Liquidsoap

source('#a,'b,'c) ?

Details

  • Type: Sub-task Sub-task
  • Status: Closed Closed
  • Priority: Important Important
  • Resolution: Fixed
  • Affects Version/s: 1.0
  • Fix Version/s: None
  • Component/s: Liquidsoap
  • Labels:
    None

Description

It may be something that I do not understand. but I think that operators like output.ao should require a source of type
  source('#a+1, 'b, 'c)
i.e. should not have to care about anything concerning video and midi..

I do not think that the current type system can express it but I may be wrong..

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
David Baelde added a comment -
The type system can express this. However, the difficulties caused by variability would apply even for an operator that only looks at the audio channel. This is because changing a content layer in a frame is a global operation for all channel types: if you want to change the video layout you need to change the global layout. So the operator would have to deal with data spread over multiple layers.

This makes me think of something trickier, related to the mismatch between local/precise types and the global/rough layer system. If you take such a source (assuming we can build one, perhaps by muxing?) and drop the video and midi channels, you get a ('#a,0,0) but I don't think that the layers will have been glued back together...

We'll have to discuss this more. I'm not sure how important this is and how we should address it.
Show
David Baelde added a comment - The type system can express this. However, the difficulties caused by variability would apply even for an operator that only looks at the audio channel. This is because changing a content layer in a frame is a global operation for all channel types: if you want to change the video layout you need to change the global layout. So the operator would have to deal with data spread over multiple layers. This makes me think of something trickier, related to the mismatch between local/precise types and the global/rough layer system. If you take such a source (assuming we can build one, perhaps by muxing?) and drop the video and midi channels, you get a ('#a,0,0) but I don't think that the layers will have been glued back together... We'll have to discuss this more. I'm not sure how important this is and how we should address it.
Hide
Permalink
Romain Beauxis added a comment -
Thanks for the comment, I did not realize the internal complexity..

All in all, it seems to me that variable sources will have to get converted to sources with fixed values at some point in any reasonable script..

Should I close this issue?
Show
Romain Beauxis added a comment - Thanks for the comment, I did not realize the internal complexity.. All in all, it seems to me that variable sources will have to get converted to sources with fixed values at some point in any reasonable script.. Should I close this issue?
Hide
Permalink
Samuel Mimram added a comment -
Shouldn't we merge with LS-396 which is a general "we should check types" bug report?
Show
Samuel Mimram added a comment - Shouldn't we merge with LS-396 which is a general "we should check types" bug report?
Hide
Permalink
David Baelde added a comment -
This particular point about output.ao() is not an issue, let's not accumulate sub-tickets.
Show
David Baelde added a comment - This particular point about output.ao() is not an issue, let's not accumulate sub-tickets.

People

  • Assignee:
    Unassigned
    Reporter:
    Romain Beauxis
Vote (0)
Watch (0)

Dates

  • Created:
    Updated:
    Resolved: