add support for Eigen::Transform

[WanZhiQiu-ac]

No description provided.

[WanZhiQiu-ac]

The unit test failed at expect(buffer == "[-0.3492063492063493,0.9206349206349207,0.17460317460317448,0,0.0317460317460318,-0.17460317460317443,0.9841269841269841,0,0.9365079365079365,0.3492063492063492,0.03174603174603163,0,1,2,3,1]"); on gcc-x86. I suspect it might be due to a precision issue. Feel free to make any necessary edits.

[WanZhiQiu-ac]

Now I am certain it's a precision issue. I checked other unit tests, and none of them have such high-precision testing, so I removed the high-precision unit test. Perhaps this issue should be addressed in a separate PR? Or maybe it doesn't need to be fixed?

[stephenberry]

I don't think this is a Glaze issue, because it is robustly tested for floating point round-tripping (every possible float is tested and random doubles). This is probably a precision issue with the way floating point assembly is generated for Eigen. Floating point math operations with Eigen aren't guaranteed to give the exact same result across compilers.

[stephenberry]

Thanks for adding this support! Do you think it is ready to be merged?

[WanZhiQiu-ac]

I don't think this is a Glaze issue, because it is robustly tested for floating point round-tripping (every possible float is tested and random doubles). This is probably a precision issue with the way floating point assembly is generated for Eigen. Floating point math operations with Eigen aren't guaranteed to give the exact same result across compilers.

Totally agree

[WanZhiQiu-ac]

And yes I think it is ready to be merged

[stephenberry]

Sweet! Merged it in.