News for hackers. Stuff that matters.
Re:Type Systems (2004) [pdf] by adrianratnapala on Thursday August 31, 2017 @08:48PM (#15145294)
Checking your units is a basic sanity check that we teach students. But as we advance, physicists throw away that wisdom. Theorists especially have an annoying habit of doing this, because all the repetitions of "c", "hbar" etc "k_B" etc clutter up their equations.

But I preffer to keep that stuff in, even when doing pretty abstract stuff. Essentially because I wanted the type annotations. For example my favorite convention for the Minkowski metric is:

     [c^2  0   0   0]
     [ 0  -1   0   0]  
     [ 0   0  -1   0]
     [ 0   0   0  -1]

That way your vectors are just $(t, x, y, z)$, where $t$ has units of time and the rest of space. Your covectors do come out oddly as $(c^2t, -x, -y, -z)$, but that's what makes them covectors.
Re:Type Systems (2004) [pdf] by catnaroek on Thursday August 31, 2017 @09:40PM (#15145496)
Alternatively, you can work with x^0 = ct instead of x^0 = t.
Re:Type Systems (2004) [pdf] by adrianratnapala on Friday September 01, 2017 @12:01AM (#15146023)
Yes, and and that a much more common convention than the one I favour. They do it partly because it keeps the factors of c out of the visible equations -- but I like my those cs.

Another reason is that people instinctively feel that the elements of the coordinate tuple all have to have the same units. But that's not necessary when you have a metric around. Indeed keeping irregular units is helpful because it prevents you from accidentally forgetting to multiply through by the metric.

Re:Type Systems (2004) [pdf] by catnaroek on Friday September 01, 2017 @08:35AM (#15147822)
The c factor doesn't really go away. It just gets shifted into 4-vectors.