I know. This then implies the NR is in error. Do we report a bug? [For the reason that being thrown off the scent I spent hours looking all in the interfaces to no avail.]
I think this is absolutely a bug. It should not say shorten-pair applies to TextSpanners when it does not.
There are three possible solutions I can see:
- Fix the description of shorten-pair to indicate which spanners it applies to
- Create a new interface e.g. shortenable-spanner-interface that contains shorten-pair, and then give that interface to all the grobs that accept shorten-pair (see dynamic-line-spanner-interface
for a similar single-property interface).
The padding works. Seems inconsistent to me. Why do we leave TextSpanners out of the group?
- Add the shorten-pair property to the spanner-interface (or maybe the text-spanner interface) and make sure it works with all of the objects having that interface.
My preference right now would probably be #2.
I have not made an exhaustive study of this, but my initial thought is that there are two kinds of spanners.
Spanner type A goes from one grob to another (e.g. TextSpanner, glissando). These spanners can be shortened by adding padding, because they will need to move away from the grobs that they connect when the padding is increased.
Spanner type B goes from one column to another (e.g. hairpin, pedal bracket). These spanners don’t have grobs at their ends, so padding will not be effective at shortening them. Hence the need for another method, as met by the shorten-pair
property.
One could add the shorten-pair property to the type A spanners, but it’s not necessary to do so to shorten them.
I don’t know what I prefer for the long-term solution. It would take more study and consideration of the pros and cons of each method.
Thanks,
Carl