[makeotf] substitution without target item does not fail

[frankrolf]

Unexpectedly, a features.fea looking like this seems to be completely acceptable to makeotf:

languagesystem DFLT dflt;

@letters_lowercase = [ a b c d e f g h i j k l m n o p q r s t u v w x y z ];

feature ccmp {
	sub @letters_lowercase;
} ccmp;

(note no target item in the substitution).
This could be a feature – just making sure it’s not accidental.

Example project attached: subs_no_target.zip

[skef]

I think that syntax just amounts to an implicit "by NULL".

[skef]

Yes, I looked at the code and this is treated the same as if there was an explicit "by NULL", so I think it's OK.

[frankrolf]

This whole scenario was encountered by a student in a FDK workshop – they reminded me of what the effect was. Basically, this substitution kills the a in the target font (nothing else):

Still expected?

[skef]

I guess the first question is whether putting in an explicit "by NULL" changes the outcome. If it doesn't, then it's more or less expected. If it does change the outcome then there's a discrepancy.

The other question is whether "by NULL" is implemented correctly for this case.

[frankrolf]

more verbose, as suggested:

sub @letters_lowercase by NULL;

same result, only a is affected

---

try to affect more glyphs than the a

sub @letters_lowercase by NULL NULL;

fails

---

different feature tag

feature smcp {
	sub @letters_lowercase;
} smcp;

no difference

---

different logic

@letters_lowercase = [ a b c d e f g h i j k l m n o p q r s t u v w x y z ];
@null_group = [ NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL ];

feature ccmp {
	sub @letters_lowercase by @null_group;
} ccmp;

fails

---

deliberate

feature ccmp {
	sub a by NULL;
	sub b by NULL;
	sub c by NULL;
	sub d by NULL;
	sub e by NULL;
	sub f by NULL;
	sub g by NULL;
	sub h by NULL;
	sub i by NULL;
	sub j by NULL;
	sub k by NULL;
	sub l by NULL;
	sub m by NULL;
	sub n by NULL;
	sub o by NULL;
	sub p by NULL;
	sub q by NULL;
	sub r by NULL;
	sub s by NULL;
	sub t by NULL;
	sub u by NULL;
	sub v by NULL;
	sub w by NULL;
	sub x by NULL;
	sub y by NULL;
	sub z by NULL;
} ccmp;

works

---

sneaky NULL

@letters_lowercase = [ a b c d e f g h i j k l m n o p q r s t u v w x y z ];
@null_group = [ a b c d NULL f g h i j k l m n o p q r s t u v w x y z ];

feature ccmp {
	sub @letters_lowercase by @null_group;
} ccmp;

fails

[anthrotype]

this is documented here: https://adobe-type-tools.github.io/afdko/OpenTypeFeatureFileSpecification.html#5a-gsub-lookuptype-1-single-substitution

If the replacement glyph is the reserved word NULL, then the substitution has no replacement, removing the input glyph from the glyph sequence:
substitute a by NULL;
Omitting the by clause is equivalent to adding by NULL.

[frankrolf]

Thank you Cosimo. I guess my original question is answered there, however some of the results (only the first item of a group is replaced) are surprising.

[skef]

however some of the results (only the first item of a group is replaced) are surprising.

Yes, this seems like it's a bug (assuming it's something we can fix, which seems likely)