🔎 Playlist item offsets

[jubabrut]

Describe the feature

When I try to change position or length of an audio clip in the Playlist, with that code

tracks = [track for track in project.arrangements.current.tracks]
tracks[2].__getitem__(0).position = 100

I have this error
pyflp.exceptions.PropertyCannotBeSet: Event(s) () was / were not found

It will be interesting to be able to change the position, the length and the offset of an audio clip

What version of PyFLP are you using?

last commit (27th oct 22)

Screenshots, Additional info

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[demberto]

While actually I have not implemented this, the exception was a result of a bug in here:

PyFLP/pyflp/_descriptors.py

Lines 121 to 129 in 5034910

	@final
	def __set__(self, instance: Any, value: T):
	if self._readonly:
	raise PropertyCannotBeSet(*self._ids)
	event = self._get_event(instance)
	if event is not None:
	self._set(event, value)
	raise PropertyCannotBeSet(*self._ids)

This was quite a big bug making nearly everything unsettable. This reminds me of testing the setters as well.

[demberto]

Technically it is settable via the huge PlaylistEvent but that needs low-level API knowledge

[demberto]

Done, implemented in latest commits. New release comming soon 🎉

[jubabrut]

Thanks
perfect it works too
what is the difference between the length and the offset (end minus start) ?

Moreover I don't understant the "end value"
because I created the smallest audio clip possible

And this is the offset value
offset : (0, 1084851705)
length : 1

[demberto]

Start Offset:

Length is the length of the item as it is visible to you in the playlist.
End offset is the internal position from the end of a sample or a pattern. Now you might think that start_offset + length is enough for that, why another field?

And the answer to this is, I don't know myself 😄 FLP is a weird format.

[demberto]

@jubabrut Will it possible for you to test PyFLP and find out bugs and suggest new ideas like these? Doing so will be a great help for me and will fasten the release cycle. A stable version will be ready sooner 🤗

[jubabrut]

Do you know how the offset value is construct ?
Right now I don't fina logic with a value like 1174289603

Did you notice a full audio clip offset are (-1082130432, -1082130432), so weird

For the moment I try PyFLP and discover all features, that's why I find some bugs on different functionnalities.
But yes I wiil try to pass more time on this project in order to help you to increase the release cycle and find bugs.

What is your goal with this library? Because it is a huge work
For me my wish list will be to be able to generate music with AI but it seems complicated today to created channel, pattern, etc. directly with the program

[demberto]

Do you know how the offset value is construct ? Right now I don't fina logic with a value like 1174289603

Did you notice a full audio clip offset are (-1082130432, -1082130432), so weird

I need to take a look at it, maybe its incorrectly parsed, maybe it is a float, maybe it is not an absolute value at all.

What is your goal with this library? Because it is a huge work

At first, I aim to make all existing structures parseable and editable. I try to find the representations of various quantities like the TimeStretching.time feature you had problems with. That is the hardest part.

I am also looking at the ability to clone or add certain entities like channels or MIDI notes for example. Once things like channels can be cloned, you can play with the values and modify them to your liking.

Some particular areas needing work are:

• MIDI and Internal controllers
• Patterns
• Automation clips

I generally try to add features when I am genuinely sure that the integrity of existing data is not compromised. PyFLP guarantees to not touch the data you didn't modify. FL studio on the other hand, rewrites everything from scratch, since it knows how the format works entirely.

PyFLP is not an FL Studio replacement. It can most certainly be used in project file converters and real-world DAWs that aim to support FLP format (if they ever get written in Python 😄)

[jubabrut]

Ok because I really don't understand how to use offset value.
Look at this example, that has no sense
PyFLP test 4.zip

Track 0
1536 ==> length
(-1082130432, -1082130432)
Difference: 0 ==> end - start

Track 1
384
(0, 1134572607)
Difference: 1134572607

Track 2
384
(1134572607, 1142961215)
Difference: 8388608

Track 3
384
(1142961215, 1148210782)
Difference: 5249567

Track 4
384
(1148210782, 1151349637)
Difference: 3138855

About TimeStretching.time it is interesting because in fact I searched a solution by using Python, to solve a problem I have with FL Studio. (A bug is open by Image Line because strange behovior happens when you switch a sample with different BPM. Position, length and offset change randomly).

[demberto]

(A bug is open by Image Line because strange behovior happens when you switch a sample with different BPM. Position, length and offset change randomly).

FL has grappled with timing issues since eternity 😄

Ok because I really don't understand how to use offset value.

I think its interpreted incorrectly (might be a float, might be PPQ dependant, dunno). But atleast the start, end values seem to match up with the arrangement. I will take a look at this. You will have better luck if you look at the individual bytes instead of these integers, use int.to_bytes(value, 4, "little")

[jubabrut]

I really don't understand how to use to_bytes
it gives me something like

Track 1
b'\x00\x00\x00\x00'
b'?4\xa0C'

Track 2
b'?4\xa0C'
b'?4 D'

Track 3
b'?4 D'
b'^NpD'

[demberto]

Yes sorry I meant

tuple(int.to_bytes(value, 4, "little"))

which will give you something like

(1, 2, 3, 4)

Now you get 4 numbers instead of one and maybe it can help you find out some relation between the arrangement and them.
Although I don't think there's anything useful out of these numbers, maybe they represent a quantity I don't know about.

[jubabrut]

do other python libraries provide the same result for offsets?

[demberto]

Wdym? There are no other Python libraries which can parse FLPs to the best of my knowledge. This implementation of playlist item offsets is from MonadGroup's FLParser which parses it as a float or uint32 as well.
I think i initially parsed it as float as well and got similar results. The playlist items are locked to the resolution of the PPQ (unlike automation points, which are completely free), so I think float would be too accurate for that and waste space

[jubabrut]

I was just thinking about FLParser but I didn't test that library
Thanks again for the time and your patience.

[jubabrut]

From my first constatation, the offsets values indicate the time where to cut the sample, by not taking account the arrangement, the general bpm or the time stretching.
See below it is the same samples (the second one is time stretched) but cuted on the same location as the first one.
And the offset end values are the same.

It is an element to reply to the difference between offset and length.

So I need to found a relation between
offset end 1138116901
the real cut at: 0,42857143 seconds (exactly in my example)

[demberto]

Interesting

[jubabrut]

Problem solved here, it's a float32 representation
thanks to chrslg

https://stackoverflow.com/questions/74246934/what-relation-between-10-digits-and-datetime

[demberto]

Oh quite some discussion going over there. I did try float32 once, but got some weird results. time to retry. Thanks for taking the time to research :))

[demberto]

@jubabrut https://chat.stackoverflow.com/transcript/message/55452241#55452241