#3208 Wrong MultiMeasureRest glyph

[Anonymous]

Issue 3208: compressed MMRs choose correct glyphs

usable-duration-logs now affects only the choice of glyph for the
representation of a single measure. For multiple measures, glyphs for
longer rests up to maxima are used.

This is a second attempt to make it work and has little to do with my
first attempt at https://codereview.appspot.com/333340043 Therefore I
created a new Rietveld issue.

Also the vertical positioning of half and shorter rest glyphs in MMRs
now is handled correctly. BUT: this breaks make check because of a
different positioning of breve rests in TabStaffs. For details see
discussion below.

http://codereview.appspot.com/335410043

Originally created by: colingh...@gmail.com

Thomas Morley reported here:

http://lists.gnu.org/archive/html/bug-lilypond/2013-02/msg00180.html

as follows:

With measures longer than two whole notes LilyPond chooses different
glyphs for MultiMeasureRest.

\version "2.17.12"

\new Staff {
  \compressFullBarRests

  \time 7/4
  R1*7/4
  R1*7/4*2

  \break

  \time 9/4
  R1*9/4
  R1*9/4*2
  \break

  \time 18/4
  R1*18/4
  R1*18/4*2
}

Discussion on the user-list:
http://lilypond.1069038.n5.nabble.com/Glyph-for-longer-MultiMeasureRest-td141438.html

According to Gould LilyPond's default seems to be correct.
Werner objects.

At least there's need for discussion, I think.

Attached an image showing "LilyPond's default" vs "Correct?"

Regards,
  Harm

[Anonymous]

Passes make, make check and a full make doc.

[Anonymous]

Patch on countdown for January 4th.

[Anonymous]

I am going to leave this on countdown as it looks like there is still some discussion. However if I have misinterpreted this, then it can be pushed.

[Anonymous]

I can see from Rietveld that this is not yet quite ready for 'prime time', I will set this back to Review - while Malte gets his clarification. If nothing has changed by the next countdown, I'll put his back to 'Needs Work'.

[Malte Meyn]

Thanks, that sounds reasonable.

[Malte Meyn]

So far I’ve got answer from all four German publishers I contacted but not from Faber Music (or even Elaine Gould). Three of these answers were very clear, I’ll wait for a clarification on the fourth and then summarize what they said and make a patch.

[Malte Meyn]

Here are the answers I got from the publishers:

Publisher 1
• Has no house rule for this situation.
• But recommends to use a brevis rest for 1 measure in 4/2 time; for more measures use a “modern” horizontal beam rest (MMR.expand-limit = 1).
• Alternatively do the same for 4/2 as for other time signatures like 4/4 (“consequently modern practice”).

Publisher 2
• Use the same notation for all time signatures including 4/2 (what Publisher 1 calls “consequently modern practice”).
• MMR.expand-limit = 9

Publisher 3
• Use brevis rest for 1 measure, longa for 2 measures, longa + brevis for 3 and so on (like LilyPond’s default behaviour).
• MMR.expand-limit = 12 (or maybe 8); after seeing my example they voted for 8 (“else it looks like stacking wood”)

After thinking about these answers I would suggest the following:
• Change LilyPond’s default expand-limit to 8 (publisher 3, Gardner Read “Music Notation”) or 9 (publisher 2, Elaine Gould “Behind Bars”) instead of 10 (current default). IMO there are arguments for both 8 and 9:
∘ 8 is one or two (in case of 4/2 time) maxima rests, not one or two maximas + a little bit.
∘ 9 is the largest single-digit number.
→ I’m undecided
• Have a boolean grob property of MMR that controls whether longer rests are used for longer measures (publisher 3, current default) or whether all times are handled the same (publisher 2 and maybe 1). And make the default that it matches current default (after all, LilyPond doesn’t want really “modern” engraving, does it?).
• Now think whether we need usable-duration-logs and if so what it should do:
∘ Control the appearance of single measure rests.
∘ Control the appearance of compressed rests.
→ I would suggest not to use it. According to Gould smaller than whole rests should be placed where notes of that lengths would stand, so if one wants 1/8 rest in 1/8 time just use r8 instead of R8. And I don’t see why anyone would want to disable brevis, longa, maxima rests. (Ok, maybe maxima … in this post there is an example of 8 measures rest that uses two longas instead of a maxima … but after looking at that edition (NBA) I have the impression that they do weird stuff with rests …) But that’s only my opinion.

[Malte Meyn]

Oh, and if we decided to keep usable-duration-logs it would be necessary to fix the staff-position of smaller rests: currently all symbols just hang from the second line so not only quarters and eighths are displaced but halves look like wholes.

[Benkő Pál]

for default expand-limit I vote for 8.
about usable-duration-logs: I also suggest not to use it; however, the main usage would be to use dotless full measure rests e.g. in 3/8 time (so R4. could look like r4 or r2).

[Malte Meyn]

Would you really notate a rest in 3/8 as quarter or half instead of whole or dotted quarter?

Apart from that, publisher 4 has anwered: Basically their answer confirms that we should support both versions. They do the same as publisher 2 for Bach/Handel and later; additionally they set a “1” above single measure rests (see also [#5251]) and have less space between the rest symbols (is there an option to control this distance?). For older music they are ok with the thing publisher 3 and LilyPond do.

[Benkő Pál]

Would you really notate a rest in 3/8 as quarter or half instead of whole or
dotted quarter?

I definitely would not, I just said that as the only argument I can
imagine someone could come up with in support of usable-duration-logs.
Undotted quarter would match the principles of mensural notation
(rests are never dotted and always perfect if mensuration says so);
half would be the result of scaling the "whole rest in 3/4 time"
principle. but those were never used, I guess; Bach uses whole rests,
see the Gloria movement (bottom of page 2 in
http://ks.petruccimusiclibrary.org/files/imglnks/usimg/0/05/IMSLP102259-PMLP04197-Bach_Bm_Mass_parts_278110185_bass.pdf
).

[Malte Meyn]

Undotted quarter would match the principles of mensural notation
(rests are never dotted and always perfect if mensuration says so);
half would be the result of scaling the "whole rest in 3/4 time"
principle.

Mensural notation has perfect and imperfect division, yes, but only for maxima, longa, brevis and semibrevis (= whole) notes/values. So I see no point in having undotted minima and seminiminma (half and quarter) rests: this principle of mensural notation doesn’t apply to these values. Or am I mistaken? Of course, modern notation often shifts durations (f. e. tempus imperfectum cum prolatione major could be written as 6/8 time) but this is “modernized” notation and should not use quarter but dotted quarter rests for semibrevis rests.

[Malte Meyn]

I changed my mind: Allowing shorter rests doesn’t harm—no one has to use positive numbers in usable-duration-logs but if someone really wants to do it even though Gould et al. disadvise, we shouldn’t take away this functionality.

This also means that we don’t need such a boolean grob property: Instead you can just override usable-duration-logs to #'(0) as it has been done in every single discussion about this property that I could find on the internet. Yay, backwards-compatibility! ;)

Of course we could discuss whether it would be nice to have a shorthand like

modernMMRs = \override MultiMeasureRest.usable-duration-logs = #'(0)

next to compressFullBarRests and expandFullBarRests in ly/property-init.ly.

Concerning the default value of expand-limit: This doesn’t really belong here but would be another issue. I’ll send a feature request to the bug list for discussion.

[Malte Meyn]

BTW until 2.15.8 there was a boolean property called use-breve that was replaced by usable-duration-logs, round-up-to-longer-rest and round-up-exceptions. Going back to a boolean property would be something like a regression ;)

[Benkő Pál]

Undotted quarter would match the principles of mensural notation
(rests are never dotted and always perfect if mensuration says so);
half would be the result of scaling the "whole rest in 3/4 time"
principle.

Mensural notation has perfect and imperfect division, yes, but only for
maxima, longa, brevis and semibrevis (= whole) notes/values. So I see no
point in having undotted minima and seminiminma (half and quarter) rests:
this principle of mensural notation doesn’t apply to these values.

Yes, that's why I said "principles", not "rules" or "practice". I
think the same principle works when Bach uses 9/8 in one part, 3/4 in
another (e.g. Jesus bleibet meine Freude in BWV147) or 12/8 in one
part, 4/4 in another (e.g. Mein teurer Heiland in Johannespassion).
But it's definitely not extended to multi-measure rests, as the
examples above prove.

I changed my mind: Allowing shorter rests doesn’t harm—no one has to use
positive numbers in usable-duration-logs but if someone really wants to do
it even though Gould et al. disadvise, we shouldn’t take away this
functionality.

This also means that we don’t need such a boolean grob property: Instead you
can just override usable-duration-logs to #'(0) as it has been done in every
single discussion about this property that I could find on the internet.
Yay, backwards-compatibility! ;)

Of course we could discuss whether it would be nice to have a shorthand like

modernMMRs = \override MultiMeasureRest.usable-duration-logs = #'(0)

Oh, these are all great!

[Malte Meyn]

Issue 3208: compressed MMRs choose correct glyphs

usable-duration-logs now affects only the choice of glyph for the
representation of a single measure. For multiple measures, glyphs for
longer rests up to maxima are used.

This is a second attempt to make it work and has little to do with my
first attempt at https://codereview.appspot.com/333340043 Therefore I
created a new Rietveld issue.

Also the vertical positioning of half and shorter rest glyphs in MMRs
now is handled correctly. BUT: this breaks make check because of a
different positioning of breve rests in TabStaffs. For details see
discussion below.

http://codereview.appspot.com/335410043

[Malte Meyn]

Breaks make check (rest-positioning.ly). For explanation see Rietveld.

[Anonymous]

Passes make, make check and a full make doc.

Reg test diffs attached

[Benkő Pál]

to me this is progress; some of the tests look (now and before) plain
silly to me:
rest-dot-position.ly: aligned longa and breve rests should not touch each other
TAB staves: is it actually expected that a longa and breve rests don't
go from line to line?

[Malte Meyn]

rest-dot-position.ly: The regression here is [#5217] (random placement of dots) and has nothing to do with my patch; of course one could change this touching rests in a separate issue.

multi-measure-rest-staff-position.ly: This is ok IMHO because staff-position = 0 should behave the same for Rest and MultiMeasureRest. But maybe we should add a convert-ly rule that shifts these overrides?

rest-positions.ly (screenshots 2 and 4, the actual error is only in screenshot 4): That’s on purpose but needs fixing nethertheless, this patch should be waiting for [#5258]. Gould doesn’t say anything about whether they should go from staff line to staff line but IMO that’s worth a discussion (not here but at [#5258]). One would have to draw these rests as rectangles instead of taking them from the font.