Regression with addMetaNote that effects time signatures, fermatas and tempo

[cianoc]

I first noticed this with the bartok example where it's all in common time. Here's an example:

{-# LANGUAGE OverloadedStrings #-}

import Music.Prelude

main = defaultMain music

music = let 
    meta = title "Test 123"
        . composer "Some guy"
        . timeSignature (time 3 4)

    in meta $ pseq[c, d, e, f, g]

[cianoc]

%%% WARNING: GENERATED FILE. DO NOT EDIT. %%%

\version "2.20.0"

\paper {
    #(define fonts
       (make-pango-font-tree "Times New Roman"
                             "Arial"
                             "Andale Mono"
                             (/ staff-height pt 20)))


    % First page vs. subsequent pages
    indent          = 0.3\in
    short-indent    = 0\in
    % horizontal-shift

    % paper-width
    % left-margin
    % right-margin

    % Horizontal space taken up by music
    line-width      = 210\mm - 2.0 * 0.4\in

    page-breaking = #ly:one-page-breaking
}

\header {
  title         = "Test 123"
  composer      = "Some guy"
  tagline       = ""  % removed
}

\layout {
}

#(set-global-staff-size 10)

<<
    \new StaffGroup <<
                    >>
    \new StaffGroup <<
                    >>
    \new StaffGroup <<
                    >>
    <<
        \new Staff {   \set Staff.instrumentName = "Piano I" \set Staff.shortInstrumentName = "Piano I" \clef treble {   \time 4/4 c'1-\mf
                                                                                                                     } d'1 e'1 f'1 g'1
                   }
    >>
    \new StaffGroup <<
                    >>
    \new StaffGroup <<
                    >>
>>

[cianoc]

[cianoc]

So I'm happy to try and take a go at this - but I'm probably going to need some pointers as where to look.

[cianoc]

Tempo is also being ignored. This bug also seems to occur with musicxml export, though the musicxml file currently crashes musescore for unknown reasons :(

[hanshoglund]

This looks like a regression. I would go about it like this:

• Check if the problem extends to StandardNotation (the pipeline is Music -> StandardNotation -> [Xml/Ly]).
• Depending on the outcome, continue to "bisect" the pipeline to find the exact spot where the information disappears. You can use MonadLog for debug printing, or rely on tests an REPL to test the pure functions.
• Finally I would add a golden test (e.g. using tasty-golden or hspec-golden) to prevent the issue from reappearing.

Let me know if you need any further pointers!

[cianoc]

Ah thanks! I've fixed the time signature bug (caused by a regression due to the Option datatype being deprecated - which is a little weird, but a trivial fix). But now I've discovered a wider issue with addMetaNote that I'm looking into.

[hanshoglund]

Reopening as fermatas and tempo needs to be resolved too.

[cianoc]

So I've looked into this some more and for the code where tempo would be generated you have "TODO add tempo". So I'm guessing this got removed in a refactoring. I couldn't see anything for fermata.

Is there any reason for tempo to be tied to a barline like it used to be? Obviously it makes it easier to code, but it does feel a little limiting. Particularly as fermatas will be independent of the barline and I have to work out how to code that anyway.

[hanshoglund]

I looked through the code and it seems these features exist in the Lilypond AST (Music.Data.Lilypond). There was a bug in the Lilypond pretty-printer which I have fixed in #70/#71.

It looks like the code to render these from StandardNotation.Work where never added to the toLy function. I think this is straightforward to do, following the approach in toXml.

I don't think there is any non-technical reason reason to connect tempo/fermata to barlines.