BS.2076-2 tracking issue

[tomjnixon]

To Do

• [x] parse version -- in #59

  • how to model this, perhaps BS2076Version(version: int) or OtherVersion(name: str)
  • warn if the version is not standardised
  • error if the version is not supported at all (as BS.2076 makes forward-incompatible changes)

• new audioObject parameters

  • [x] positionOffset -- in #59
  • [x] gain -- in #59
  • [x] mute -- in #59
  • [x] alternativeValueSet -- in #59
    • [x] gain, mute, positionOffset
    • parse alternativeValueSetIDRef
      • [x] check that they are unique within audioProgramme and audioContent
      • [x] check that they are unique within the matched set of audioProgrammes and audioCntents
        • presumably it's ok to reference an audioObject more than once as long as only one alternativeValueSetIDRef is used?
  • [ ] parse gainInteractionRange and positionInteractionRamge and check the gain and positionOffset are within the specified bounds -- parsing is done, but the checking isn't really a v2 feature and is not done
  • [x] check that all these are only specified in leaf audioObject -- in #59

• [x] track uid to channel reference: -- in #59 (originally https://github.com/ebu/ebu_adm_renderer/pull/32)

  • [x] don't add audioChannelFormatIDRef in AudioID -- in #59
  • [x] add support in item selection (specifically channel_format_for_track_uid) -- in #59
  • [x] check for version >= 2 (or empty document) -- in #59

• [x] gain in DirectSpeakers -- in #59

• [x] gain in HOA -- in #59

• [x] gain in Matrix -- in #59

• [x] extend importance handling to DirectSpeakers/HOA/Matrix? -- in https://github.com/ebu/ebu_adm_renderer/pull/71

• [x] fractional time format -- in #59

• [x] gainUnit everywhere -- this is a really bad idea but needed for compatibility -- in #59

  • [x] warn about compatibility when using this -- https://github.com/ebu/ebu_adm_renderer/pull/72/commits/a719ca4b8101ee2eff9f8f43bb29b082bc52eb10
  • [x] check it's only specified in a -2+ document

• [x] change objectDivergence azimuth default from 45 to 0 -- in #59

  • warn about using default in any version -- in #59

things that don't need implementing

• no need to do anything with authoringInformation

• coordinates of screen which have changed are not used

• could remove the "more than one audioProgramme; selecting the one with the lowest id" warning, but it's still useful

• maximum divergence is not validated

• HOA

  • order and degree now override equation (good), so we could remove the exception when equation is specified, but I don't see the need and would have to change it to a warning in -2 anyway

• dynamic metadata for matrix is probably not a priority

misc

• add a mechanism to warn about use of non-standardised features?

  • this is mostly quite obvious, so I don't think there's any need

• warnings above are based on this policy: warn when using features which:

  • can cause silent differences in behaviour when interpreted according to different versions -- i.e. software is not going to break because of a new attribute or changed default, but the output may be wrong

  • and can easily be replaced by equivalent metadata which doesn't have the same problem

[tomjnixon]

all done except for #65 and #66, which are small enough to track separately