Dez

[giamic]

Introduce a new converter from Rn directly to Dez. Add a couple of tests

NB: This breaks the tonality because I don't remember how to read all the different tonalities from a score in music21. Let's talk about it.

[MarkGotham]

Thanks @giamic !

When music21 parses rntxt ...

• it stores roman.RomanNumeral objects on a stream.Part ...
• which can be extracted with .recurse().getElementsByClass(roman.RomanNumeral) ...
• and on which key and chord are attributes.

So the way to get changes to either is if this_object.attribute != prev_object.attribute. (This can be by keeping a record go the previous state, or there's actually music21 .previous() functionality for this now.) Debatable whether to do this for chord as it's feasible that an exact repeat would be wanted, but definitely for key changes.

Obviously the key signatures on the part are not it at all.

Happy to chat.

[MarkGotham]

i.e. / e.g.

from music21 import corpus
analysis = corpus.parse("bach/choraleAnalyses/riemenschneider005.rntxt")
rns = analysis.recurse().getElementsByClass(roman.RomanNumeral)
previous_key = "FAKE!"
for rn in rns:
    if rn.key != previous_key:
        previous_key = rn.key
        print(rn.measureNumber, previous_key)

[giamic]

Nice! Thanks for letting me know. Then we should already have everything we need. I will update this PR maybe this Wednesday with the full solution!

[giamic]

I have implemented the change you suggested and it seems to work. I have updated the reference dezrann file to the new one, which has the following differences:

• The tonality labels are no longer all at the beginning of the dezrann file. This change can be easily reverted with a few more lines of code, but I thought it'd be preferable to keep the code as clean and simple as possible
• Some of the chord figures have changed because they no longer have to pass through the tab conversion. Please check that all changes make sense because I think that the previous conversion might have been buggy
• The key figures have changed. Now it's no longer Db but D- major, following music21's convention

Let me know if it all looks good to you!

[MarkGotham]

Hi @giamic, thanks for this! Some good progress here. I've added a few small/obvious edits and added a quick, direct rn to dez conversion for discussion (no score required).

As I see it, here is the current status:

1. remove unnecessary intermediate tab state = DONE by you, thanks 👍
2. remove unnecessary score import = under discussion, sketched by me just now
3. remove unnecessary wrangling with measure offsets = under discussion, sketched by me just now

Wrt 2./3. the new rn2dez shows how we could get the offsets directly from the analysis, and even directly from the RNs without recourse to the measure (which dez doesn't use anyway). I argue that we should do this because it helps:

• save on processing
  • score parsing is probably the heaviest part of this process and
  • the .currentHierarchyOffset iterator on one class is especially fast/light).
• avoid any discrepancies there might be
  • between score and analysis (get all data from one place) and
  • arising from counts/addition on offsets via measures.

What do you think? It could stand alone (apart from the Tab logic here) or else be integrated?