Hi everyone,

I am starting to use pd's datastructures. But I am very surprised that there is no "drawcircle" object, with argument such as center coordinates and radius.
How do you draw circles ?

By the way, if anyone knows where to find some advanced documentation on data structures, I've used PD's data structures tutorial by MR. Karman, which was very good, but I would like to have more complete information.
Thank you,

Flipp, can you tell me what I do wrong here?

Hmm, no. ...It seems to be all correct.
The GOP-size blowing up can be considered normal, I think (actually I think the struct-stuff is a little buggy...) Everything should be looking ok later on.
Although that didn't happen when I did the same with the patch. ..I use windows and pd0.42.5...
Maybe have a look at the attached patch. I only selected "GOP" and set the x- & y-range from 0 to 299.

...from time to time it would be quite handy to have some developers around here..

http://www.pdpatchrepo.info/hurleur/10.onoff-GOPtest.pd

I'm just curious...does it work?!

Berenger, thanks for posting the s-abstractions. Very nice to learn that an xy-pad in vanilla Pd is actually possible!

Flipp, your GOP version of 10.onoff.pd works here (I happen to be on Linux today), in a sense.The objects appear at microscopic scale, not like the size they have in the original patch. But the key point, which I always overlooked, is the fact that data structures behave like arrays, where 'range' is an important parameter to set. Normally, range is most important for arrays and x-y size is most important for regular GOPs, but in case of data structures, both seem to be important.

Okay, learned a lot from this thread so far. Thanks.

Katja

Yeah, actually the GOP-range works just like the arrays range. If you select the array's canvas-properties there one can see the same behavior. ...I guess that arrays must be something like inbuild structures. If one just knew the name of that structure one could build a vanilla [arraysize]...

Just to point it out, the "range" (in pixels) stands for an imaginary frame inside the subpatch in which all the graphics should be placed. This then is scaled to the "size" of the subwindow. So for a 1:1 scaling one has to meet "range_max - range_min = size"

Jonathan Wilkes is very familiar with the data structure C code in Pd and indeed he once mentioned that 'array' is a data structure preset with limited properties (fixed color and line thickness). He has written a 'canvas get method', which can request all sorts of info from any canvas:

http://lists.puredata.info/pipermail/pd-dev/2011-05/016855.html

So far, the method is not yet implemented in Pd but it is relatively easy to compile vanilla Pd with Jonathan's patch file, and he's done a very cute interactive test patch, featuring data structures again.

Now back to the circle, initial topic of this thread: it would be cool to have an abstraction [complexplane] featuring the unit circle, where you could set a complex number as a point or vector. With such an abstraction, all sorts of visualization demo's could be done, even animations. For example, the working of resonant filters could be demonstrated, or the multiplication of complex roots of unity like used in fft. I'm tempted to start working on it, now that this thread has made some aspects of data structures clearer.

Katja

Hi there,

this discussion was for me a pretext to go further with my data-structure learning.

Here's a patch drawing a polygonal approximation of a circle, with controllable number of vertices and radius.
An array holds the (x,y) pairs of vertices plus one, which is a copy of the first pair.
[plot] does the work, i'll go on further trying other representation objects.

Hope this will save time to somebody. (I mean for those who don't intend to write abstractions )

Nau

http://www.pdpatchrepo.info/hurleur/circle_example.pd

Here is a visualization of the complex plane, with a draggable complex point for which xy coordinates, radius and angle are shown. No animations yet.

Nau, your solution with [plot] is much better than [drawpolygon]. I'll use it in a next version of [complexplane].

Katja

http://www.pdpatchrepo.info/hurleur/complexplane.pd

OK now it is an abstraction [complexplane] with axes, unit-circle and a draggable complex vector which can also be set via inlets. By way of test, an animation of [cpole~] is done.

A pole-zero GUI (like MaxMsp has it) would be possible with similar technique. But the math behind pole-zero plots is not trivial.

Anyway, I'm starting to understand data structures better now. The only thing I could not get done is a fat (draggable) point according to Flipp's description. The complex datapoint in [complexplane] is a square instead but I would like a circular point better. If the two sets of coordinates are identical, the polygon is just not drawn. Flipp, could you find a way to replace the square in [pd $0-radius-template] with a fat (draggable) point?

Katja

Edit: oops, it was drawing ellipses instead of spirals, uploaded an update.

http://www.pdpatchrepo.info/hurleur/complexplane.zip

Hi Katja,

didn't you get the materials of my data structures workshop in Weimar? In one of them there's a small animation of some circles getting bigger - data-s_07.pd, the circles template.

João

Hey João! A real data structure expert is joining in, great. Yes I have your workshop patches. It's magic. Data structures are so powerful, if only you know how to handle them. I completely forgot about the click messages.

It should be possible to make a [led] abstraction, like the Max object, a toggle with the appearance of [bang]. And what about [knob]? Would be cool to build a set of data structure GUI abstractions. In vanilla Pd.

Katja

Hi,

about [led], I'm not sure if J Kreidler already did a similar thing. There was also an object somewhere that was a bang which could be rescaled in both dimensions.

A data-s-gui abstraction library could be nice, yes. I already did a version of [swatch], but it wasn't ready in time to be included in 0.43 (it also has a circle). Does the attached file work?

I myself can't help much now (am too busy), but to give advices is cheap:

• in case you want to do gui abstractions, don't try anything that has too much movement (data-s are tcl, and tcl is slow);
• to avoid memory space conflicts from using several templates with the same name, it's better to use $0 in all templates+scalars. for that you should create the scalars at loadbang, e.g. like I do in [swatch], [jmmmp/pd-colors] (a good resource for anyone using colors with data-s) or in [jmmmp/bezier] (a more complex example)

Although data structures can be very powerful as a composition/data display tool, since Pd was started [and apparently Pd started because of them] they didn't evolve that much. People can do lobbying to Miller, to see if the puts some more work into them. I already sent some mails several times, suggesting the simple operations I usually mention in my workshop, which require lots of coding.

Another thing that might be looked up by a programmer can be, porting the ftm library from Max/MSP. It's much more robust and mature than Pd's data structures, and, quite frankly, we would all win if we could use this - http://ftm.ircam.fr/index.php/Main_Page and http://sourceforge.net/projects/ftm/.
Besides all the data display+handling apparatus, ftm also has some very good audio objects, which use the data stored by the libraries. IOhannes already did some work in the port, but apparently it stopped.

João

@jmmmp said:

didn't you get the materials of my data structures workshop in Weimar?

Is it possible to download it from somewhere ?
I couldn't find it on the net, and it seems very interesting !

Thanks,

Nau

Hi Katja,
attached is the edited patch. I have created the point and I "styled" the radius-line accordingly (hope that's after your fancy). (And I adjusted some small things). The point is draggable, although the zone in which it reacts to the mouse is rather small. ..maybe it interferes with the other drawings?!

Another way to draw the point would be a separate [struct] (and scalar) for it (eg. like you did with the unit-circle), but that's more complicated. ...and can be even more cpu-intensive, I think..

Hey João,
could you share your workshop-patches overhere as well, please?!
And is there a reason why you just go to "899" instead of "999" for the color in your pd-colors patch?? I thought rgb-values go up to 999!?!
...and I can't find your bezier-patch anywhere, seems to be missing...

Flipp

http://www.pdpatchrepo.info/hurleur/complexplane2.pd

Flipp, thanks for revising my patch. The background-foreground style is nice indeed, gives better contrast. But now the surprise: on my computer (OSX) the fat point is invisible! I tried Pd 0.42 and 0.43. Apparently it is not a matter of Tk version, but what else?

Anyway, the draggable area should be comfortable so it is probably better to make a scalar with more coordinates. The circle in João's [swatch] is a good example. By the way, [swatch] & co make my Pd crash when closing the patch, don't know why, these patches are quite elaborate.

João, it is true that data structures are CPU-intensive, because they are Tk which is an interpreted language. But isn't it the same with Juce? I remember the same frustration from using MaxMsp. With the upgrade from 4.6 to 5 the graphical facilities were greatly expanded but it got no faster.

Data structures are attractive for me because of the fact that it is all vanilla Pd. I am planning to collaborate on a math&music programme for high school kids, and to make it work on arbitrary school computers, we don't want the complicated dependencies of Pd-extended. CPU performance is not so much of an issue, as the patches will be small demo's only. On the other hand, if we want real flashy GUI, we'll probably resort to OpenFrameworks and libpd (but I still have to learn that).

Katja

Hi,

here are the patches of the workshop. There are other things I wanted to program, but that will have to wait until another workshop comes through...

About [pd-colors], I think I didn't make it until 999, because afaik after 899 the last row just repeats itself. Doesn't matter how high you go, it just repeats the last row always.
You can check it yourself by creating more scalars with [until]. If I remember correctly, if you just change the number before [until], it creates more scalars, positioning them in the right place.

Can you check if [bezier] comes with pd-ext-0.43? I think it did come with the version I have here.

Suggestions/critique is always welcome.

João

http://www.pdpatchrepo.info/hurleur/data-s.workshop.jmmmp.zip

Hi Katja,

I don't know Juce, so can't do any comparations.
Strange, I never had any problems with [swatch]. I noticed that patches with data structures crash more often in ubuntu than in windows (which is strange, since Pd is developed in unix), specially when closing them - at least with other more complex patches I have. But as I remember, never had that problem with these small abstractions. Is there any way of getting an error report, and send it to the developer gurus?

I will also look at your plane patch as soon as I can, unfortunately am too busy these days.

Best,

João

Hi Katja,

just a couple of small details about the complex plane patch (I can just look at it for some minutes):

• uau, that's a complicated circle you drew. Maybe you can adapt the circle I did in [swatch]?
• the order of creation of the scalars is the only way to determine which are shown on top of others. Probably you already know that, but just to make sure.
• there's one bug bound to happen, that you also see documented in [bezier]: when the grab points go outside the gop frame, the data structure will stop responding. It will still work, but you won't be able to grab it, you have to reset it from the outside; (look at the help file of [bezier])
• this can be avoided, if you put some limits to the action field of the scalars. Is this patch just to serve as an illustration of incoming data, or can people use it as control gui for another patch?
  If the 1st, then it's not a problem. If the 2nd, then you can program some limits to the template, so that the grab points don't go outside the gop box. I don't have the reference right now, but look for the "float aaa:aaa)" kind of command in my workshop files.

Best,

João

Very interesting, thank you.
The circle was my very first data structure patch, for tuto purpose, and I got so much back...

I wish I'll find [bezier] and [swatch], I didn'tn search them seriously yet, and I am on pd-extended 42 something.

Nau

João, there is a crash when closing [swatch-help] or [swatch-gui-help] on OSX. Sure I get a crash report for it, but I can not find the exact cause. It must be something with deallocation of GOP data, as it does not happen with [swatch] itself. In the meantime I also found that my [complexplane] makes vanilla Pd 0.43.1test5 crash (but not Pd-extended). Need some time to isolate the offending code, before filing a bug report.

About the circle again: which object in [swatch] makes the circle? There are two polygons and one curve. (João, [swatch] is not included in the set you posted in this thread, I have it from your Pdcon11 workshop).

Indeed it seems that objects using the scaling format 'float aaa:aaa)(bbb:bbb)' do not get out of range, and other objects using variable coordinates can be made insensitive for mouseclicking in runtime with flag -x. So I modified [complexplane] in this sense, here it is again attached.

Katja

http://www.pdpatchrepo.info/hurleur/complexplane.zip

Hi,

the circle in swatch is made by [drawcurve 0 2 p1x p2y p2x p3y p3x p2y p2x p1y p1x p2y]; all points are declared in the [struct] in the same patcher. I don't know if this is the most efficient way (better ask F Barknecht, J Wilkes or someone else), but the technique is the following, as I recall:

• define 5 points in a drawcurve (last point is the same as the first, to make a closed circle)
• since the whole circle (that is, 5 pairs of coordinates) has to be dynamic to move around, I use the change flag to always reset the display of the circle everytime the values change.
  (something similar happens in bezier; again, that might not be the most efficient way of doing this)
• I think that [drawpolygon 0 2 p0x(0:258) p0y(0:65) p0x. p0y.] acts as boundary controller for the circle, by limiting the values sent to [set $0-rgb-sel p1x p2x p3x] and [set $0-rgb-sel p1y p2y p3y].

In general with structs, a good way to know which one is what is to connect a toggle to the drawing object (0 hides structs, =!0 shows them).

For the limiting of boundaries you only need the aaa:aaa) part. The bbb is for rescaling. If you leave it out, it uses the same values as the input.

Aha, my examples weren't included in the zip file. Most of them should be in Pd-Ext anyway. I'll put in a new file later.