@porres I am more than happy to answer any questions you have about using it or how it works. As I said in the other thread, I hate doing documentation, it tends to turn into a rabbit hole for me and eat up a good deal of time; once I start thinking about usage I start seeing how some things are not right in the abstraction and I start fiddling which means a change to the docs and it quickly turns into a feedback loop. That small change to the manual I suggested in the other thread caused/required me to reread much of section 2 and reread a few sections many times even though I had it all typed out after reading two short section once.

I don't think extending pd would solve this issue, the question was not really how to do this in an isolated situation but about the methodology. How do you do something like functionA (functionB (1 2) 849 functionC ($1 $2) asdf) in pd? In a one off situation it is generally a fairly simple problem to solve but to make it so at anytime you can do the equivalent of [abstractionA [abstractionB 1 2] 849 [abstractionC $1 $2] asdf] is a much more complicated issue and one which externals don't solve and often make it seem more complicated than it actually is because they have a tendency to make pd feel like a higher level language than it is, keep the user from ever really learning the nuances and techniques of the low level dataflow paradigm that is pd. The one off solutions which often come up in these threads result in what I think @ddw_music refers to as "spray everywhere," to try and build that solution into every abstraction so you can use it everywhere but to get the equivalent of abstractions calling abstractions would require you to fill your abstractions with [v ]s and [send]s and [receives]s in the vain attempt to cover every possible situation and very complicated/convoluted bit of inter abstraction communication. All my posts since the first asStream screenshot post have been about methodology, how to build higher level paradigms in a low level language like pd and avoid spraying everywhere, this is what asStream is meant to show. Like any well founded low level language pd is quite good about giving you the tools to build out higher level functionality and adapt it into your own language. At least that is what I got from this thread.

asStream has the potential to be a really useful abstraction but still has some issues, I can help you understand it and flesh it out, just ask your questions, they will probably make it a better abstraction than either of us could make on our own. As an exercise I will attempt to repatch it in your heavily subdivided patching style on my next break from my projects, I will probably fail but it will be interesting, I get the logic of the style but it obfuscates the dataflow for me and anytime I want to understand the work of a subdivider I have to remove all that subdivision, I can't even see how it could be subdivided into discrete, well named subpatches of highly singular/specialized function, for me it already is that.

FWIW, I interpreted "spray everywhere" to mean that Pd "functions" spray their return values to every place that could potentially need it, depending on the control flow. That's a common way of dealing with Pd's lack of intrisic "return to caller" when there are just a few callers.

@porres

for now it's just quite abstract and over my head

I made a concrete example -- with scores! If the programming-ese terminology is too difficult, maybe focus on that example.

I hear that it's impossible to do something "x", or too hard, and I hear about complex programming paradigms, and then it comes down to a simple solution like using [value].

That's a good point. Sometimes it takes effort to find the simplest way.

@oid

I don't think extending pd would solve this issue

Yeah, the core of the issue is that Pd's stack works differently from most text languages since Fortran.

Tbh I think variable-target [send] is fine, once it's been established that [send] is going to be involved in pretty much any Pd solution. "Return to caller" as in any structured or object-oriented language is nicer, but [send] is not bad.

what I think @ddw_music refers to as "spray everywhere,"

Here, I'm referring specifically to https://forum.pdpatchrepo.info/topic/15085/return-to-caller-followup-from-symbols-explicit/8 , where the [send] pushes the value out to all matching receivers -- to every potential caller -- where it's swallowed by an [f]'s cold inlet, and only the specific caller pushes the value forward. If you have 50 potential callers, then the data have to be replicated 50 times.

If the data are sub-lists instead of single floats, that would be a lot of multiple allocation. My/jameslo's solution doesn't do that; nor does yours.

to get the equivalent of abstractions calling abstractions would require you to fill your abstractions with [v ]s and [send]s and [receives]s in the vain attempt to cover every possible situation

[list-map] in list-abs shows a nice approach, though, decoupling the iteration logic from the user-operation: the abstraction outputs a trigger to perform the operation, and then the result of that operation is put back into the abstraction to finish the job. That could inspire a super-minimalist approach (which has the benefit of assuming nothing about the returned value's type, whereas most other solutions posted here seem to assume it will always be a single float):

Now, I didn't think of this until the discussion went through a few iterations... see porres's comment about the eventual solution often being much simpler than the problem seems at first.

hjh

@ddw_music said:

"Return to caller" as in any structured or object-oriented language is nicer

I think this thread may have convinced me that the idea of return to caller is antithetical to dataflow, it goes against the flow of data and my solution which you said side stepped it is just restating the problem into dataflow. Not that this was my intent, didn't even realize I sidestepped it until you mentioned it but it got me thinking. Not sure if it is even possible or practical to avoid but I think I am going to try and be more rigid with the flow of data for awhile and see what happens.

Here, I'm referring specifically to

That is what I originally thought but the previous post confused me, I now realize that you weren't addressing my patch directly but offering a summation of what you got from the thread in your reply to me.

whereas most other solutions posted here seem to assume it will always be a single float):

Just replace [array] with [text] and make a tiny change in stream.pd/add other abstractions asStream can load/change the arguments it is loaded with. I switched from [text] to [array] because you used an array in your SC example and I switched tactics to confront your example head on. Originally I thought your ignoring the suggestion of [value] was that you wanted to use things other than floats and then thought that what you were having issues with is the methodology. This is probably a part of the problem porres mentioned about people having issues transitioning to pd from another language, you have a good grasp on programming and when you ignore a solution which seems the easy/obvious solution to us it is easy to assume you have a good reason for ignoring it. These sorts of assumptions are probably reasonably common in such situations, we all gave up on the suggestion of [value] early on and it came back into the discussion purely on accident.

It has been an interesting thread, the failures in communication ended up surprisingly insightful on many fronts.

I'm sorry I don't read the full thread, English is not my natural language and it's a bit technical…
Nevertheless I'd like to note that even in written languages the return to caller process at low level is coded by giving the called method an address to store the result in and then the caller can read it.

Pseudo assembly code

MAIN:
give me address to store N bytes
push this address to the return register
call the method
METHOD:
do stuff
get the address from the return register
write N bytes at this address
:END 
read N bytes at the address
do stuff with this result

The full code is unfoldable as a control flow.
The N bytes can be a pointer to an address the method allocates so the return data size can be set by the method.

The easiest way to simulate this in puredata is to create a memory address as [value] do for float. The caller can then give the address to the method with an argument or a message.
As data structures are pointer-accessible they can certainly help to create such a feature.

@FFW said:

Nevertheless I'd like to note that even in written languages the return to caller process at low level is coded by giving the called method an address to store the result in and then the caller can read it.

I think (not a major expert here) that most text languages do this on the stack: an operator or function leaves its result on the stack, and returns to the point in the code that called it. Nested calls like myFunction(random(20) * 2, random(20) * 3) are easy for an interpreter to handle if they're:

1. Push 20
2. Do "random" on the top value.
3. Push 2.
4. Do * on the top 2 values.
5. 6, 7, 8. Similar 4 steps.
6. Do "myFunction" on the 2 values left on the stack (and leave its result there).

In that sense, high level languages are just a less mind-bending way to write FORTH.

(oid is right that Pd uses a stack for something quite different, such that returning data back upstream really can't be a thing. But workarounds aren't all that painful in the end.)

hjh

@ddw_music said:

I made a concrete example -- with scores!

which one exactly do you mean?

Probably best to skip this post and just read the next, will leave it here since it could result in useful replies from those who understand the internals of pd and perhaps some people here are as interested in these sorts of details as I am,

@ddw_music I think pd uses the stack in exactly the same way but the difference here is that pd lacks a return stack (and stack/instruction pointer manipulation); abstractions are actually macros which are expanded, they are not functions which are pointed to and returned from. The unique ID given to an abstraction when it is expanded allows us to use them sort of like functions and mimic scope but pd is a very simple globally scoped imperative language.

Edit: I guess pd does have very basic IP manipulation, [send] is essentially a goto. The big difference with pd and dataflow is that the IP carries a bit of our data with it as it moves through the patch. The result of an operation does not go onto the stack but is carried with the IP. Or maybe data is only carried with the IP in a conceptual sense and result actually is TOS? Think I will have to investigate pd's stack when I have the time.

Edit2: I guess wires are also gotos then. I suppose what this really is is that pd is a 2d language, [send] and wires only seem like gotos if we try and think of pd as a 1d text language.

Edit3: (last one) Pd would have a return stack, but the returns always point to branches. A [t b b b b] is essentially push and drop commands for the return stack, rightmost bang pushes, leftmost drops. End of the branch having an explicit pop so it can return to the [trigger] or where ever the code branched. More like other languages than I thought in stacks, assuming my reasoning is sound, the big difference being that pd has no concept of anything remotely function like. IP probably carries the pointer as well as the data, would save popping everytime it return which means no return stack and back to my original point of pd lacking a return stack (subrosa edit4.) But that would mean the IP caries a return stack in a rather funge like way, which it might, the return stack is part of the IP object and not the interpreter proper (subrosa edit5.)

I guess I went off on a tangent there, went into implementation and lost focus. The important thing there is that abstractions are not functions which you return from or objects you point to, they are macros which expand in place. We can sort of use them like objects or functions but they are macros from the users perspective and that is what we need to keep in mind.

@oid said:

Edit2: I guess wires are also gotos then.

I realized a bit later what the difference is -- In Pd, wires are procedure calls, not gotos.

Continuing the tangent, feel free to disregard

Pascal distinguishes between procedures and functions, where a function has a return value and a procedure does not. (C gets rid of the term "procedure," but you can still have void functions.) In SmallTalk (and SC inherits this), there is always a return value, which may be ignored -- hence no such thing as a procedure. The last thing a function or method does is to return a result and drop its own execution frame off the stack. In rand(20) * 2, after rand returns the random number, there is only the result -- the fact that rand was called at all is forgotten. (Returning from a procedure just drops the execution frame, with no return value.)

In dataflow, there are only procedures and the last thing an object does is to procedure-call the next object -- the entire history stays on the stack. In [random 20] --> [* 2] --> [print], "random" is more like: 1. Get from the RNG; 2. Call descendants with this result; 3. Then backtrack up the stack.

It's kind of like, Random(20, descendants: [Multiply(_, 2, descendants: [Print(_)])]). I could hypothetically create these objects in SC and it would work (but, not much point lol).

The two approaches are mirror images: lisp would write (print (* (random 20) 2)) where the last operation to be performed ("print") is at the head of the tree. Structurally SC is the same. Syntactically it may be different, since a method selector may be either pre- or post-fixed, but (rand(20) * 2).postln still parses as follows. Then it renders bytecodes depth first! So internally it ends up like FORTH ("push 20, random, push 2, *, postln").

   postln
     |
     *
    / \
rand   2
 |
 20

But in Pd, the last operation is structurally the innermost leaf, not the head.

I think this explains a bit why it took me a good couple of years to wrap my head around dataflow, even though (I thought) I knew a thing or two. It's literally upside-down world for me. Still struggle with that sometimes.

hjh

@porres said:

which one exactly do you mean?

https://forum.pdpatchrepo.info/topic/15085/return-to-caller-followup-from-symbols-explicit/9

With a bit of clarification at https://forum.pdpatchrepo.info/topic/15085/return-to-caller-followup-from-symbols-explicit/13

hjh

@ddw_music said:

In Pd, wires are procedure calls, not gotos.

Are they actually procedures or just conceptually? I would assume the latter but if they actually are procedures can you point me where to look in the source? I get the depth first stuff, I just turned myself around in those edits, hence the follow up post and the disclaimer added to the original post. I did a bit of digging in the source and I think my original point was correct, pd does not do anything weird with how it uses the stack, it is just that without a return stack we can't actually return which means we have to go about/think about things in a different way.

lisp would write (print (* (random 20) 2)) where the last operation to be performed ("print") is at the head of the tree

But there is no tree in lisp, right? it is just a list data structure for the REPL to evaluate. I can't quite grasp how lisp works under the hood, I poisoned my brain with one of those build your own lisp tutorials which I later learned did not actually function in anyway like lisp proper, it was just an easy way to do a lisp like language and I have been lost ever since. Lisp is one of the languages I really want to learn and understand because it is supposedly great for building languages but there is something about it which I fail to grasp.

Regarding Forth since you keep bringing it up, it is no where near as mind bending as people think, in practice you don't spend much time doing that low level RPN stuff; you define some words and use those words to define other words and before you know it you have a purpose built high level DSL for your task. But you have to learn all that low level RPN pushing and popping stuff first. I am a big fan of Forth and what got me interested in language design/implementation was the whim of implementing a language in pd, so I found a language which would seem suitable for that whim and it was Forth, I learned Forth by implementing it in pd.

@ddw_music said:

@oid said:

Edit2: I guess wires are also gotos then.

I realized a bit later what the difference is -- In Pd, wires are procedure calls, not gotos.

In my mind wires (and hidden wires [s] [r] [v]) are variables and boxes are procedures…
You share data throw wires as you transmit them with variables or stacks or registers. The order you link the boxes give you the order the data stack is filled.
By the way, it's also the order the process stack is filled because the full tree is flatten before it computes.

@oid said:

@ddw_music said:

In Pd, wires are procedure calls, not gotos.

Are they actually procedures or just conceptually?

The distinction I'm thinking of is that gotos have no history, while procedure calls do. When a chain of objects runs out of connections, pd and max do backtrack, so there must be a call stack for the objects that have been traversed so far (this, I'm certain, is not conceptual in pd -- there are stack overflow errors, for one thing). A goto is just a jump, with no memory of how it got there.

hjh

@ddw_music All that stuff about instruction pointers and the like was me getting turned around and falling back onto Fourth because that is where my mind goes when I think about stacks. In pd there is no instruction pointer so never any need to return or backtrack or goto. In pd's core, every internal and external object is a literal OOP object and the inlets and outlets are methods of that object. When we create a new object pd creates a new C object of that type and makes an entry for it in glist with its pointer and UI details; when we connect a wire to its outlet pd retrieves the pointers to those objects from the glist and then goes through the outlet method. I am not quite sure how it deals with reaching the end of the branch, I suspect that the last object always has its outlet method run to point it at the branch point and when we append a new object pd just passes that to the new object when it updates the previous end of chain object's outlet method to be the new object.

Some details there might be a tad off, I am not quite good enough with C pointers to really follow pd's source yet, but I am fairly sure that is the mechanics of it. I have no idea about how Max handles this stuff but I would assume it is quite similar.

@FFW said:

By the way, it's also the order the process stack is filled because the full tree is flatten before it computes

Is there a process stack? Can you tell me where to look in the source for it? What is actually on this stack? Is my above reasoning regarding the patch just being a chain of objects incorrect? The patch being a chain of C objects explains why we can't turn off message rate and what happens when pd is running out of cpu but I have not quite grasped how pd's scheduler actually works. I'm guessing my previous post is difficult for a non-native English speaker since my grasp is not great and my explanation is probably not the clearest and I am not sure how well you know pd's internals either. I am almost at the point where I will post on the mailing list to get a response from the devs but I want to get enough of it that my post is not just a very verbose way of saying "I don't get this, explain the entire inner workings of pd for me!"

@oid "Process stack" is my own term to describe how orders arrive at the processor. I'm thinking at low level, below the C one.
Let's consider this tree and translate it in pseudo-assembly or pseudo-forth:

: BANG
push bang to stack

: MESSAGE_One
pop stack if bang push "One"

: MESSAGE_Two
pop stack if bang push "Two"

: PRINT
pop stack, send to console


: RUN
BANG MESSAGE_One PRINT BANG MESSAGE_Two PRINT

RUN

Trigger is unused at run time, it's only used at "flatten" time to put the methods in the right order. the :RUN content is the process stack I talk about.

However audio tree is flatten to create a DSP chain but maybe the control tree is not…

@FFW I sort of thought you meant something like that but wasn't sure. I don't think message rate stuff is flattened and can't find anything in the source but I don't really know, I just don't see much advantage in doing it with how slow and simple message rate is. Thanks for the clarification.

@oid said:

@ddw_music In pd there is no instruction pointer so never any need to return or backtrack or goto.

You absolutely must backtrack in the case of any object that has multiple outputs. Backtracking is most straightforwardly done with a stack.

A stack is only a data structure that defines two operations: add something new at the end (push) and take the last item off of the end (pop) -- last in, first out. "Stack" doesn't define what it should be used for. It may be data, as in FORTH; it may be execution frames; can be anything.

If necessary, I could find it in the source code later, but let's assume that Miller Puckette is no fool, and he knows that tree traversal is best done with a stack, and that he isn't going to waste time trying to implement some alternative to a stack because of a preconception that stacks are for x or y but never z (and further, that he wouldn't post "stack overflow" errors to the console if that was misleading).

I am not quite sure how it deals with reaching the end of the branch...

I'm also not sure of the implementation details in C (don't really need to be), but if each object gets pushed onto a message-passing stack, then it's easy: when this object is done, pop an object off the stack and see if it has outputs that haven't been issued yet; then issue them, or if there are none, pop again and repeat until the stack is empty. Because a stack is LIFO, this is always (literally!) backtracking.

If you've ever used Max's graphical debug mode: it literally shows you the message-passing stack in a window. I'll admit that I don't have the line-number evidence to show exactly how pd does it in C, but I'm not making this up...

Trigger is unused at run time, it's only used at "flatten" time... However audio tree is flatten to create a DSP chain but maybe the control tree is not…

I'm quite sure there's no flattening for control objects. (See Max's debugger, or I bet pd's trace mode will show the same.)

Edit:

trace: bang 
trigger: bang      -- not flattened
message: One 
print: One
trigger: bang      -- and, after doing "One," it backtracked to here
message: Two 
print: Two
backtrace:  -- what happened before the [trace] object?
bng: bang 

Note that you don't know "what happened before the [trace] object" without a stack

Again, if I'm really saying anything controversial, when time permits I can go diving into pd's sources and find the evidence. But IMO the [trace] output is enough.

hjh

@ddw_music I never said pd did not have a stack, I said that I did not think it had a RETURN stack. I don't think I can reasonably respond to your post without risking more confusion.