pmcharrison / hrep Goto Github PK

It would be good to add a method for audio synthesis with sox. It makes sense to synthesise tones according to the idealised harmonic expansion used in pi_sparse_spectrum.

Example sox command: play --norm=-3 -n synth 2 pluck A2 pluck A3 pluck A4 channels 1

expand_harmonics() is slow - it should be optimised in C++.

Best to remove the c() method for chord representations, as it can yield unexpected results for users assuming that it'll behave like a normal numeric vector.

smooth_pc_spectrum(pc_set(0), force = TRUE)

fails with error

Error in assert_force(force) : 
  cannot deterministically convert this object to a 'pi_chord' representation. To force such a conversion, set force = TRUE.

If core methods such as .pc_set and pc_set were written in C++ this would really speed things up for batch conversion operations. This should be a development target once the API is definitively pinned down.

Currently an as.data.frame method exists for milne_pc_spectrum; we could also add as_tibble and as.tibble methods.

metadata of a sequence (coded_vec) contains $keysig (a signed integer) and $mode (integer, mostly 0 or 9?) but their mappings are not explained anywhere. It would be useful if I can use $keysig to transpose pitch class chords with respect to a reference pitch class (e.g., C=0) but the relationship between the $keysig value and the actual key is unclear (e.g., 'bach-chor001', which seems to be Riemenschneider No.1 (BWV 269), has a $keysig=1 but is in G major. A clear explanation would be much appreciated.

It would be better to rename view() to plot() to avoid clashes with tibble and the tidyverse.

The view() method has yet to be enabled for coded vectors.