I'm on this project where we want to do #realtime #radar but are sort of starting with nothing (apart from world-class radar transmitters, receivers and expertise...)

One very smart but non-#software person wrote a bunch of good #signalprocessing #code and some "gets the job done" #gui code

Or it did until we went higher bandwidth

Last week I rewrote all the non-sigproc parts into #pyqt and #pyqtgraph. Today I benchmarked both.

Exactly the same speed....except pyqtgraph is

THREE ORDERS OF MAGNITUDE

faster than #matplotlib

#python peeps, please hear me. mpl has its place and uses. High data rate animated displays is not that place.

It's one of THOSE days

I overslept by an hour. The tuxedo #cat had to use his little paws to roll me out of bed. (He was hungry but worse I was not On Schedule, something he cares about a LOT)

But also I only had to get up early because #IT made some change on my laptop that prevents me from logging in to #wfh

When I got here, I found that a request I put in to unlock an unrelated account on a different network had ALSO not been done. But at least the ticket is closed!

And the thing I have to work on is annoying. A project I definitely want to be on needed a display. I made one in #python #pyqt #pyqtgraph.

Oh, they meant a #web #app. OK, so I redesigned the backend to both would work.

Everyone likes THAT except for some external guy who AFAICT hasn't done any actual work on this project. The only reason his opinion matters is that he's the one actually deploying the app...?

And instead of modifying the code to be how he wants it he wrote a sample app for ME to follow?

And I don't think that's even going to work. #plotly needs a session ID to get persistent zoom and some other things right. This guy's design means everyone has to share a session ID. I explained this to him.

So the outcome of starting from his code is either that I'll be proved wrong or I'll have wasted a day proving HIM wrong. Two wonderful possibilities.

I love #pyqtgraph for 2D scientific and #dataviz #graphs in #python. And now I'm starting to really love #vtk for 3D.

https://vtk.org/

It has a lot of the same #science adjacent type plotting. It works well with python. (And there's a #javascript #api but I haven't used it...yet). And it is *extremely fast*.

The only problem is that there's a lot to it and I don't really understand the workflow. (I feel like there's an extra layer. There's a "source", a "mapper", an "actor" and a "renderer". One of those seems superfluous.)

There are a lot of examples on their site and elsewhere. And there's a lot of documentation. But I haven't yet found a good combination tutorial that has worked examples that are also explained.

I really just have to use it more and get it into my head.

Fortunately, I have a weird #math/#geometry/#orbitalmechanics thing I want to try to solve (or at least visualize). Intersections of multiple ellipsoids and cones....

Welp. I spent WAAAAYYYYY too long computing zoomed ticks/labels on my custom #python #pyqtgraph polar plots

But at least I'm not doing a tremendously great job and the code is incomprehensible!

To be fair, I looked at some other graphing packages and none of them really worked how you'd expect in all cases.

Desmos has the same behavior as mine as 0,0 crosses over the viewable area.

Both Desmos and another only add ticks in the radial dimension past a certain point.

That seems not helpful, so I'm beating the pros! And I shoved my last remaining error in a try/except. So....done?

#software #engineering

A while ago I created my own polar #graph (θ,r but also built az,el on top of that) inside the excellent #pyqtgraph (#python)

I also implemented a spaced tick computation similar to the one it uses. However, I didn't make it work in zoom for lack of time

I got a bug report about how the ticks/grids/labels "disappear" during zoom (i.e. stay at a larger scale while the zoom goes inside) so I thought I'd just fix that

This turns out to be non-trivial. Or at least I can't think of a simple way to do it

I was getting the xy coords of the corners of the zoombox and computing the θ,r to figure out where I was. That doesn't really work, tho. Imagine the zoom box just to the right of 0,0: θ->0,180. Now imagine it just barely containing 0,0: θ->0,360.

I think I need to measure angles from 0,0 but that means....uh....I need a whiteboard and some rubber duckies. I wish my nerd kids were awake at this hour...

#data #software #engineering #math