Eric Lyon

Is there a tight JavaScript methodology  for sending messages into specific inlets of an object that was created with JavaScript code? The object might be a native Max object that has more than one inlet, or it might be an abstraction that needs to have messages sent not to the containing patcher, but into one of its inlets. 

One could of course create a temporary message object, connect it to the inlet in question, set the message into the message object, bang the message into the desired inlet, disconnect the message object, and then remove it from the patcher. But that feels overly dramatic. It would be very enjoyable if a more direct approach were possible.

Greetings,
Is there a tight JavaScript methodology  for sending messages into specific inlets of an object that was created with JavaScript 

programmatically-messaging-a-specific-inlet-on-a-javascript-created-object

Did you ever figure out an elegant way to send messages to inlets via JavaScript?

I just bumped into this issue. I am trying to build a Max patch entirely (or as much as possible) in JavaScript, and this issue has become a major impediment.

Any code examples you have would be much appreciated.

I never saw an update for JavaScript in Max to provide this functionality. It would still be nice to have it.

Programmatically messaging a specific inlet on a JavaScript-created object

Tom Swirly

I have no trouble finding boxes from inside my JS patch, or sending messages to those boxes - unless the box is a bpatcher.  

If it's a bpatcher, if I try to e.g. send the message goodbye, I always get an error "patcher: doesn't understand the message goodbye".

I attach sample code (unfortunately, I don't see how to save the JS with it, but I have the JS in a comment in the code).  You can see that the JS code does locate both the message box and the bpatcher, and successfully sends hello to the message box, but gets an error trying to send goodbye to the bpatcher.

I feel I have solved this before in the past but I cannot remember how.  Suggestions most welcome!


I have no trouble finding boxes from inside my JS patch, or sending messages to those boxes - unless the box is a bpatcher.  
If it's a bpat

sending-a-message-to-a-bpatcher-from-within-js··

I'm going to work around it for the moment, but while I am here, let me ask the other question that's been on my mind - from Javascript, how do I send a message to an inlet of a box that is not the leftmost one?

To summarize - in Javascript, right now I know how to send messages directly (without a patch cord) to the leftmost inlet only of any box  which isn't a bpatcher or patcher, by using the message() method.  I cannot send to any inlet other than the leftmost one, and I cannot send to batchers or patchers at all.

For the second issue - i'm pretty sure there is no way to send to any inlet other the first.  In this case i put a receive object before the object in question and use route or unpack to route to the correct inlet.

Yeah, that's what I'm doing, but the point of the exercise was to avoid having all those patch cords everywhere.  It's a shame that I have the very message I want to send and the very object I want to send it to, but there's no way to actually send it without creating a bunch of wires for this single purpose...

I'm still hopeful I can directly send to bpatchers from my .js script.  I can't be the only person who wanted to do that...

For the first - not sure how to send to the inlet itself, but you could send to an object inside?

Only if that object has a uniquely named receive somewhere in it.  I could do it with the usual receive #1-foo trick but such solutions are fragile.  More, it means I have to rewrite these boxes that I'm already using elsewhere.

This issue comes up for me all the time - I have a large number of otherwise identical boxes that I am controlling from a Javascript script that has all my actual logic in it.  

All I want to do is to send it a message in its inlet - just like I could to any other Max object.  Why is bpatcher different?

So I'm believing that I'm not going to get any more answers here, and that my results are unfortunately correct.

To summarize - it seems as if the JS message method only allows you to send to the left-most inlet, and only to boxes that aren't bpatchers or patchers.  It's unfortunate - without these critical limitations, particularly the second one!, message would be an amazing general-purpose tool for building Max patches almost entirely in JS, instead of just a curiosity.

same problem found with bpatchers.
I'm using a named 'trigger l" connected just after the inlet inside the bpatcher.
my outer JS can send to this bpatcher/trigger.
With a named trigger object, no needs to acces the global send/receive in JS, and to give a unique name to the receive. the uniqueness is given by the bpatcher script name.


The code to get the named trig inside the bpatcher:

 i'm using an array of bpatcher, with a trig inside each one named "trig"

objs_array[i].replace(obj_maxpat); // load the bpatcher maxpat

trigs_array[i] = objs_array[i].subpatcher().getnamed("trig"); // get the trig inside it that is just after the inlet

trigs_array[i].message("blablabla"); // send a message outputed by the trigger

This is a neat idea, but it doesn't really hit the spot.  The trouble with it is that you have to rewrite each box you use this way to have the named trigger in it.  If I were going to do that, I could just add a named receive in each box, and use #1 to make sure it's a unique name.

What I want to do is simply talk to any box in my patch without having to rewrite that box specifically for that purpose.

yes, i know. but thats the only workaround i found.
I just put a "trigger l" after my inlet in all my bpatcher so i can access them through javascript.
I don't know if someone is 'still' in charge of JS for maxpat in cycling cos all those questions about criptic messages to Bpatcher has been asked many times and no official answers.

https://cycling74.com/forums/there-is-no-way-to-make-a-bpatcher-and-include-in-presentation-mode
https://cycling74.com/forums/bpatcher-and-rect
https://cycling74.com/forums/bpatcher-args-multi-segment-name

Oh, not at all carping about your solution, which is I think as good as it's going to get.

I still have resentments from spending a year on a JS project that I could never get to stop crashing in Max 6, and then later discovering that JS in Max 6 was not thread-safe so you couldn't reliably use it for more than one input, so it makes me grumpy when I have to spend a few hours more of my life working around long-term issues that are well-known, and I'm therefore sorry if I come off as a grouch!

Considering that Cycling 74 has been taken over by Ableton, a company with an extremely poor history in fixing bugs (there are numerous serious bugs in Live reported by me and others that are over five years old and have never been addressed), my theory is that this will never get fixed.

By the way, did you get any answer from cycling, respect to the relative path for require being not compliant with commonJS?
As I stated in others posts, i think cycling should set up a wiki dedicated to all those issue and cryptic messages in JS with max.
Happily, rob is answering regularly  for the jitter/GL part of JS...


Ah, interesting question!  I'm flattered to be recognized.

But no.  To be honest, I'm pretty sure that I convinced them that I was some eccentric, because the response was basically, "Why would you want that?"

The idea that you might want to use some of seemingly-infinitely many node.js packages was perhaps a little science-fictional at the time, but these days it seems much more desirable.

<rant>

To be frank, the level of support for music and audio programs in general in the industry is just terrible.  Pretty well every program I use, there are critical bugs that have first been reported many years and then re-reported by numerous people, without ever being followed up on.  I've whined ;-) on this forum about Ableton quite a bit - how M4L only allows you one MIDI input and output so at a certain point everyone ends up using an unsupported, orphaned freeware box that lets you speak directly to MIDI devices, how it won't let you use System Exclusives or Program Changes.  (And that's only one of dozens of serious issues I have documented with Ableton.)

But Native Instruments, for example, is even worse.  Almost none of their packages have even one level of undo.  Worse, their interfaces are tiny, unresizeable windows, and typically there is a "mutate" button that changes all your parameters squashed right up against a button you use frequently, so it's very hard not to occasionally click it - and unrecoverably destroy all your work in that session.  (I get into the habit of obsessively using Save-As with index numbers to partly mitigate this issue, but that has its own drawbacks.)

I moved recently to Reaper and it seems to be much better that way, but we'll see how it goes in the future.

</rant>

You do know none of those things mentioned in your above post are bugs right? And the seriousness of those issues is totally dependent on what you are doing and your particular workflow?

I don't mean to  invalidate those requests about how Live handles MIDI, because I agree, but to call them critical bugs is a bit extreme. 

A MIDI programming tool that only works on channel 1 for input and output and throws away a majority of the  MIDI message types - including program changes! -  is broken.  The MIDI standard has been around since 1983 - it's tiny, a few pages.  If you can't support that standard, at least as far as not throwing information away, you have no business calling yourself a MIDI programming system.

From my notes, here's a list of the MIDI messages that M4L silently discards:

* program change
* system exclusive
* MIDI time code
* song position pointer
* song select
* tune request
* end of exclusive
* timing clock
* start
* stop
* active sensing
* reset

This is a majority of the MIDI message types.  And I'm not asking for the program to do anything fancy with them, other than not throwing them away.

Some of these are obscure, but some of these are extremely important, like program changes and system exclusives.  Indeed, I have a dozen or so hardware synths in this room with me and all of them use both program changes and system exclusives.

You could argue that this isn't a bug - it's simply surprising and undesirable behavior that isn't documented and doesn't match the MIDI standard.  I'd certainly accept "Shoddy work and no pride in their own craft" as another description of this.

Don't get me wrong - it's not Max's fault.  

Max handles all MIDI messages perfectly well, handles all MIDI channels fine, handles multiple interfaces if you have them.  But Max For Live simply drops a bunch of messages on the floor, throws all data onto channel 1, and then gives Max the pathetic remains.

I paid good money for M4L when it came out, even though I already owned Live and Max, and it was basically useless for my needs because of this.  At the time, we were told that SysEx was coming soon, but it's been nine years and I no longer believe this.

There are numerous outright bugs with Ableton that they are never going to address, but aren't related to Max so I didn't mention them.  For example, if you stretch time on a MIDI track, it stretches only note-on and offs - it leaves all the pitch bend and continuous controller information right where it was.  Ableton's response on this that that was exactly how it was supposed to work  - but this means that if you play a MIDI performance which uses pitch bend, modulation wheel or aftertouch, and wish to stretch or edit it in such a way as your notes and your controllers stay together - well, there's just no way to do this. 

I reported that almost a decade ago and I periodically see other people finding this out online - usually after they trashed some performance they recorded on MIDI.  Ableton will never fix that.

I personally have a commercial digital audio application I wrote entirely myself, and I've been writing computer music programs since the late 70s, so it's not like I'm speaking out of ignorance.

You could argue that this isn't a bug - it's simply surprising and undesirable behavior that isn't documented and doesn't match the MIDI standard.

That's all I'm doing. File all this stuff under feature requests, not critical bugs. 

Saying, "Sorry, we only support one MIDI channel, something we don't mention anywhere in the heavy PR campaign, and we're never going to change this" - that is a critical bug in a MIDI programming system.

I, personally, would be ashamed to do such shoddy work.

I'm curious - can you point to other MIDI programming system that only supports a single MIDI channel?  For that matter, can you point to another DAW where stretching a MIDI track puts the controller information out of sync with the notes?

I use multiple MIDI channels all the time in Live.
I don't think Live has ever branded itself as a MIDI programming system, although it does support (in a somewhat limited fashion apparently) MIDI.
I suppose its good you don't work at Ableton, then. 
I don't use any other DAW for MIDI work, so no I can't. 

My point is - I don't see the point in making such a huge stink about these 'critical bugs'. File them as feature requests, and move on creating things. 


No, in that one feature I'm talking about Max For Live, as I was very clear - I mention M4L four times.

Max For Live does bill itself as a music programming system.  It only works on one channel in, and one channel out.  Go ahead, try it yourself now.

I bought Max For Live for a specific project where I needed to use multiple MIDI channels.  It was useless for that but of course, try to get your money back, and they won't do it.  I was unable to "move on creating things" as there was no way to accomplish what I needed with just one MIDI channel.  I eventually came up with a complex workaround, after wasting a few days of my life.

Ableton has no place to file feature requests.  For a short while, a third party ran a website where you could file bugs against Live, but then Ableton's lawyers got it taken down.   You can contact tech support, but there's no way to track your requests, and so far not one issue I've raised has been dealt with - years later.

I have of course contacted them about these issues.  I am generally very polite in these matters, though you have probably detected my patience is pretty well shot after a decade.  Their response has almost always been to claim that this is how it's supposed to work - yes, even the matter of "any edit to a MIDI track with both notes and controllers breaks the track".   Often they ask, "Why would you want to do that?" - even about things like wanting to record a MIDI track without playing to a metronome (part of my discussion about "editing MIDI tracks")

I should not have to explain to a music company why I don't always want to play to a metronome!

This idea, that I should just accept software that I have paid for that does not actually do what it claims - because I'm sorry, but a MIDI programming tool that doesn't actually support more than one channel and throws away a majority of the MIDI message types is broken - is a poor one.  Other companies do far better than that.  Reaper costs a fraction of what Ableton Live does and they're incredibly responsive and polite.   


Regarding access to box inlets: this is a huge gripe of mine.  My experiences are the same as those already cited, and I really wish there weren't the need for workarounds (I code extensively in JS within Max, and most of my connections are scripted using obj.message() methods).   

As far as sysex goes:  it's part of the upcoming Live 10 release.  Better late than never ;)

Multichannel MIDI:  complete agreement with Tom on this, but I don't see it as a bug,  instead I think it's just a low-level  design flaw.  I'm sure if it was easy to fix, Abes would have done it by now.  I hope to see a future solution that parallels what they have done with multichannel audio for m4l in Live10, but I'm not holding my breath.  In the meantime, there ARE solutions for using multichannel MIDI via various third party externals.  I quietly published one in my own m4m7 repository for OS X, or you could use either of these:
https://www.theimpersonalstereo.com/max-externals/
https://sigabort.co/m4l_tools.html

One thing to note, though, is that Windows's handling of MIDI ports is horrid, and that greatly limits what can be done using these solutions on that platform.  I suspect that may even have some bearing on the reasoning behind why we haven't seen an official solution for this problem yet.

@Tom:  maybe join the Live 10 public beta? I believe they DO want to hear what their users have to say, and that provides  a forum to make suggestions.  Being in the minority when it comes to use-case scenarios isn't pleasant (believe me, I understand), but the more voices heard, the better.  On the other hand, if it hasn't happened because of some technical limitation buried in decades old base code, I guess we're beating a dead horse ;)

Ah, no, I've left Ableton forever, and joined the ranks of the Reaper people.  Reaper is absolutely everything I ever wanted in a sequencer or a DAW!

I had a chance a few years ago to chat to an engineer who had seen the Live codebase, and most of these problems are probably unfixable - for example you'll never be able to open two Live projects at the same time as "the current sequence" is a singleton; you can't stretch both MIDI notes and CCs or pitch bends at the same time because the notes are stored as part of the clip whereas CCs and pitch bends are attached to the track; and you can't really do anything clever with program changes because they're a completely special case, also attached to the track.

In particular, the one thing that I want - to simply be able to play my MIDI instrument into a track without a metronome and then stretch it so it fits into an even number of bars - will always be impossible.  I did this with Performer (before it was Digital Performer), I can do it with Reaper and any other sequencer I used, but the underlying data structures in Live make this essentially impossible.  Indeed, a lot of the source of my frustration is that Live support actually accused me of being "difficult" for not wanting to play to a metronome.  :-(

This question was regarding a little hobby project of mine, to control a bunch of DMX lasers I have lying around, so it's a pure Max problem when it comes down to it.  The rest was wrangling with a random . :-)

Thanks for your civilized and very enlightening comment.

Did you ever figure out an elegant way to send messages to inlets via JavaScript? 

I just bumped into this issue. I am trying to build a Max patch entirely (or as much as possible) in JavaScript, and this issue has become a major impediment. 

Any code examples you have would be much appreciated. 

Sending a message to a bpatcher from within JS..

Trevor being Trevor

I have a max patch that dynamically creates and removes a ton of objects via the this.patcher.newdefault() and this.patcher.remove() methods to essentially build and destroy a dynamic UI via Javascript as needed. The objects are stored in global arrays by type (for example, all the textedit objects) so the functions that create and destroy the UI both have access to them.

It's been relatively smooth. However, as I have continued to build using the same methodology I encounter an odd crash now and then. After troubleshooting, the problem comes down to the use of a this.patcher.remove() in specific instances. Literally, if I comment it out the line(s), the crash does not occur. 

If anybody with more knowledge on Max debug output could give any insight into following info I think I could get to the bottom of it. I am happy to provide more info if needed, just wanted to avoid writing a novel of a post.

These are snippets of the Max debug output when running a bunch of this.patcher.remove() methods. I am wondering if the bolded lines are potential issues that are compounding into a crash later.

Looks Normal:
object_unregister: slating unique sym u857004814 for reuse
unregistering: u562004824 in 00000000000B1890. ref count now 0

Should I be concerned?
unregistering: u597005447 in 00000000000B1890. ref count now 0
object_unregister: slating unique sym u597005447 for reuse
object_detach: object 000000002D0E4170 wasn't attached to object 0000000029D6F890 
unregistering: u915005449 in 000000000014C738. ref count now 0
object_unregister: slating unique sym u915005449 for reuse
unregistering: u587005469 in 00000000000B1890. ref count now 0

Another snippet after adding in some lines that makes it crash in certain instances.
registering: u184003150 in 00000000009267B8
registering: u363003151 in 00000000009267B8
unregistering: u512002235 in 00000000009267B8. ref count now 0
object_unregister: slating unique sym u512002235 for reuse
warning: object 0000000027D28780 was freed but was still attached to 1 registered object(s) 
loadbang_clearqueue
loadbang_clearqueue
registering: u176003152 in 00000000008DC8A0

MaxCrashSnippet.png

I have a max patch that dynamically creates and removes a ton of objects via the this.patcher.newdefault() and this.patcher.remove() ... 

need-insight-on-crash-and-debug-output-for-this-patcher-remove

Need insight on crash and debug output for this.patcher.remove()

tyler mazaika

I'm making some jspainter files for tab/live.tab which cache the "unselected" state for the whole set of buttons as an Image().

It seems like if I don't make my mgraphics context for the cached image 2x resolution (and then downscale during paint() ), it always looks blurry.  

Compare the text in the yellow image and with the blue image where the cached Image() was drawn at 2x.

This seems weird because if I just draw everything on the fly the appropriate scaling happens automatically.  

image.png

I'm making some jspainter files for tab/live.tab which cache the "unselected" state for the whole set of buttons as an Image().
It seems lik

mgraphics-image-differs-from-standard-drawing-on-2x-retina-displays

FWIW the least painful solution seems to be to use scale(2, 2) when when drawing the cached image (on a 2x-sized canvas) and then scale(0.5, 0.5) before calling image_surface_draw() in the target drawing space.

mgraphics Image() differs from standard drawing on 2x / Retina displays ?

Neil Baldwin

As per the title really: when using applydeepif() in a patcher to recursively retrieve nested objects in subpatchers, can you be sure of the retrieval order? For example if several objects of the same name were fetched (multiple instances of the same bpatcher contained in a subpatch), could you read/write to those objects in JS in the numerical order they were fetched and be sure you're reading/writing to the correct object?

If not, how does Max/JS handle (same) named objects inside multiple instances of the same bpatcher?

As per the title really: when using applydeepif() in a patcher to recursively retrieve nested objects in subpatchers, can you be sure of ... 

applydeepif-guaranteed-retrieval-order

That’s a very interesting question. According to the Max 8 Documentation, the applydeepif() method of a Patcher object applies a function to all objects in a patcher that satisfy a test function, and recurses into subpatchers (depth first). However, it does not specify the order of retrieval for objects with the same name in different subpatchers. One possible way to handle this situation is to use the box property of the Maxobj object to get the parent object that contains the subpatcher, and then compare its scripting name or position to determine which one comes first. Alternatively, you could use unique names for each object in different subpatchers to avoid confusion

It seems indeterminate from testing. I think I'll have to bite the bullet and do some refactoring so that all the things I need to target have unique names, as you say.

Applydeepif: guaranteed retrieval order?

Alex Guillard

I  have some kind of problems with the Node API for Max.
Recently, I created a way to communicate in TCP with 2 Max applications installed on 2 different computers. To do that, I just use a Node JS server and client, with the node.script object. Everything works fine actually and I can communicate between the 2 computers, except that sometimes, after a while, the Node Max API gives me this timeout error :
Max API Message Request timed out. id: u52248380302
Error: Max API Message Request timed out. id: u52248380302
    at makeError (C:\Users\XXX\max_project\resources\packages\Node For Max\source\lib\api\index.js:9:14)
    at Timeout._onTimeout (C:\Users\XXX\max_project\resources\packages\Node For Max\source\lib\api\index.js:178:12)
    at listOnTimeout (internal/timers.js:531:17)
    at processTimers (internal/timers.js:475:7)

I tried to increase the time before timeout in the index.js file, but nothing helps.
When I start to use the client/server, I bombard with messages, but it doesn't crash. It's only after a while that it does that... Has anyone ever encountered this problem and know where it may be coming from ?

Hello there,
I  have some kind of problems with the Node API for Max.
Recently, I created a way to communicate in TCP with 2 Max application

node-error-max-api-message-request-timed-out

This usually is due to heavy back pressure on the communication between the two processes. Would you mind getting in touch with support and include the relevant pieces of your patch / JS code if possible?  Feel free to reference me so that they can forward it if needed.

We are considering options for eventually making this easier to handle for the user by slightly adjusting the API behaviour so every data point about you experiences like your is helpful and appreciated. But meanwhile I'm happy to check your project and see if there are maybe some adjustments that could already help.

Hi, I'm having the same issue... Is there any configuration or something that could help max handle heavy communication with Node? Could logging to the console a lot be a reason for this crash?

Node Error : Max API Message Request timed out

Florian Demmer

With the just released 8.0.3 update of Max 8, Node For Max now supports bundling dependencies (aka the node_modules folder) within Collectives, Standalone Applications and Max For Live Devices. In order to give you a jump start into using this feature, this is a short guide on how to do it.

Let’s assume for the following that we are working on a Project called N4M-Bundler.

We start off by opening Max and creating a new Project using the File Menu called N4M-Bundler. In return Max will show us the Project Window, which can be used to organize the project’s files and contents. Since we don’t want Max to try to organize contents automatically, the first thing we’ll do is disable the “Keep Project Folder Organized” option in the Project Inspector. 

To open the Project Inspector use the Project Settings Menu:

In the Project Inspector disable the option "Keep Project Folder Organized":

You can now close the Project Inspector. We will now use the + Button in the bottom left corner in order to add our main patcher to the project by choosing "Add New File…" and creating a Max Patcher called n4m-bundler.main.maxpat

In order to verify what just happened we can use the Finder (Mac) / File Explorer (Windows) and look at the N4M-Bundler folder in the user’s Documents/Max 8/Projects folder. It should now contain a folder called patchers that includes our freshly created n4m-bundler.main.maxpat patcher.

In order to have a place to organize our content for node.script we will also create a folder called node_content next to the patchers folder. Our folder should look something like this: 

N4M-Bundler/
├── patchers/
│ └── n4m-bundler.main.maxpat
└── node_content/

2/ Adding node_content to the Project’s Search Path

In order make the files in our node_content folder available to the project we need to add it to the project’s Search Path. For that click on the Gear Icon in the Project Window and use choose the "Search Paths" option. Max will open another window that allows you to manage the Project's Search Paths.

Using the + Icon at the bottom left of this window we can add another entry, which should give you something like the following:

Now click the Choose Button in that new line and add the node_content folder we created previously. After that you can rename the entry to Node Content but most importantly you have to enable the “Embedded” option. Your Project's Search Path Window should now look similar to this:

We will now create the simple script for this test scenario. For that open the node_content folder in a text editor of your choice and create a file called n4m-bundler.js with the following content:

const Chance = require("chance");
const maxAPI = require("max-api")
const chance = new Chance();

const generateRandomName = async () => {
        const name = chance.name();
        await maxAPI.outlet(name);
};

setInterval(generateRandomName, 1000);

This uses a package called chance to generate a random name every second and outlet it to Max.

4/ Patching everything together and installing the chance module

We will now edit our Project’s main patcher and install the required chance package from NPM. The simple idea is to show the random name in a comment box. Start off by create a node.script instance pointing to our n4m-bundler.js file. Before we do anything else we will install the chance package by sending a message "script npm install chance" to the node.script object:

Once done you should find a node_modules folder and a package-lock.json in the node_content folder and can remove the message from the patcher. If you like, you can use Finder or File Explorer again to verify that your project directory looks something like this:

N4M-Bundler/
├── patchers/
│ └── n4m-bundler.main.maxpat
└── node_content/
│ └── node_modules/
│ │ └── chance/
│ │ └── *** a whole bunch of files ***
│ └── n4m-bundler.js
│ └── package-lock.json

We will now use the prepend and comment objects to show the generated, random name. And add the script start and script stop messages to control our script:

Clicking the script start message should start the script and show a random name in the comment box every second. You can now stop the node.script using script stop.

We have now all the pieces of the puzzle at hand and can build our project. Choose “Build Application” from the File Menu, make sure the Format is set to Application and proceed. Max should now build a Standalone application containing our Patch, Script but also the chance module as a dependency.

The same would be true if we attempt to build either a Collective or a Max For Live Device.

Feel free to let us know if you have any further questions or feedback on this.

image2.png

image3.png

image5.png

image6.png

image1.png

image4.png

With the just released 8.0.3 update of Max 8, Node For Max now supports bundling dependencies (aka the node_modules folder) within Collectiv

node-for-max-bundling-in-collectives-standalones-and-max-for-live-devices

Thank you for your tutorial, but it seems there is something wrong here.

1. You wrote: "We will now create the simple script for this test scenario. For that open the node_content folder in a text editor of your choice and create a file called n4m-bundler.js".  But in the file listing the n4m-bundler.js file is located in the root project folder N4M-Bundler:

N4M-Bundler/
├── patchers/
│ └── n4m-bundler.main.maxpat
└── node_content/
├── node_modules/
│ └── chance/
│ └── *** a whole bunch of files ***
├── package-lock.json
└── n4m-bundler.js

2. You wrote: "We will now edit our Project’s main patcher and install the required chance package from NPM. <...> Once done you should find a node_modules folder and a package-lock.json in the node_content folder and can remove the message from the patcher." But the node_modules folder is created in the project root folder as well and not in the node_content folder:

├── patchers/
│ └── n4m-bundler.main.maxpat
└── node_content/
├── node_modules/
│ └── chance/
│ └── *** a whole bunch of files ***
├── package-lock.json
└── n4m-bundler.js

So, there is a question: why do we need this 'node_content' folder? Looks like the Node For Max bundling works well without creating the additional folder: add main patcher and js-file to the project and then point to the 'node_modules' folder to the project search path.

Can you, please, bring a little light on this issue?

thanks for pointing this. And yeah sorry, it was in fact a mistake in these ASCII folder structures.

If things work for you without the node_content folder: great! However, we do recommend the usage of a separate folder (node_content in this example) for the node related content in projects. The reasons for that are mostly practical. It's easier to find/navigate the Node related content and we provide a sort of space for npm to do it's disk related work (writing/altering node_modules, creating debug logs etc) in order to avoid having it change the project's root folder. Also in more complex cases you'll be including more user authored JS files for Node For Max  that are required/used for your script. This way adding a folder to the project ensures that they will be bundled as well.

Hope that makes sense? Thanks for following up and pointing out the little issue in the folder structures,

Yes, now everything works as expected. Thank you for this useful tutorial!

Hey Florian,

A very detailed guide, thanks. When i run the script npm install <package-name> in my case, axios, the package is installed and the patch does what it's supposed to do, a http request.

However, my finder does not show any package.json or node_modules created. So the npm is installing the packages elsewhere. This would be a problem for bundling and distribution. Any ideas?

glad it's working for you. In terms of the installation location - it's kinda hard to pin that down given the limited information but I can try.  To make sure you don't have any naming conflicts you can send a reveal message to node.script that should show the location of the file in your Finder or File Explorer. Is there a node_modules folder next to the file that gets revealed? If you are facing a naming conflict, we highly recommend namespacing your files, fe n4mtest.index.js where n4mtest would be the project specific namespace.

Fwiw,  npm install will  extend the dependencies of a pre-existing package.json file but not create one (only the package-lock.json gets created). The minimum requirements are a name and version field (you can learn more about the package.json here):

{
  "name": "mypackage",
  "version": "0.1.0"
}

Hope that helps a bit with understanding the functionality of npm install? Curious to see if there is in fact an issue with multiple files having the same name in your search path or if something else is at play.  Let me know and I'm happy to assist further.

Hi Florian,
Thanks for this, super helpful for what I'm trying to do. I'm running into an error where the node_content folder can't be added to the  application, and I'm not sure why.  The Max Console reads as follows...

Adding file message.mxe64
Adding file prepend.mxe64
Adding file comment.mxe64
Adding folder node_content
Unable to add node_content!
Investigating Audio Status.maxpat...
Saving Patcher Audio Status.maxpat...
Copying External live.numbox
Copying External adstatus
Copying External umenu
Copying External metro
Copying External number
Copying External substitute
Copying External toggle
Copying External pak
Copying External speedlim
Copying External delay
Copying External button
Copying External js
Copying External bgcolor
Copying External universal
Copying External ubutton
Copying External zl
Including File interfacecolor.js
Copying External led
Copying External textbutton
Copying External panel
Copying External spray
Copying External Uzi
Copying External loadmess
Copying External !-
Copying External buddy
Copying External dspstate~
Successfully built project N4M-Bundler to application.

(emphasis mine)

There are no other errors, and the application opens just fine, but obviously can't find the n4m-bundler.js file. I'm running Windows 10, 64 bit.

I'm at a loss as to why this could be happening and I appreciate your help. 

Thanks!
Stepan

Thanks for the reply Stepan. Please get in touch with our support as we might have to get a copy of your project in order to be able to reproduce the issue.

The issue " STEPAN DYACHKOVSKIY " mentioned is solved?
I have the same problem.

Adding folder node_content
Unable to add node_content!

I'm also running Windows 10, 64bit.
Thanks in advance.

I cannot reproduce the issue here on my end. If you still encounter issues with bundling on Windows please get in touch with our support so we can help you in a more detailed fashion.

Just a quick reply for the others having problems to have a working build on Windows (Max 8.1.3/64 bits).
If you try to build the standalone from the project window (menu : "Manage Project > Build Collective/Application..." ) it doesn't work (you should have the message 'Unable to add node_content!' in the Max Window).
If You build using the menu "File > Build Collective/Application..." in the main patcher window, it works flawlessly.

Hi Florian,
Never got a chance to thank you for writing this.
Just a quick note that second picture caption has a typo: Disbale
And I would like to request this information to be added to Max built-in documentation.
(node.script help patch)

@Omar thanks for the pointer, just fixed the typo. AFAIK the info is bundled with the Max application in the Working with Projects, Max For Live Devices, and Standalones. You can find the same content here:

https://docs.cycling74.com/max8/vignettes/03_n4m_projects_devices 

It would seem that this process is broken on MacOS Monterey. Using the node.script "open folder location" icon, it would seem that all of the javascript files are being sandboxed in a directory in the user library, and none of the node modules are being copied over.

Hi Eric,
thanks for the reply. Was this working successfully before on a different version of macOS?

Would you mind getting in touch with our support with a bit more insight on your project, workflow and version of Max that shows that issue in order for us to be able to investigate further if that's a general issue or something specific?

Thanks
Florian


Yes this is an internal application that we’ve been using since 2018 with previous versions of macOS.

I will reach out to the support team as you requested.

I still have the same issue Eric mentioned, Max 8.3 OSX Monterey. The nmp modules are not copied into  the dir Content/Resources/C74/packages/Node\ for\ Max/source/node_modules/ ,  and the package.json (v. 2.0.4)  does not declare them while  they are into the Project's package.json. So that the app tells that the script can't start (not ready).
Is it already solved in some way ?

why don't you copy files manully ?
At same time you could trash unneeded files from the package....


Yes, sorry to have not mentioned I already did it, the result being the same , node.script tells : 'node script not ready, can't handle message...' . Same result with message 'script stop' after having sent 'script start'.
This appears as soon as I declare  in the script require('sqlite3'), even if :
- I do not call at load time any sqlite function (but only require) ;
- declared [standalone @database 1] in the main patch (or not, it should not be necessary since sqlite may run apart in some node process) ;
- and copied manually all of the required modules and their dependencies.

Here is the node.script : 

const path = require('path');
const Max = require('max-api');
const sqlite3 = require('sqlite3');

Max.post(`Loaded the ${path.basename(__filename)} script`);

Max.addHandler("testing", () => {
 Max.post("test");
})

When starting it, node.script debug tool tells "Process Running" (in green), but the first Max.post does nothing, and the 'testing' message from Max returns 'Node script not ready can't handle message testing'.

Building from Max/File/ menu or from Project/manage does not make a difference.

A point would be that the file 'package-lock.json' which is at same level of the app's Content node_modules directory is not updated even if I consolidated the project, so that it does not mention any of sqlite3 and its dependencies. If I copy the 'package-lock.json' from the project folder, the last declaring  sqlite3 package (version 5.1.6), it makes no difference.

I'm running Monterey on x86 macbookpro (Max 8.5.3).



then your problem is sqlite related and not failed build process ?

May be, not so clear. May be the build is not working properly : if the project is well defined, the generated app is not complete and requires manual intervention after build but this manual intervention does not yet fix it. And it may have something to do with sqlite3, or the OS ? so that the Cycling annoucement relating that it's better to deal with database using node is not true for me, until now the sqlite embedded in Max built works much better.


I will have a try asap on MS-Windows but anymway I'd like to make it work on OSX


So, we did a try of node.script using Max8.5.3 on Windows10  which is very strange, 
see  my new topic: https://cycling74.com/forums/node-script-on-windows10
The same with Max OSX Monterey is working, but not yet with a Max built app when requiring sqlite3.

Any idea ?



Very nice and amazing work . Must be appreciated and proud of this. Congratulations.

Hi Fred, would u mind providing a minimal example project that can be used to reproduce the issue? I'm happy to look into it and see what might be the culprit but would prefer to limit the scope and take it one issue / confusion at a time as there are a lot variables at play in your reports in this thread.


One general question - are you successfully running npm install on the node.script object prior to building / bundling? I'd also be curious if you are only encountering this with the sqlite3 package or also other npm dependencies given that sqlite3 contains a native dependency and something might go wrong with node-gyp in case they can't be fetched as pre-built binaries.

Hey,
when i instantiate the node.script object with n4m-bundler.js .
the max console prints "node.script: n4m-bundler.js: No such file or directory" , although i have added the search path in the Project Search Paths.
I need to add the search path to the Max Search Path ( options->file prefs-> ... )

furthermore when I send the msg "script npm install chance" the node module is not in the folder, nor a .json. I tried to find it via "reveal" but it only points to "n4m-bundler.js".

Node For Max Bundling in Collectives, Standalones and Max For Live Devices

frank desben

Dear all,
I am getting this error message when trying to use the freesound example (from the Cycling74/n4m-examples github page) with node.script:
"node.script: No value for key FREESOUND_CLIENT_KEY in .env file. Please make sure to create a file called .env with a Freesound API Client key."
Can someone tell me how to fix this?
Thank you! Frank

Cycling '74 Forums

I never saw an update for JavaScript in Max to provide this functionality. It would still be nice to have it. Eric

Hi OP, Did you ever figure out an elegant way to send messages to inlets via JavaScript? I just bumped into this issue.

I have a max patch that dynamically creates and removes a ton of objects via the this.patcher.newdefault() and this.patcher.remove() ...

FWIW the least painful solution seems to be to use scale(2, 2) when when drawing the cached image (on a 2x-sized canvas) and then ...

Thanks Ali. It seems indeterminate from testing.

Hi, I'm having the same issue...

Hey, when i instantiate the node.script object with n4m-bundler.js . the max console prints "node.script: n4m-bundler.js: No such file or ...

Just a quick heads up that the freesound example got a little update which fixes the issue you were seeing.

cycling74.com/forums/any-plans-to-update-support-for-recent-versions-of-js#reply-58ed21d5c2991221d9ccad8c I think this is sill valid.

Sorry my silly question, I just wish to get a broad view of the JavaScript integration in Max.