Aras Pranckevičius
I was today years old when I realized that just like OBJ files are really useful to write while debugging 3D stuff, SVG files are useful for 2D cases! So obvious, but I never did it before! Well, turns out an old dog can still learn some new tricks. Example: https://projects.blender.org/blender/blender/pulls/156622
Mar 31 at 5:56pmWeb
Show threadScherzog von Beast OilMar 31

@aras I recently briefly used SVG as a makeshift timeline visualization for profiling!

Unfortunately I soon found out that neither web browsers nor inkscape can really handle Large SVGs so I ultimately switched to simple binary dumps and a standalone timeline viewer built with SDL_Renderer...

Show threadAras PranckevičiusMar 31
@scherzog yeah very much like you wouldn't want to dump "something really complex" 3D into OBJ, there are practical limits how much 2D stuff you can dump into SVG before it becomes not usable :/
Show threadArseny KapoulkineMar 31

@aras Yeah I used SVG dumps a few weeks ago to debug!

The big downside is that "obvious" viewers are pretty bad at handling large files :(

One other annoyance is that SVG doesn't seem to have layering/Z indices, which makes it non ideal for dissecting parts in the viewer, and forces a specific output flow in the code which can be inconvenient. OBJ doesn't have either problem.

Show threadMikko MononenMar 31
@zeux @aras i have used svg groups for that. Surely depends on use case if they work
Show threadJosh SimmonsMar 31
@aras one fun thing is that Firefox is *significantly* faster at rendering svgs than chrome etc. so if you have a moderate chonker it's worth trying that. Bonus points for embedding JS to manipulate the svg. :')
Show thread🌈 Andrew ☄️Mar 31

@aras also if you’re debugging a non-Cartesian graph like structure then graphviz dot files are often handy. It’s a simple text graph notation like “A -> {B C}”

https://en.wikipedia.org/wiki/DOT_%28graph_description_language%29

DOT (graph description language) - Wikipedia

Show thread🌈 Andrew ☄️Mar 31
@aras I even used it a lot in the early stages of a Delaunay triangulation implementation (which was Cartesian) as it was easier to set up than SVGs
Show threadSean BarrettApr 1
@aras I'm guessing your rasterizer isn't actually NP-complete, but https://en.wikipedia.org/wiki/Boolean_satisfiability_problem
Boolean satisfiability problem - Wikipedia

Show threadFabian GiesenApr 1
@nothings @aras *adjust glasses* it's a separating axis _test_. The _theorem's_ name is the Hyperplane Separation Theorem (or Hahn-Banach for the most general version).
Show threadFabian GiesenApr 1
@nothings @aras this is a -1 nerd point mistake and will be added to your permanent record!!1111
Show threadAras PranckevičiusApr 1
@rygorous @nothings I am aiming for my nerd score to wrap around to INT_MAX!
Show threadTom ForsythApr 1
@aras @rygorous @nothings But where will you find the nuclear weapons?
Show threadSean BarrettApr 1
@aras SAT means clamp not wrap, you poor fool https://en.cppreference.com/w/cpp/numeric/add_sat.html
std::add_sat - cppreference.com

Show threadAnders StenbergApr 1
@nothings @aras Seeing this out of context for a second I thought std::add_sat had managed to get saturation wrong.
Show threadRkApr 1
@SonnyBonds @nothings @aras yeah, me too! and the fact that it is even an option to consider speaks volumes about <gestures widely> everything!
Show threadAras PranckevičiusApr 1
@nothings oh shoot! I’ll reply on SAT
Show threadPer VognsenApr 1
@aras @rygorous @nothings Your scholastic aptitude test score won't help you here.
Show threadFabian GiesenApr 1
@pervognsen @aras @nothings nor will your summed area tables!
Show threadKojack 🦎Apr 1
@aras One of my students was making a game about wandering around a city talking to people. He wanted to know where players were going during play tests. So I wrote an SVG generator for him that logged a map of the area and showed the trail made by the player.
Show threadLeszek GodlewskiApr 2
@aras I used SVG as debug output for a local light shadow atlas reorg algorithm at Frostbite. It came up with a sequence of moves that could safely be performed in parallel with a single compute dispatch without requiring a second atlas for ping-pong, but the sequences were sometimes rather unobvious, and thus needed visualising. 😁