 0.069734)
   (KRN O 77 R 0.069734)
   (KRN O 41 R 0.069734)
   (KRN O 51 R 0.069734)
   (KRN O 135 R 0.069734)
   (LABEL O 13)
   (LIG C i O 16)
   (LIG C l O 17)
   (KRN O 47 R 0.069734)
   (KRN O 77 R 0.069734)
   (KRN O 41 R 0.069734)
   (KRN O 51 R 0.069734)
   (KRN O 135 R 0.069734)

which confirms my preliminary analysis.

Of course a kerning pair between 0 and the en-dash can be specified, but it will come into action only if


is input (maybe via direct input of the en-dash with a mechanism similar to what inputenc does for LaTeX).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .