Hi guys,
I have just a quick technical question: how does PD work?
I mean, each object has its own *.c file, but then how does PD "put together" these objects?

I'm trying to read the source code, but I'm not really into C language. Could you explain me roughly the working principle behind Pure Data?

Thank you in advance for your time and help!

Anna

I'm not a programmer. However, you can find some info about the architecture behind Pd at http://aspress.co.uk/ds/sample_chapters.php

So once you are in that site, clic on the button Abridged PDF (Pd intro) . That should allow you to download a .pdf file that consists of a 100-pages sample of Andy Farnell's book "Designing Sound".

In fig 2.5: Pure Data software architecture, you are going to find a nice scheme.

(If you cannot download it from that site, there should be plenty of other sites that allow you to download that 100-pages sample).

external objects each have their own ".c" file often, but that is not how pure data deals with it's "vanilla" objects generally.
Miller wrote an class system from scratch in c. Pd uses these classes for actual objects used in the graphical environment (what the user sees) and a few others that work behind the scenes. An example of the latter is the clock object, which can store functions in it (called callback functions) that are called at a certain time.
When the program runs (called "run time") pd vanilla uses a loader to load binary files with such objects into pd, and creates the graphical representation and initializes any other information for the class. it loads each compilation unit (usually a .c file) separately, each of which may have many classes in them. pd calls an initialization startup function for the file that does all of this.
there are also external loaders that do the same things but in different ways, for instance pdlua which uses a lua loader to load classes written in lua

Thank you both for answering promptly!
I'm sorry for the delay, I've been reading some documentation, including the one Landon proposed but with no success

Let's suppose I would like to create a very simple version of PD (without audio, midi, GUI, watchdog, etc...). I'd like to define just one object (a "+" operator, for example) and then be able to perform a simple operation like 2+2

I know, this might sound stupid, but my aim is to compute the result starting from a *.pd file independently from how many blocks have been implemented in that file.

Thank you!

could you please elaborate?
Specifically, what is it you want to do? You want to put a [+ ] object into a .pd file, compute 2+2 and put the result where?
Why do the number of blocks matter? (by blocks you mean objects?) Why shouldn't this version of pd have other objects in it?
Also what is meant by "starting from a *.pd file independently"?

@wateredge You're welcome. That's all I can offer, but @seb-harmonik.ar and other friendly forumites here will surely give you great tips.

yeah, I know, probably it's not clear. In few words I would like to recode PureData from scratch for an academic project. That's why I would like to start with simple arithmetic blocks. Of course there will be no GUI, so it's going to be a very minimal version of puredata

that's quite an undertaking,
so from my basic understanding of pd internals this is the basic outline of what needs to be done:

• a class system
• a message passing system (with a symbol table, etc.)
• a graph system so that outlets of objects send their messages from outlets into other objects.
• implement some basic classes (such as +) in this system
• make a scheduler in order to schedule clock callbacks (pd uses a list of t_clock objects) if you want to support timing

I would also suggest looking at the code of libpd because it is basically what you're describing (pd without gui, able to run patches directly from c) but I guess that won't help if you don't know c well