API: CountEnvelopePointsEx() not returning nr. of points in AI (FIXED)

[nofish]

FIXED (v5.979)

See

https://forum.cockos.com/showthread.php?t=199142

[nofish]

bump

Also, as mentioned in the linked thread, CountEnvelopePoints() doesn't work the same anymore as before the introduction of AI's (when AI's are present on the track).

[cfillion]

Still happens in 5.70.

[nofish]

bump

as this also affects SWS currently.

So at least a statement if this is considered a bug or not a bug / 'live with it' and SWS / scripters should work around it would be nice.

[juliansader]

A related problem is that Automation Items are not included in track or item state chunks. Scripts can therefore not count the points by parsing the state chunk.

[nofish]

Hm..actually I'm not not so sure currently if it's completely fixed.

When I run the following code

function msg(m)
  return reaper.ShowConsoleMsg(tostring(m) .. "\n")
end


track = reaper.GetTrack(0,0)
env =  reaper.GetTrackEnvelope( track, 0 )

countPts_underlyingEnv = reaper.CountEnvelopePointsEx(env, -1);
countPtsAI_1 = reaper.CountEnvelopePointsEx(env, 0);
countPtsAI_2 = reaper.CountEnvelopePointsEx(env, 1);
countPtsAI_3 = reaper.CountEnvelopePointsEx(env, 2); -- doesn't exist

msg("Points in underlEnv: " .. countPts_underlyingEnv)
msg("Points in AI1: " .. countPtsAI_1)
msg("Points in AI2: " .. countPtsAI_2)
msg("Points in AI3: " .. countPtsAI_3)

I get this:


Could it be that the point count in AIs is (wrongly) doubled somehow ?

(R5.78)

edit:
Point count in AIs seems erratic.
I just retried above script with a fresh portbale R5.78 (32 bit) install in a project from scratch and get this:



In first AI it seems ok, second one seems doubled.

[Justin]

Duplicated.

This appears to be that it includes neighboring loop iterations in the envelope point list, sometimes even when those iterations are not visible. For now as a workaround you can call reaper. GetSetAutomationItemInfo() to get the start/length of the item, then loop through the points to see which ones are actually within those bounds.

[nofish]

Thanks Justin.

Maybe not necessary to post the .rpp as reproduced here from scratch ?
(fresh portable R5.78 32 bit)

Point count in second AI seems doubled:



edit:
Missed your edit, nevermind.
Thanks for looking into it.

PS:
Sorry for speaking to soon / sloppy testing.

[Justin]

Sorry for the confusing edit!

Looking at this more, we need to keep extra points outside of the visible area (even when looping) for obscure reasons (and also if bezier points are used), so it might just be good advice to check the points against the bounds of the container to see if they are meaningful...

[nofish]

Ok, so it sounds like it'll stay this way (for the forseeable future) ?

Asking because I noticed this when working on SWS stuff and now wondering if I should use your suggested workaround or wait for a change / fix ?

[Justin]

Yes, definitely use the workaround -- it might change at some point, but if it does it would still often contain points that are outside the visible area (so that linear or bezier transitions can be interpolated), so in this case you'd want to validate the points anyway.

[nofish]

Ok, thanks.

[nofish]

Originally Posted by Justin

For now as a workaround you can call reaper. GetSetAutomationItemInfo() to get the start/length of the item, then loop through the points to see which ones are actually within those bounds.

One question about this, sorry.

When I have a situation like this:


Left edge upper point is in underlying envelope, lower point is in AI, but visually the seem (alomost ?) at the same position. How should I do the bounds check to only and reliably catch the point within AI ?

Can I assume point_pos within AI is always > AI start and < AI end or can they fall exactly at AI bounds too ? (hope it's clear what I mean ?)

[Justin]

Originally Posted by nofish

One question about this, sorry.

When I have a situation like this:


Left edge upper point is in underlying envelope, lower point is in AI, but visually the seem (alomost ?) at the same position. How should I do the bounds check to only and reliably catch the point within AI ?

Can I assume point_pos within AI is always > AI start and < AI end or can they fall exactly at AI bounds too ? (hope it's clear what I mean ?)

If you only want the points within the AI, you should just request the points for that AI, then bounds check them to >= AI start and < AI start+AI len. You might need to look at the next points on each side if you need to transition to those values too?

[nofish]

Originally Posted by Justin

then bounds check them to >= AI start and < AI start+AI len.

This is the bit of info I needed, thanks!

[nofish]

Another thing...

I'm trying to count the actual visually present points in this envelope.
This is with 'AIs attach to underlying env. on both sides' enabled:

function msg(m)
  return reaper.ShowConsoleMsg(tostring(m) .. "\n")
end

track = reaper.GetTrack(0,0)
env =  reaper.GetTrackEnvelope( track, 0 )

countPts_underlyingEnv = reaper.CountEnvelopePointsEx(env, -1)
countPtsAI_1 = reaper.CountEnvelopePointsEx(env, 0)

countActualPtsAI_1 = 0
AIpos = reaper.GetSetAutomationItemInfo(env, 0, "D_POSITION", 0, false);
AIend = AIpos + reaper.GetSetAutomationItemInfo(env, 0, "D_LENGTH", 0, false);
for i = 0, countPtsAI_1-1  do
  retval, time, value, shape, tension, selected = reaper.GetEnvelopePointEx(env, 0, i)
  if time >= AIpos and time < AIend then
    countActualPtsAI_1 = countActualPtsAI_1 + 1
  end
end

pointsTotal = countPts_underlyingEnv +  countActualPtsAI_1

msg("Points in underlEnv: " .. countPts_underlyingEnv)
msg("actual Points in AI1: " .. countActualPtsAI_1)
msg("points total: " .. pointsTotal)

When I use your suggested workaround with the bounds check the points at AI edges are counted in the underlying env. (ok, as per Options) but also in the AIs (because using the bounds check method) which results in a total point count > the actual point count on this env.

(Background: I'm currently trying to make the SWS envelope actions work with AIs and this would be a problem there as it works with counting / looping through the env. points and caching them for manipulating. So it would cache more points than actually present.)

I'm not sure how I should get the correct total point count in this case.
I think a IsPointContainedInUnderlyingEnvelope(AIidx, ptIdx) could help in this case to avoid 'double counts'.

Or is there another way I'm missing ?

[nofish]

CountEnvelopePoints() seems also give wrong results in some cases:

function msg(m)
  return reaper.ShowConsoleMsg(tostring(m) .. "\n")
end

track = reaper.GetTrack(0,0)
env = reaper.GetTrackEnvelope( track, 0 )
countPts_underlyingEnv = reaper.CountEnvelopePoints(env)

msg("Points in underlEnv: " .. countPts_underlyingEnv)

[mpl]

I would request a change for enumeration points of AI not in visible area, but real count, in case of looped AI it is a number of points in source AI. Other things is math.

Also maybe not related, but there is not any API for dealing (enumerate, getsetchunk, delete) with pooling AI sources (which are in the RPP header).
(It is one of the reason I still not included AI context into InteractiveToolbar yet).

[cfillion]

Originally Posted by mpl

Also maybe not related, but there is not any API for dealing with pooling AI sources (which are in the RPP header).

Being able to get/set the source length would be useful! https://forum.cockos.com/showthread.php?t=202186

[juliansader]

Originally Posted by nofish

CountEnvelopePoints() seems also give wrong results in some cases:

I think REAPER must necessarily add points to the underlying envelope, otherwise there is no node to connect the downward sloping segment to the left with the flat segment to the right.

[mpl]

Originally Posted by juliansader

I think REAPER must necessarily add points to the underlying envelope, otherwise there is no node to connect the downward sloping segment to the left with the flat segment to the right.

If I understand right, something like EvaluateEnvelope() used internally for these cases.