wanted: form and color detector (not throughput)

[ludorino]

I have my first 2-layer MAM running (with different crystal creation for both etc.), and I'm having a lot of fun. Thanks for this cool game. I love it!
Now there is two cool next steps: Sure, going 3+ layers, yeah. But I would also like to optimize some stuff. Particularly when switching tasks, there is a lot of faulty shapes "in the flow". I have some auto trashing at some but only few points. I would really like to add more trash control to fasten up the transition from one task to the next, however:

The current wire stuff only allows the "virtual" creation and modification of shapez and logic gates. We can also use these wires to control the "non-virtual" belts and/or test the passing-by shape against a virtual shape on the wire. So, the wires only work wire-->wire and wire-->belt (belt filter), but not belt-->wire.
To be correct, that is except for a very manual button and the "belt detector" which is measuring the throughput by counting passed shapes (without a reset btw!?!?). Aside from this (IMO not very useful) sensor, there is no sensoring device or method to use seen/placed "real" stuff to trigger/set wires (belt-->wire).

I guess, the wires would be even more fun with more "sensoring" devices and such. For example, a form/shape and color detector could be working "inverse" to the belt filter: the shape passes and its "value" (shape/color/...) would be passed to the wire. That way, it would work similar to the signal producer, however, not manually set but dynamically changed accorging to a passed shape. Analogously, there can be more sensors like a layer detector or whatever. I would further guess, that more devices could be needed to make sure, the detected bad shape really goes to trash somewhere later down the belt. For instance a time-delay gate might be needed as well as a count-down gate etc. etc.

I am quite sure, the devs have already thoughts about all this and I would be eager to know, what's coming in some next update(s). Leak it out! :-P

[CaptainFlint]

The belt sensor it exactly what you are looking for. The speed measurement is purely decorative, but the sensor has a wire output, and this output receives the shape that passes through the sensor right now (or has "null" signal if there is no shape currently).

However, there are some gotchas lying around for you to step into. The main problem is, whatever shape you detect with that sensor, you cannot do anything with it right there and then, because you can't combine the sensor with any other building. The shape has to pass further along the belt - and by that time whatever signal you had calculated in your wire circuit will also disappear, it will reflect the (possibly) other shape that came under the sensor instead of the previous one.

Therefore, using sensors is quite fiddly and not as reliable as many players would like. You have to ensure regularity in shapes input, guarantee that the belt has always the same throughput, and adjust the buildings further along the line with precise timings, so that when a shape passes the sensor, you are processing exactly the same shape somewhere else, using this signal. Basically, you need the time-delay gate you're talking about - but there is no such thing in the game currently.

[ludorino]

Thx!!! :-) I must say, this is far from obvious, but agreed, I could have tested that! :-)
OK, I will have to think about how to use this without a time-delay gate or something else missing.

[ludorino]

It's a pity that the button does not have such "hidden feature" too: allowing to set it's value by wire, thus, having a 1-bit memory.
I'm wondering, if one could build a latch without 2 NAND gates and bulky wiring...

[CaptainFlint]

You can combine a button with a signal generator using the control gate. So that when the button is pressed, the shape from the signal generator goes down the wire, and when the button is released nothing is sent (or a different signal is sent). Or vice versa.

[ludorino]

Hmm...
https://steamcommunity.com/sharedfiles/filedetails/?id=3402191313

[Khagan]

I've found the belt filter to be all that is needed to keep the belts clean. You should know what shape is expected at each point, so let that through and send anything else to trash. The belt sensor has very occasional use for flow control, but I don't think I have one anywhere in my MAM.

[ludorino]

That's right, but nevertheless, it would allow very interesting contraptions that way. As others pointed out, it shouldn't _only_ be MAM as a goal.

Here is some working (but slightly lagging) sensor-based thrasher:
https://steamcommunity.com/sharedfiles/filedetails/?id=3402259673

[VictorNet]

I don't think there's any point in purging a MAM. I tried and all I got was an unstable outlet with gaps on the conveyor belts. I pushed my research to the point of finding a solution to purge the paint (valve and painter with looping forms). In reality, the best way to go from one color to another quickly is to use normal hoses and banish space hoses. A normal pipe behaves like an electric wire - it contains nothing. A space pipe, on the other hand, can contain a considerable quantity of paint.

Translated with DeepL.com (free version)

[ludorino]

Actually, it's more a question of fun and not so much of return-on-investment ;-)
Purging the paint was also on my mind, thx for the idea with the loops and over-painting. But yes, ROI-wise it's not "worth it", I can do what I did the whole time: just wait and let time fix the inconsistent "state" producing faulty shapez. ;-)

I did not get the thing with the "hoses"? What is a "hose" and what did you mean there?
Edit: I guess, I get it now -> belt for paint = hose, ok.

[Heiinz]

funny. i build a shape and a color detector (from signals) before i noticed that there is a thing that does both a once.

[ludorino]

It's a pity that the "Unstacking" does not works inverse. Sure, the unstacking is correct the way it is = from top to bottom. But for the MAM, I have to find a way to work the layers the other way. Do you do this for arbitrarily many layers automatically, or do you have fixed MAMs for 1/2/3/4-layer shapes?

[Heiinz]

Originally posted by ludorino:

It's a pity that the "Unstacking" does not works inverse. Sure, the unstacking is correct the way it is = from top to bottom. But for the MAM, I have to find a way to work the layers the other way. Do you do this for arbitrarily many layers automatically, or do you have fixed MAMs for 1/2/3/4-layer shapes?

i have a complete modular MAM where i currently try to figure out how i will do color mixing but i can copy the modules indefinitely because it is a manifold without any trash.
the modules are always all 4 pieces of a layer.

[Khagan]

Originally posted by ludorino:

It's a pity that the "Unstacking" does not works inverse. Sure, the unstacking is correct the way it is = from top to bottom. But for the MAM, I have to find a way to work the layers the other way. Do you do this for arbitrarily many layers automatically, or do you have fixed MAMs for 1/2/3/4-layer shapes?

Recursion works here. I have a chain of subfactories, each responsible for one layer. The one at the front strips off the top layer of the shape description, and passes the rest to the next along the chain. The constructed shape goes in the opposite direction: the rear subfactory makes the bottom layer, and passes it forward, and then each layer gets stacked on top as it moves to the front. If the shape has fewer layers than the maximum the factory is built for, the rearmost couple of subfactories have nothing to do (which is fine since all the workers are on zero-hour contracts).

[CaptainFlint]

Originally posted by ludorino:

It's a pity that the "Unstacking" does not works inverse. Sure, the unstacking is correct the way it is = from top to bottom. But for the MAM, I have to find a way to work the layers the other way. Do you do this for arbitrarily many layers automatically, or do you have fixed MAMs for 1/2/3/4-layer shapes?

I built my MAM in such a way that it can be as many layers as I want. Although I would need to increase the amount of unstackers in the analysing scheme, I currently have only 4, so I can't use it for Insane shapes. But if I add another unstacker, the rest of the MAM will remain the same and will be suitable for building shapes with any amount of layers from 1 to 5.

Some details below (under spoiler, in case you want to figure everything out by yourself):
I agree, unstacker does seem counter-intuitive at first, but I managed to think "in reverse" and work it out. I just completely ignore the total amount of layers in the shape, and start counting from the top - exactly as unstacker returns them. So the top layer for me is always number 1. The layer below is number 2, and so on. Any extra layers (if any) are considered empty. So for a 2-layer shape, for example, the top layer is 1, the bottom one is 2, and layers 3 and 4 are empty. It does feel a bit "wrong" at first, since the bottom layers kinda go "below" the shape, while the empty layers above remain "ignored", but it works.

And then I combine the final shape taking these layers in reverse order: the layer number 4 (if present) comes into the No.3 builder, and the No.3 layer shape is dropped on top of No.4 (and then goes through crystallizer, if it's a CROSMAM). Then the resulting 3+4 shape goes into No.2, which is dropped on top of that to get 2+3+4, and so on. If there are fewer layers, the last builders just don't do anything (plus, obviously, I have filters to make the shape pass the stackers if there is no previous layer).

P. S. Typing too slow, it's basically the same as what Khagan described just above.

[VictorNet]

@Ludorino Sorry if my terms aren't clear, I use a translator. What I mean to say is that space pipes store liquid whereas normal pipes do not, so normal pipes should be used to avoid the color change taking too long.