humbletim / glm-js Goto Github PK

Hi there. Awesome work. Thanks very much for working this out.

Noticed that no matter the type of vector built, FloatArray32 JS type is what is used under the hood.

const glmIvec = glm.ivec3(-3, 0, 0);
const glmUvec = glm.uvec3(-3, 0, 0);
const glmVec = glm.vec3(-3, 0, 0);

console.log('glmIvec', glmIvec);
console.log('glmUvec', glmUvec);
console.log('glmVec', glmVec);

I would expect that once the library takes care of the glm type specifics on the semantic side, the closest JS type array match would be used. This would be for memory efficiency also, but mainly to avoid misunderstandings/bugs when memory strides are passed into the graphics API functions.

Taking the chance to ask if PRs are welcome, maybe to fix this or for example to add typescript types to the library.
Thanks.

Using the library on a page in IE11 and the following errors are showing int he dev console:

'glm_vec3_Error' is undefined

Also If I type our glm in the console. it is showing as null, in other browsers it shows the full glm object.
Typing GLM in the console in IE11 shows an object but it does not have the function vec3() that I need.

For more precision calculation it would be nice to have access to dvec and dmat datatypes.

I previously mentioned this here but figured it would be more appropriate to write a new issue for it. I've been using this library pretty heavily to build a 2d graphics application, but I've been hurting from a lack of mat2s. I expect a lot of people might have similar use cases since 2d graphics application are pretty common in web browsers, likely even moreso than 3d applications, and people could benefit from having a common frame of mind for both 2d and 3d problems.

rough notes:

• document which versions and dependencies are needed (and where to put them)
• add example to readme of how to run a command line build and the tests
• maybe wire into package.json so 'npm test' can be used
• integrate the native GLM comparison tests into main build/test process
• maybe get git hooks / one of the automated CI services going to run tests automatically on commits

note: comments below brought forward from resolved issue #2

@Beuc

Note: I wanted to test my changes but I cannot run 'make test' or 'make build' from nodejs (tested with v4.7.2 and 6.9.2), I get various errors:

• looks at closure-compiler somewhere in /tmp
• requires build/VA_ARGS.js which needs to be build manually (missing make dependency)
• uses the non-standard 'load' keyword
• errors related to "ES5 strict mode"
  I'm not sure how I'm supposed to run all this :)

@humbletim

And yeah... my hacked-together build system needs some TLC. Fortunately, you can also bootstrap glm-js from source (ie: without any build system or minification); here's a quick snippet to demonstrate:

• web page
• the underlying gist

Shouldn't this be something like if(T in F)?

Hi,

I noticed that glm.ortho is not present in glm-js.
C++ GLM implements both glOrtho and gluOrtho2D under the "glm::ortho" function using function overloading.
https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glOrtho.xml
https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluOrtho2D.xml
http://glm.g-truc.net/0.9.8/glm-0.9.8.pdf

I use glOrtho2D to initialize a 2D projection matrix so as to draw sprites.

Last push date is 5 years ago! Is the project dead?

I might be using this wrong but I just found out about it today and so added to these benchmarks

http://greggman.github.io/webgl-matrix-benchmarks/matrix_benchmark.html?tests=glmatrix,twgl,glm-js,glm-js-gl-matrix

Also it's possible rotate is buggy. I benchmark both glm-js and glm-gl-matrix. They're exactly the same except for which library is included but glm-js returns different results for rotate then glm-gl-matrix

The code to test glm-js is here

https://github.com/greggman/webgl-matrix-benchmarks/blob/master/libs/glm-js/glm-js.html

Hi,

Thanks for writing a JS library compatible with GLM!
Much appreciated to try and port my https://en.wikibooks.org/wiki/OpenGL_Programming#The_basics_arc tutorials :)

I found a bug: glm.lookAt returns a quat rather than a matrix.
Cf. http://glm.g-truc.net/0.9.8/glm-0.9.8.pdf §5.2.
At first glance, stripping "glm.quat" in the various implementation of 'lookAt' in source code fixes the issue.

Note: I wanted to test my changes but I cannot run 'make test' or 'make build' from nodejs (tested with v4.7.2 and 6.9.2), I get various errors:

• looks at closure-compiler somewhere in /tmp
• requires build/VA_ARGS.js which needs to be build manually (missing make dependency)
• uses the non-standard 'load' keyword
• errors related to "ES5 strict mode"
  I'm not sure how I'm supposed to run all this :)

Hi,

I noticed that the matrix swizzle is incomplete.
For instance I wanted to use .xyw, which is missing, when converting a 3D transformation matrix to a 2D one.

    var mvp = ...
    var mvp2D = glm.mat3(mvp[0].xyw, mvp[1].xyw, mvp[3].xyw);

I had to manually expand it (see https://gitlab.com/wikibooks-opengl/modern-tutorials/blob/master/2d-webgl/sprites.html).

https://github.com/cheako/cheako-vulkan/blob/5fe88eabd4d742bc6554621a8509e2acb78ce2d2/015test.diff

I found out it was 2696, using duktape. I've opened a bug with them about not providing a stack trace in that instance, but it wouldn't hurt to have these evals protected by try/catch.

• Adopt a standard four spaces / indentation scheme.
• Get rid of mixed spaces/tabs! (ugh, shame on me ;) )
• Add an ESLint config and integrate linting with the testing process.