I would rather use Python with the FontTools. You can get a working version with AFDKO, the Adobe font development kit.

Here is the kind of code that, once AFDKO is installed, would probably work with your files, given what you say:

---- cut here
#!/usr/bin/env AFDKOPython

from fontTools.ttLib import TTFont
import sys

filename=sys.argv[1]
ofilename=filename[:-4]+"_new.ttf"

idstocopy = [1,2,3,4,5,6,7]

names={}
f=TTFont(filename)
for n in f["name"].names:
if n.nameID in idstocopy:
if n.platformID == 3 and n.platEncID == 1:
names[n.nameID] = n.string.decode('utf_16_be')
for n in f["name"].names:
if n.nameID in idstocopy:
if n.platformID == 1 and n.platEncID == 0:
n.string = names[n.nameID].encode('mac_roman')
f.save(ofilename)
---- cut here

If you save that script in an executable file adjnames, then adjnames font.ttf creates a font font_new.ttf with the adjusted mac names.

With a simple find, you can then batch process all your files.

Thank you, Michel, for pointing me in the right direction. AFDKO Python is the way to go.

But there is a bug in your script, and I need an enhancement. I hope you can point me to documentation for Python and for its AFDKO extensions.

The bug is that in the first table entry, for Copyright, the value is entered in the Macintosh version as "Copyright" instead of the name of the copyright holder as shown in the Windows version of the entry.

The enhancement is that some of the files (I just chose another arbitrarily) have more than seven table entries, and the extra table entries don't have any Macintosh version at all. It would also be nice if the script could go through the whole folder of TTF files, operating only on the ones that have either no Macintosh entry for the name or have "New" as the Macintosh name.

I could do it myself if I understood the language. Unfortunately, the only programming languages I ever learned to use were Fortran, C (without the ++) and BASIC. But I could learn another.

The fact that some entries may be present for Windows and not for the OS X adds some difficulties. New records need be created. Moreover, if you create a record where there is already one, you get two records for the same parameters (I checked with TTX); that is probably not good.

The only documentation I could find is on the Microsoft site and I found snippets with Google. Not fantastic. Here is some improved code that handles missing records for the mac and takes care of the US English language. This time, the font file is overwritten. That means that if you put that script in your path and call it adjmacnames then

find . -name "*.ttf" -exec adjmacnames "{}" \;

recursively modifies all the fonts in the current directory. You can add values to idstocopy without problem because the code now checks what can be overwritten and what record is to be added. If you find improvements (or some decent documentation), please tell us

#!/bin/sh

echo "Processing $1 ..."
AFDKOPython <<EOF
from fontTools.ttLib import TTFont
from fontTools.ttLib.tables._n_a_m_e import NameRecord
import sys

idstocopy = [1,2,3,4,5,6,7]

winnames={}
macnamerec={}

f=TTFont("$1")

for n in f["name"].names:
if n.nameID in idstocopy:
if (n.platformID,n.platEncID,n.langID) == (3,1,0x0409):
winnames[n.nameID] = n.string.decode('utf_16_be')
if (n.platformID,n.platEncID,n.langID) == (1,0,0x0):
macnamerec[n.nameID]=n

for i in winnames:
if i in macnamerec:
macnamerec[i].string = winnames[i].encode('mac_roman')
else:
myRecord = NameRecord()
myRecord.nameID = i
myRecord.platformID = 1
myRecord.platEncID = 0
myRecord.langID = 0x0
myRecord.string = winnames[i].encode('mac_roman')
f['name'].names.append(myRecord)
f.save("$1"+"tmp")
EOF

mv "${1}tmp" "$1"

Michel: thanks for the help.

I found the bug: the list of "idstocopy" has to include 0 (zero) in order to overwrite the Copyright entry.

Better yet: omit the list of "idstocopy" and the corresponding "if" statements; the script will overwrite any existing Macintosh entries and do nothing if the Macintosh entry does not exist.

However: I will avoid complications and not try to create new Macintosh entries. The fonts won't be distributed outside my own computer and I don't need to have the designer, license, etc. in Mac format. Really, all I need is distinct names. Thanks anyway. In fact, thanks very much. You've been a great help.

I still don't understand the import statements and the object member names, but that can wait till I need to know.

The import statement simply allows using in your program functions that are not provided by default by Python but are accessible via packages. For instance, to get the square root of 2, you may type in Python

import math
math.sqrt(2)

or, if you want to type directly sqrt(2) without specifying each time that you are using the sqrt function of the math package, you can type

from math import sqrt
sqrt(2)

You simply tell python where to fetch the functions you are using. Different packages could define functions with the same name but that provide different funtionalities.

Michel

One more comment. Some of the TTF files can't be processed by the adjnames script, giving errors like "UnicodeEncodeError: 'charmap' codec can't encode character u'\x99' in position 7: character maps to <undefined>", and a *_new.ttf file is not created. I propose to fix those by hand, by going into Windows and using my registered copy of Font Monster.

Hmm... What about replacing n.string.decode('utf_16_be') by

n.string.decode('utf_16_be').strip(unichr(0x99)) ?

Or add 'ignore' to encode('mac_roman') as follows:

n.string = names[n.nameID].encode('mac_roman', 'ignore')

Michel, that helped for a few font files, but it left about ten or fifteen of them still not properly named for the Mac.

At that point, doing what works (even if it has to be done one file, one parameter at a time) is more efficient than trying to tweak a batch method. I went into Windows NT on a virtual machine and used Font Monster.

Font Monster assigns one parameter at a time, but CTRL-C, CTRL-V made it easy to copy the MS values to the Mac parameters. Other than that, its main drawback is that (having been last updated in 1994) it doesn't do long file names. Since all the faulty font files have 8.3 file names (probably dating from the same period) that wasn't much of a problem.

Good that you got it right. Incidentally, there was also a way with FontForge, provided FontForge does what you want with the naming table. The idea is just to open your font, say font.ttf and generate a new ttf font, say newfont.ttf with fontforge. You can then extract from it the name table using ttx wich comes with AFDKO; the call

ttx -t name newfont.ttf

generates the file newfont.ttx wich contains only the generated naming table. You can then merge that table into the original font, so that only the naming table is modified.

ttx -m font.ttf newfont.ttx

generates the font newfont#1.ttf which is font.ttf with the new merged name table.
Of course that process can be done in batch provided you can call fontforge from the line command (which is the often the tricky part).

Here is a script that does it provided the Python fontforge files are installed for python2.7 (the one I am personally using with FontForge) and you have ttx installed.

#!/bin/sh

# given as parameter font.ttf
# generates font_new.ttf
# with mac entries copied from
# the win entries George Williams
# considered relevant

python2.7 <<EOF
import fontforge
f=fontforge.open("$1")
f.generate("$1"[:-4]+"_new.ttf")
EOF
ttx -t name "${1%.ttf}"_new.ttf
rm "${1%.ttf}"_new.ttf
ttx -m "$1" "${1%.ttf}"_new.ttx
rm -f "${1%.ttf}"_new.ttx

If the fontforge package is installed for your default python, then call python instead of python2.7.

Now that you mention it, Michel, I could have used FontForge without ttx. When FontForge opens a .ttf it copies the outlines exactly. When I generate a font I usually have the Validate box checked, but if I uncheck it and tell FontForge to generate a TrueType font it will (I think) simply spit out whatever outlines it read in. If it upgrades other tables, like hints or kerning, I wouldn't object. As I said, these are mostly crappy fonts.

I did that with one font. The output looked the same as the original in the (low res) Font Book preview (bad spacing included), and the files were the same size (as reported in Kbytes). Font Monster, FontForge, and your Python script with "ignore" all ignored the 0099 character in the original Trademark field (which I identified because FontForge displayed it with little numbers).

Next time....