I've been trying to finally learn how to use data structures, with some success this time. But I'm stumbling with data retrieval, and can't seem to get it to work as expected.

My objective is to have an array with a variable number of x-y data points (append) which can be defined textually (set) and then have its coordinates queried (get). But the query always comes out 0 0 until the data point is manually moved with the mouse.

This attachment should illustrate: ds-help.pd

What am I getting wrong?

Saved by the @Balwyn! I was vaguely aware that there were too many 'layers' in the patch that I'd posted. [append] adds a new plot which needs a new pointer, while [setsize] and [set] add to the existing plot---is that correct?

Something else I'm noticing is that the (0, 0) point of a data structure seems to be (14, 14) of the Pd canvas. Is there any explanation for this? It's easy enough to correct of course, but I wonder if I'm doing something else wrong.

@LiamG , As I see it, If you have a struct that contains an array, the pointer from the append object is the last pointer required to address the array, so store that pointer to address the array. Using that pointer you can set the size of the array with [setsize templatename arrayname]
Then set the values of the array with the same pointer by sending the pointer to the right inlet and an index to the left inlet of [element templatename arrayname] then send the element output to the right inlet of [set arrayname] <field> <field> etc) where the fields maybe x, y ,c etc. the x,y values to the relative inlets (or list to the left inlet).
My head just exploded!
Now for something completely different
the circle needs a centre of 0,0 for its position to relate to the location point, this creates an offset that needs to be accounted for in the gop.
This is rather awkward.
Basically in the gop dialog, add the size of element object to the coordinates range but leave the size the same. but there are often anomalies and in this case an extra 5 was required for the y (why?)
datastruct-help-2.pd

@Balwyn it was because of the alternative x and y values in the [plot]. Should be all clean and tidy now.
datastruct-help-3.pd

Usually, the gop x/y range should meet the size. If not, the visual elements will be squeezed or stretched, which of course can be a desired effect at times.

@ingox , your right, and I new that and I forgot that. thanks for pointing it out
@LiamG sorry, disregard the last part about the offset, I failed to notice that the [plot] in [pd $0-big] template was [plot $0-array 500 0 10 15 20] and should of been [plot $0-array 500 0 ]. 10 and 15 were shifting the circle and the 20 was ignored because you have a float x in the $0-small struct. and is only used for auto spacing when there is no float x in the $0-small struct

Brilliant! This will make thing much simpler. Can't wait to share the finished patch with you guys.

@Balwyn @ingox

I'm almost finished with this project (still a bit of a secret) but I've run into another minor problem. I've been using [setsize] to determine the size of a data structure array, as you guys taught me. The question is, how do I set an array of size 0? Ideally I'd like to be able to make all the points disappear, but [setsize] seems to have a lower limit of 1. For an example, just take a look at the [setsize] helpfile.

Is there any other fancy way of removing all elements from a data structure array? Bare in mind that I want to be able to set it back to a positive number easily, so deleting the array might not be the best idea.

A one or zero to the inlet of [plot] will show or hide the elements although they still retain their value, is that any use?

Thanks @Balwyn, this will probably do it. I'll have to clear the data as well, but this can probably be done further down stream.

@LiamG Yes, hiding the plot is one option, the other one is to clear the subpatch. This seems rude, but clear and redraw is usually the only way to actually remove elements (resizing an array is the one other way but not everything is an array). So it is really totally ok to clear and redraw everything whenever needed.

Here's my latest challenge with data structures. I have a 2-dimensional graph holding a series of data-structure points, which I can drag around the canvas. Now I'd like to be able to hold down the shift key (or something) and drag the points only in the x- or y-dimension. Is this possible? The best I can think of is to save the coordinates and then use them to reset the x- or y- data when the shift key is set, but this would be a bit of a hassle. Is there some provision within data structures to be able to do this?

Counting on you @Balwyn and @ingox!

@LiamG Yes it is possible: Get the coordinate on click, set it back on change, so it effectively stays the same.

@LiamG Here is a quick demo: struct-demo.pd
Could be more sophisticated to get the correct value even if shift is pressed later, but it shows the basic idea.

Ah, I hadn't thought of unpacking the [struct] message like that! Thanks @ingox, the next version of my [sketch] project is going to be much better because of this!

@ingox This is strange--in my patch, I'm not getting any of those useful "change" messages coming out of the [struct] objects, and I can't work out why. Is it something to do with the "float c" argument? I tried adding this and it still didn't work, though I might be getting the syntax wrong.

Here is my container for the element:

and for the array:

Out of either of the [struct] objects, the only message I get is "click (pointer) 0". Any ideas?

@LiamG Yeah, it has nothing to do with float c, that is just the color. It is because in my example i add the scalar directly (can be done with scalar message or append to the subpatch). Then you get the click and change from every single object (scalar).

But when you add scalars as elements of an array, they are kind of locked into the array and don't behave as independent scalars, meaning they don't send click or change messages. When you receive a click message it will be from the main struct and the pointer will be to this main struct, but not the individual element. Sometimes the array struct will send out some crazy event, but this is more like a bug.

So this is a big mess and in my opinion incomplete implementation or even buggy behavior and has caused some headaches to some here in the forum, but this is the reality we are dealing with

Well that's a pain. I'm guessing if I place them directly, I won't be able to read the elements' positions using [get], right?

@LiamG Yes, that is correct. But you can build your own system to store pointers in a list so you can access them by number and thereby access the scalars. In the example below i did it with [list store], but it can also be done with [list] + [list-idx].

struct-demo2.pd

Here the scalars are added to the subpatch with the append method. The [append] object outputs the pointer to the scalar whenever one is added. The pointer is then stored in a list which is send to another [list store] on the right. Here you can give a number, get the pointer from the list and get the values from the scalar.

The method above to lock y0 when moving the objects with left shift key pressed still works like before.

Also there is another method that you can find in Chris McCormick's s-abstractions s-controlsurface: https://mccormick.cx/projects/s-abstractions/. The method is to traverse the pointer to the subpatch and then use [next( so many times until you reach the desired scalar. Which you then can access.

Edit: i should add that the way Chris McCormick does it, the scalars are appended to the head of the subpatch (same as i did above). Therefore the last added scalar becomes the first in the subpatch and [next( is therefore not in the expected order. Anyhow it should be possible to implement it correctly (only traverse the pointer of the append on load and on clear or instead count elements and use inverse number), but i couldn't quite figure it out on first try, so i just recommend the above method with lists which also gives the most control in my opinion

So we still use [get], we just don't use [element] anymore to select the scalar. There i was not clear in one previous post.

As far as constrained dragging, you can do it in Purr Data fairly easily:

purr-dragger.pd

As far as constrained dragging on array elements, you can do that in Purr Data as well:

purr-dragger-array.pd

A few notes:

• the "drag" event is set on the relevant [draw] command and sends notifications to the outlet of the relevant [draw] command. This allows "per-shape" events so you can tell which shape triggered the event.
• since these draw events aren't turned on by default, you can decide which events you want to listen for-- mousedown, mouseup, mouseover, mouseout, mouseenter, mouseleave, and drag. (Scroll and touch events are possible but I haven't implemented them.)
• for the sake of sanity "x" and "y" fields don't have any automatic behavior when scalar elements are drawn using [draw array]. If you want to make associations you can declare them in a message as I do in the demo.
• [draw array] has no bounding box to select it in Edit mode. Otherwise it would try to get the bounding box for all its elements and cause dropouts for arrays of non-trivial size every time you move the mouse. To get a bbox you can nest it inside a [draw svg] which creates a viewport clipped to a rectangle you can specify.
• receiving events for ds array elements is fairly new behavior. I know there's at least one crasher lurking if you try to add fields to the [struct] for an array element with extant scalars on a canvas.

Edit: clarification