The callback way is more generic and prevents consing when you don't need to store the resulting node list. You may want to simply print something or maybe modify the node in-place, for example.
> In C++, this could be achieved if all local variables were in fact std::shared_ptr captured by value.
So, in C++ you do actually get to pick what happens and there are plenty of options but for our purposes here all we want is a (mutable) reference capture.
However, experienced C++ programmers would never do this because C++ is all foot guns all the time, so you can express what you meant and it'll blow up and cause chaos because now our reference outlives the thing referred to. Oops.
In Rust we can write what we meant, but instead of the program exploding at runtime the compiler will politely point out that this can't work and why.
And so armed with the knowledge from that, we can (in Rust or with C++ although it's harder to spell in C++) write something that'll actually work.
We could move the captured variable. In Rust we just use the keyword `move`, now the captured variable is gone, moved inside the closure, and so as with the Tcl the same variable (the one moved into this closure) is used each time the closure is called, and if we make another closure that's got a different captured variable.
But we could do the "shared reference" trick, that type is spelled Rc in Rust.
The Wiki[1] is one of the primary "hang out" spots, although it's a bit different from usual online communication. But there's a lot of mutual commenting, small articles and utilities etc. on there.
"The European OpenACS and TCL/Tk conference will be in Bologna/Italy/Europe on July 10 & 11 2025." - this is crazy. Seems there are still people using OpenACS in 2025.
They did have a recent language update after awhile. That may have triggered some folks to look into it again. There is sometimes a HN effect where an initial post triggers some interest amongst enough users to get us new posts for a few weeks and then things tend to die off again. I've seen this with a lot of the more obscure languages like APL.
It would be cool to have a Tcl revival though (although I don't see it happening - I'm not in the community though so hopefully someone more informed can post). The language itself seems more capable than most give it credit for. I'm more of a Python fan myself, but can appreciate Tcl after reading through a book on it and writing a few scripts.
For those who are not aware, Tcl is actually part of standard Python distribution through TKinter.
There are many things Tcl has built in that are quite amazing, like a robust virtual filesystem support, reflective channels, and less known these days Starpacks (stand alone runtime) that bundle sources with the binary.
I am current working on bringing back kitcreator for an AI project that uses Tcl as a scripting environment over llama.cpp.
Roy Keene is the original author, and has done some really clever stuff here, like encrypting the VFS appended to the executable. I added compression to this. It provides some manner of obfuscating sources.
And actually, I am also working on using tohil to compile a static Python and load it as a Tcl extension, with the goal to have standalone Python applications bundled with their sources and completely loadable from within the VFS. This will provide a means to bundle TKinter with a “frozen” Python app.
Thank you and GP for the recommendation, just bought the book, seems pretty good! Now I wonder whether it's a good idea to replace my shell with tclsh... seems a lot more sane than bash/zsh.
Bash is pretty good for really small scripts. Anything bigger and I have reached for Perl, Python, or Tcl in the past ... depending on what IT had installed on the server.
I last got help on the IRC channel (bridged to Slack, because I don't know IRC).
In the most recent big version update there was what I'd consider a breaking change regarding text encoding handling, but it was possible to go back to the old behaviour with an additional parameter .
I introduced it into some of our release tooling in the mid-2000s. Easy to integrate, easy to understand, unsurprisingly good string/text handling, expect was very useful, and it’s not going to be used by anyone else, so no worries about version conflicts.
It ran successfully largely unchanged for around a decade.
it's a language that's trivial to implement because it's well designed and simple, it embeds very nicely, and it's fantastic for use as a debug shell and to implement guis. it's a great technician's language, if you work with technically-minded people who aren't necessarily programmers, it's a great way to hand them deep interactive power without the footguns of a forth.
I would say it's cleverly designed. Well designed? Hmm, would a well designed language have such a basic flaw as comments that can only be used in very specific places?
I understand where they came from here: the Scheme-like obsession with purity (the enshrined Endekalogue, now Dodekalogue) didn't mesh very well with traditional comment.
Yeah, Tcl has its design warts, but I don't think it has that many remaining that can't be fixed via metaprogramming. Even the popular Python manages to frustrate me with its idiotic statement/expression divide (they doubled down by making match() a statement...) and constant need to convert between generators/iterables and lists.
Thing is that R6RS Scheme (or R7RS-large if it comes out one day) is basically a better Tcl if you only consider scripting and don't need the event loop. If Tcl had played its cards right, it'd have competed with fish/rc/nushell/powershell instead, it was really ready to be a better shell well before any other.
Despite that, I find article interesting. It shows that Tcl can truely be multiparadigm programming language.
Myself, I've implemented pattern matching [1] over algebraic-type-like values and used that here and there.
[1] https://wiki.tcl-lang.org/page/Algebraic+Types
reply