: Implementing the Dutch IJ digraph as a ligature I'm designing a display typeface and having some trouble deciding how to implement the uppercase Dutch digraph "IJ". I have designed a ligature

@Turnbaugh909

Assuming that your IJ glyph is appropriate to the Dutch eye (which I cannot tell), it should be used as a ligature for I+J by default for Dutch texts.

While you can restrict every OpenType feature to specific locales, the arguably best way to do this would be via the feature locl, which is specifically reserved for such purposes. The OpenType cookbook uses this very case as an example:

feature locl {

script latn;

language NLD exclude_dflt;
lookup DutchIJ {
sub IJ by IJ.dutch;
} DutchIJ;

} locl;


If the user properly sets the locale of the text, this should ensure that the ligature is used in Dutch text, but not for any other language. (Sidenote: It should not be activated for Afrikaans, as the absence of IJ is one of the distinguishing features of Afrikaans orthography.)

I do not see any reason to not use your ligature for the IJ Unicode character (U+0132), as there is no legitimate use of this character outside the Dutch language and in Dutch it should not be used for the exceptions.


According to Wikipedia, an "I" followed by a "J" doesn't always form a digraph; would the ligature still be acceptable in those cases, or would that be considered incorrect?


It would be considered incorrect, but such cases are the exception. Also such exception exist for almost every ligature. E.g., in German, you would not have the typical ligatures involving the letter f in the composite words such as Auffahrt or Kaufleute.

There are two ways to deal with this:


The user has to use a zero-width non-joiner (ZWNJ) to suppress the ligature in cases, where it’s not intended, i.e., write FI|JI instead of FIJI (with | indicating the ZWNJ). If you expect your fonts to be used by non-professionals, you should arguably instruct them on this.
The exceptions are hard-coded into the font. Given that exceptions are rare in this case, this is rather feasible. In this article (which is a great source on the IJ in general, by the way), the Dutch type foundry Underware states that this is what they are doing. While such a hard-coding may capture the vast majority of exceptions, it cannot be perfect, e.g., it cannot possibly cover all proper names in the world.

Also, this is arguably nothing that should be handled by font features, but by dictionaries and the end user.

Vote Up Vote Down