Controller Testing Apparatus

[SeanMcOwen]

WIP

Tasks

• [ ] Allow for passing one's own signals into boundary actions (as opposed to generators in the boundary action)

Resources/Commentary

• Similar to this video: https://youtu.be/S8aaMkqtxXw
• suppose there is a logical block myBlock i want to do testing on i want to import it from where ever it is stored i want to know what its parameters are and to be able to set them or overwrite them i want to know what its inputs are (whats its domain space) so i can feed it a stream of signals i want to know what its outputs are (whats is codomain space) so i can write some simple metrics and/or make some simple viz
• is there a way i could set up a notebook where the blocks are imported the blocks are parameterized (that is their parameters are set or overriden locally) some input signal sources (with appropriate domains) are declared the wirings are assigned the locally defined model is executed the results data is available to compute over

[SeanMcOwen]

For this: i want to import it from where ever it is stored i want to know what its parameters are and to be able to set them or overwrite them i want to know what its inputs are (whats its domain space) so i can feed it a stream of signals i want to know what its outputs are (whats is codomain space) so i can write some simple metrics and/or make some simple viz

Each component has parameters_used, domain, and codomain as attributes on them. If I create an MSML function that essentially says print out component summary that nicely places them all in, would that work?

From Z: this is directionally correct. and yes i definitely need some way of declaring local wirings -- the main thing one does at this stage of testing is about wiring/rewiring adding things in and looking at how changes in wirings and changing inputs and outputs affects various signal (not so much specific values as relationships between those values).

[SeanMcOwen]

For this: the blocks are parameterized (that is their parameters are set or overriden locally)

• Does it suffice to say the parameterizations for any relevant blocks?
• Does this MSML issue solve it? https://github.com/BlockScience/MSML/issues/504
• Basically I would have a way to pass in some number of components and have it return here are the important parameters?

[SeanMcOwen]

For this: some input signal sources (with appropriate domains) are declared

Some of these are just boundary action options but are not necessarily signals versus random distribution options

• I could add boundary action options which are pass through signals and then add in parameters like "mining difficulty signal" so that instead of mu/sigma of difficulties it is like you feed in difficulty list
• On the other hand, one could take the mining wiring for example, and just cut off the boundary action and instead have it be that basically components[1:] of mining run and you pass it spaces of the domain time that you design in the notebook

[SeanMcOwen]

For this:

• the wirings are assigned

I might be realizing that maybe you are asking for a local wiring builder? I.e. right now wirings get defined in that python file but if you wanted to do on the fly you would do it with multiple lines where you say like:

spaces = [...] spaces = component1(state, params, spaces) spaces = component2(state, params, spaces)

But are you saying you want a local temp kinda builder where you can say,

myWiring = [component1, component2, component3]

then that can be executed, for easier and cleaner GUI of building?

That would end up being an MSML issue that I could build, it's just that given weekly hours > 40 in this spike phase, that does end up taking away from Quai specific time so it is a question of whether I should divert time to there ya know?

[SeanMcOwen]

Maybe even for a summary of component thing add in the mermaid graph to be displayed if we are dealing with a wiring