@KamilaSzewczyk can you plot the graph of the taylor approx and the graph of some function points calculated from the internal function implementation?
"Could not load bridge interface functions. (Error 0x00000000: "libhostfxr.so: cannot open shared object file: No such file or directory")", but I'll ignore it
@KamilaSzewczyk No, ]plot is very bare bones. It is just a cover for a tiny subset of SharpPlot using defaults for everything but the chart type. You can look at the source for ]plot to see how it does it, and use that to learn using the full SharpPlot. (There are also other ways to learn SharpPlot, of course.)
@Adám if there's a ready-made way to interface with Python, perhaps it would not be so difficult, except for mapping the Python arguments into something sensible in APL.
@Adám I don't know enough about APL, RIDE or JupyterLab to be able to answer your question. All I notice is that the Python kernel offers the debug option (from which I can inspect values of vars, manage breakpoints, etc.), whereas with the APL kernel the debug icon is grayed out.
Given that the prevalent workflow in APL (AFAIK) has significant emphasis in the think-and-explore phase, JupyterLab would be even a better IDE than RIDE. Of course there are many RIDE users who still need to be supported, but being able to leverage Jupyter work could free up some resources for Dyalog to focus on the language, the interpreter, the compiler, etc.
@xpqz Wow, that's amazing. Yeah, we should definitely support that. I find that the interactive debugging is key in APL's dev UX. Can you log an issue against the kernel to support that? (We should also start promoting Lab instead of Notebook.)
@AlexB RIDE isn't legacy. It is a quite new system, and is essential to have an interface to remote interpreters in the cloud and on headless servers like AIX.
@Adám tbo, as RIDE already has a protocol for all of this, it certainly seems to me that it should be doable in the dyalog kernel. Similarly, the editor debugger extensions for VS Code etc.
So supporting (here: developing) the Jupyter Kernel doesn't free up any resources, only adds more to our workload. But don't get me wrong: I still think we should do it.
Gil has a proof of concept implementation. It needs a bit of polish, and he said it highlighted some weaknesses in the ride protocol when used as an integration tool for m ore general tools.
https://palaiologos.rocks/posts/taylor-apl/ - taylor series https://palaiologos.rocks/posts/asmbf-ski-p1/ - register allocator in APL https://palaiologos.rocks/posts/matrix-rearrangements/ - a weird matrix problem https://palaiologos.rocks/posts/nonalphanumeric-c/ - a compiler written in APL https://palaiologos.rocks/posts/dx-rewrite/ - a (from retrospective - slightly poor) domain extension library for APL
I just got bitten by the ⎕IO problem... :-( I was testing @KamilaSzewczyk's code as modified by @Adám and I got RANK ERROR a few times, till I realized that I had ⎕IO←0
@Wezl-acautionarytale Looks good. Change "download" to [url instructions] and wrap "documentation" in [url CodeBerg]. Add the | in `[[user:Wezl|], and maybe a list of primitives? Go ahead and add it to the APL dialects' Derivatives section.
"variadic functions" should be [[ambivalent function]]s
Today I'm back to work. Even if I'm still not completely converted to APL-ism, I didn't miss Typescript a single bit.
(I might never be, actually. Even if I appreciate a ton of things APL, I think Haskell is still a better balance of verbosity. Of course, that may change...)
Yay! @xpqz, it's not cheating. The point is to be able to plot with flexibility of output from APL in Jupyter. I think this is very, very promising. For me it's just a hobby, but scientists using APL for work would be thrilled, I think.
@AlexB The cheating bit was that I hard-coded the image name reference in a markdown cell.
@MasterQuiz It doesnt' really matter, as long as you can reference it. If you want to use it in jupyter, it needs to be 'below' the folder that jupyter thinks is root.
@xpqz Of course, but you showed that it's possible. From what I guess, it is not a ton of coding that's required. Sadly I cannot offer more concrete help, because I am tied up with time and my line of work is elsewhere. But I enjoy spending time in this forum and I hope that someone might find the time to develop the support code.
@MasterQuiz What's the problem with using the whole path? In fact, it would be good practice to do so. This way the modules can be in separate directories...
@MasterQuiz Not sure you'd want to do that, to be honest. The pynapl folder contains a lot of stuff, and you'd need to root your session right in the midst of that in order for that to work, as far as I can see (disclaimer, literally only looked at pynapl for the first time half hour ago).
You can see what @RGS does towards the tail end of his video.
No need to ping me twice. have some patience instead. If I don't answer you immediately, then it is probably because I'm doing something valuable. In this case, I was giving food for my son.
