programming-language

Currently, this is defined only for arrays, but it seems like we should have

normalize(x::Number) = x / abs(x)

(I just came across this when trying to implement this algorithm for random unitary matrices, which is basically Q * Diagonal(normalize.(diag(R)), but I discovered that normalize didn't work.)

Description

Say we have a custom error defined like this in an interface:

interface MyInterface {
    error MyCustomError();
}

The compiler mistakenly considers MyInterface.MyCustomError to be of type function() pure instead of type error. The latter is what should be used b

Currently they use 20.04

The builder will need to be updated and then the version of said builder will need to be updated in .cirrus.yml.

This should get a changelog entry.

Ubuntu 22.04 will be available on or around April 21, 2022.

You can already do ucm version, but ucm --version is a pretty standard thing to be able to do, and I find that I keep typing it.

It's also expected by default for certain tools like help2man (though that one can be configured to look for ucm version).

Some quarks produce files/folders outside of it's own folder, some quarks need the system to run a certain command after installation/uninstallation or do other things to finish the setup, and while this can be done with a method in a class of the package or documentation, it would be really great to have this automated in the quark file itself.

What I propose is that the quark dictionary contain

i.e. src/one/two/three.js and src/one/two/three.erl.

We will need to track Erlang module names to ensure there are no clashes. For example src/one/themodule.erl and src/two/themodule.erl in one project is an error.