The humble bragging here smh

I like to use this style of italics for keywords. (That’s also what the Maple examples do.) My thinking is you see keywords so often that you recognize them by shape, not by reading the individual letters. And my theory is that the italic variant being a little harder to read helps my eyes skim over keywords, to focus more on words that I do need to read precisely, like variable names.

It does mean that I spend some time customizing my syntax highlighting theme to make it work the way I prefer. I’ve got examples set up on my blog. Although that’s not Maple - it’s a different font with cursive italics called Cartograph CF.

Now how to change the authentication failure message to read, “That procedure is not recommended”?

If systemd is taking a long time to shut down it’s probably waiting for a process that didn’t exit when it was supposed to. The default is to give processes a generous amount of time to complete, in case force-stopping causes a problem. Other init systems might be more aggressive about force-stopping. You can configure systemd to wait a shorter period of time by setting DefaultTimeoutStopSec

I read today that they’re using NixOS as a base for the new official distros. https://www.frandroid.com/os/gnu-linux/3062047_adieu-windows-securix-et-bureautix-le-linux-de-letat-aux-noms-dirreductibles-gaulois

Yes, that’s right. Everything except for Javascript’s usual dynamically checked types are erased.

I’ve already conceded on point 6 that types are not technically functions, and that I’m making an analogy. I think analogies are in the territory of opinions.

If you really want to argue technical correctness on point 5 then let me try to illustrate more clearly what a type parameter list is. First I’ll cite some sources to clarify terminology.

From the Wikipedia article on Generic programming (emphasis mine):

Generic programming is a style of computer programming in which algorithms are written in terms of data types to-be-specified-later that are then instantiated when needed for specific types provided as parameters.

That sentence cites the book “Programming Languages: An Active Learning Approach” by Kent D. Lee, which uses both the terms “type parameter” and “type argument”.

The Typescript Handbook page on Generics uses the terms “type parameter” and “type argument” many times. Here are some excerpts:

The type of generic functions is just like those of non-generic functions, with the type parameters listed first

Generic classes have a generic type parameter list in angle brackets

(I know we’ve been talking about functions, not classes. But both have type parameter lists, and it’s the same concept in both cases, so I think this line is relevant.)

When we use GenericIdentityFn, we now will also need to specify the corresponding type argument

I think we can clearly establish this terminology as it relates to one of the examples from my article:

function useState<S>(initialState: S): [S, (value: S) => void] { /* ... */ }
//               ╰┬╯
//       This is called a "type parameter"


//       This is called a "type argument"
//       ╭────────┴──────╮
useState<"open" | "closed">("open")

And you can see here how type arguments bind to type parameters in a way that is directly analogous to value arguments binding to value parameters:

function useState<S>(initialState: S): [S, (value: S) => void] { /* ... */ }
//               ╰┬╯ ╰────────┬─╯
//                │           └────────────────────────┐
//                └───────────────────────┐            │
//                               ─────────┴────        │
//        type argument binds to type parameter        │
//        ────────┬────                                │
//                │                             ───────┴───────
//                │     value argument binds to value parameter
//                │     ──────┬───────
//       ╭────────┴──────╮  ╭─┴──╮
useState<"open" | "closed">("open")

Your counterargument that “Generic functions can have the type specified” is just a different way of saying that a type argument binds to a type parameter.

Just in case the issue is with the use of the word “list”, a generic function can have multiple type parameters. Thus, a list. Here is another example from my article that demonstrates a list of type parameters:

//                  Two type parameters make a list
//                   ╭────────────┴─────────────╮           
function getOrDefault<Obj, Key extends keyof Obj>(
  obj: Obj,
  key: Key,
  def: NonNullable<Obj[Key]>
): NonNullable<Obj[Key]> {
  const value = obj[key]
  return value != null ? value : def
}

Note that getOrDefault has a type parameter list, and also has a value parameter list (or simply “paremeter list” if you prefer). Therefore it has two parameter lists.

Is the problem that when I use the words “parameter list”, to you that means specifically what I’ve been calling value parameters? And that’s not the same thing as type parameters? I think I’ve been very clear that I’m talking about two lists with different kinds of parameters. If that’s the issue then please make some allowance for the way I’m using terminology, and for the context in which I’m using it.

Maybe the issue is that functions can be called with only value arguments? Or that non-generic functions don’t have a type parameter list? I think I addressed this in my article. Type arguments are implicit, so they can be omitted from call sites. But the type parameter list exists in the generic function definition either way. A non-generic function doesn’t have a type parameter list - but I think an equally valid interpretation is that it has an empty type parameter list, which is expressed in the language by omitting the angle brackets.

Anyway, there’s no way you can say it isn’t true that some functions have two parameter lists when type parameter lists are a thing, and value parameter lists are a thing, and generic functions have both.

You can say that I’m overgeneralizing by saying that all functions have two parameter lists when actually only generic functions have two parameter lists. But I think you’d have to be pretty pedantic to say that makes my point wrong. I think it would be glossing over a nuance. If this is the point you disagree on then let me know.

Or you can say, as I do, that all functions do have two parameter lists, but for non-generic functions the type parameter list is empty, so we don’t write it. That’s my interpretation - or in other words, my opinion, which is not a matter of fact.

atrocious code block font

That’s author’s perogative I’m afraid. But if you have more specific feedback I’ll listen. For example, I’m thinking maybe you object to the handwritten italic font? I like to use that italic for keywords, because I recognize those by shape instead of by reading anyway. I think the use of italic helps me to skim over the keywords, and focuses my eyes on the non-keyword words that I need to pay more attention to. But then there’s always the question of where to draw the line. Like, should built-in types like number and string be italicized?

some of non-code annotations are misaligned (on phone) e. g. nullable string

Thanks for pointing this out! It turns out my monospace font doesn’t have glyphs for box drawing characters, and different web browsers were using different fallback fonts with very different glyph sizes. I configured another font for box drawing, and made sure that it uses glyphs with the same size as the other monospace glyphs. (No, I’m still not going to change the font. It’s my favorite, even if it makes me work a little harder.)

5 is about Generics with emphasis on React. functions don’t “take two parameter lists”. Generic functions can have the type specified.

Yes, functions do take two parameter lists! There is a type parameter list, and a value parameter list, clear as day! The type parameter list is how you specify the type. Just like the value parameter list is how you specify the arguments.

That section is intended to help readers to understand generics by explaining it in a way that they probably haven’t seen before. I wanted to relate the unfamiliar concept to a familiar one.

6 is wrong. types cannot “be functions”. You can do some type algebra on types.

Again, I disagree; and I provided multiple examples. I know that generic types are technically not functions. But I think they sort of do the same thing, which is to transform inputs into some output. In most languages in generics it’s not especially helpful to think this way. But in Typescript there is a lot of sophisticated stuff you can do with “utility” types, and I think it’s helpful to think of those as functions, but at the type level. Once again, this is about relating an unfamiliar concept to a familiar one.

But I did add a note to make it extra clear that types are technically not functions, even if they function like functions.

It’s intended to acquaint one with the language, and there is a lot do that. I don’t know what is the right way to write this kind of stuff; but my thinking was if I were learning I’d rather start with this kind of piece before getting into a book.

I use sops for NixOS, but those secrets aren’t accessed in CI. For actual CI I’ve used a combination of 2-4 above.

There are a few options:

1. Sops-nix or Agenix store secrets encrypted in the repo. Each local machine needs to be set up with a PGP or an SSH key to decrypt and encrypt as necessary. This is what I do with my NixOS configuration.

2. Manage secrets externally to repo code. I like to use direnv; sometimes I configure the checked-in .envrc file to source another file with secrets, that is not checked in.

3. Don’t use secrets locally. If secrets are things like deploy keys, and I want all deploys going through CI, then I don’t want secrets configured locally. Instead running a deploy script locally should be a dry run, which doesn’t need secrets.

4. Generate secrets at runtime. This is for cases where the project runs a cluster of services which need to authenticate with each other. For tests with locally running test services, authentication is confined to this isolated system. So secrets can be generated at test time, and written to env or config files that are not checked in.

Well, I’m in that Nix shop category. For example, I run tests for my OSS project using lots of Python versions. There’s no need to use a Github Actions matrix for that - Nix does it more efficiently. Doesn’t even require containers or VMs. And the whole setup runs exactly the same locally, on Github Actions, or on Garnix. (But Garnix has better Nix caching than other CI runners, so that’s what I use.)

How is that? Do you mean that week 1 can occur twice? Or is there another issue?

IIUC you can have days at the start and the end of the year that are all in week 1, and those are two different weeks. But one of those would be, let’s say, week 1 year 2025, and the other would be week 1 year 2026. According to the blog post if you have a week 1 at the end of a year the ISO year number for that week is the next year.

I showed the comic to my spouse because I do the same thing!

An application package is a compressed file with extra conventions. I think rpm packages are usually compressed with xz, but might use a different compression program. If I’m understanding what I’m reading correctly, rpm2cpio does some introspection to determine with extraction program to use. Maybe 7zip does the same thing, or maybe it works if the rpm was compressed with a compatible program.

In this context “Linux desktop” means not a server. Or in other words, a Linux computer with a graphical user interface. So it includes laptops. I can’t think of any software that would work on a desktop that wouldn’t work on laptop.

You’ll sometimes see the “desktop” clarification because technically the majority of Linux computers are headless servers. (Headless means no screen.)

It’s wonderful that we’ve moved on from a time when scrolling wms were kludgy experiments! It looks like now we have at least two solid options.

Plenty of tools are using the system keychain. There are good libraries that provide a generic interface to gnome-keyring or kwallet depending on what is running. When I was working with Node I used the keytar library for that purpose.

Edit: Oh, apparently there is a standard DBus API for keyrings. So you can use libsecret to interact with whatever keyring.

Gitignore settings don’t prevent checking in run configs if that’s what you want to do. Even if someone has the run configs pattern ignored, they’ll get those files with the repo clone. Gitignore only prevents adding files that aren’t already checked in - it’s a guard rail to prevent accidentally checking in so something you don’t want. You can override ignore settings to check in something new in with git add --force <filename>

Global gitignore is for idiosyncrasies of your computer, not for repo-specific stuff. For example if you use an editor that writes in-place backup or swap files (like .swp files with Vim’s default settings), or you’re on a Mac which writes .DS_Store files into every directory, you’re going to want those ignored for every repo. That way you can comfortably work with other people’s repos, even if they’re not configured for your particular setup.

The global ignore combines with gitignore files in each repo. So if it’s an ignore pattern that everyone working with that repo should have, put it in the repo.