@halirutan @Szabolcs @AnyoneWhoCompiles do Association-based pure-functions atuocompile at all? Do string functions, even? Say I want to use #Type == "table" && Length@#Children == 2 & as a selection function. Is that always going to be at toplevel?

I'm a bit surprised at how other SE communities are in comparison to this one. I feel like the chances of getting a useful answer here are far higher than some other communities...

I have an array of coordinates stored as a GeoPosition. Is there a simple way to convert this to x,y coordinates with some type of projection? I know I can visualize them with GeoGraphics. But I just want x,y coordinates for further processing.

@Szabolcs You can try and use GeodesyData I think

But IIRC it has a very limited number of projections

and weirdly no support for things like EPSG numbers

from the docs:

GeoPosition[GeoPosition[{40.112981, -88.261227, 192.868}, "ITRF00"], "NAD83CORS96"]

There's also GeoPositionXYZ

which seems to have similar semantics

@CarlLange how what?

@user6014 Less helpful, I suppose?

I'm doing some GIS work but I find GIS.SE to have a lot of un-interacted-with questions

Here, I think nearly every question I've ever seen has comments on it, at least

@C.E. the full CSS selector syntax is very non-trivial to implement... (and especially to do efficiently)

@b3m2a1 Yes, I can imagine.

I'm currently doing it two ways, first by seeing if I can convert the selector to a simple expression using Select and then if I can't using a BFS over the graph of my XML

But the latter is a huge pain

As I need to maintain a bunch of traversal data because CSS lets you pick arbitrarily deeply nested structures

And any sibling in the structure

I'm also not sure what node-ordering Mathematica applies to its BFS so I'll need to be a bit more intelligent about handling that, too, I think

So several years after acceptance and many more than ten upvotes, I get an "Enlightened" badge award for mathematica.stackexchange.com/questions/59463/… The last upvote was Sep 17. Can't imagine what triggered the notification.

@b3m2a1 Maybe you can do it recursively, e.g. if you have a selector like "body div.comment a" then you can first do BFS to find body, then you launch a new BFS on the subgraph looking for "div.comment a", then when you find div.comment you launch a new BFS on its subgraph looking for "a". That way you don't have to keep track of traversal data.

@C.E. the traversal data isn't so bad...well it wouldn't be if Graph weren't buggy.

For some reason it seems to think that I have no connected nodes in my tree...

ok :P

I haven't used the graph functionality seriously enough to expose its flaws, I get a feeling I should wait for the new graph functions before I try...

Ah...of course. It fails for undirected graphs even though the documentation specifies it works for undirected graphs

Oh even better. I had a mixed graph then I removed the single directed edge. But it didn't realize that made the graph undirected.

I guess it doesn't work for mixed graphs

?

It's a bit slow, but it works:

blech@"Select"["code"] // Length // RepeatedTiming

{0.001, 40}

blech@"Select"["td code"] // Length // RepeatedTiming

{0.0085, 31}

blech@"Select"["li code"] // Length // RepeatedTiming

{0.0085, 2}

I need to find a test-set of selectors though to properly try it out

The first just went directly to Select (after running through a parse/selector compile process) so that's about as fast as I can expect it to be. The second runs through the BFS as I need that parent information, too, and a BFS allows me to just do a single scan.

@C.E. fun article. One cool extra thing that'd be a fun example of how you can basically get to the point the project post you linked to got to would be to do some type of orientation estimation again at the end so that the labels you supply are aligned with the actual orientation of the card.

I feel like you must have all the code already to do that, but it'd be a fun demo I think

@b3m2a1 yeah, I already have the code for that: I just need to print the label onto the perspective corrected version of the card, and then project the label onti the image. It is the same thing that I did when I projected a picture onto a wall towards the end in this article. For the Pokémon cards I have higher ambitions, I want to do 3D augmented reality and place models of Pokémons on top of the cards.

I've started, e.g. by writing this, but it's still some way into the future because I don't quite understand how to figure out how the cards are aligned in 3D yet.

Maybe I'll add a line saying how to orientate the labels. Thanks for the idea!

@C.E. very cool! I always enjoy classic (i.e. pre-neural net) computer vision.

@C.E. Nice post! I presume you've seen PokemonData, right?

wait, it's not pokemondata

Ah, it's just a big list of Pokemon entities

Interpreter["ComputedPokemon"]["Charizard"]

wow

pity, no Image3D :)

Hi! I am new to mathematica and have some questions about the NDEigensystem function. I tried reading the documentation, and don't quite understand, would someone here be willing to talk to me a little bit about my problem...I feel the issue is probably quite trivial, and I am not specifying the correct options or something, so I am not sure it belongs on the main site.