2022-08-02 06:54:23 <lisbeths    > I posit that a dictionary should be implemented as a balanced binary search tree or a resizable array not as a linked list
2022-08-02 06:59:22 <dave0       > lisbeths: mine is a stack!
2022-08-02 06:59:58 <lisbeths    > I guess I don't know what a stack is
2022-08-02 07:01:46 <dave0       > um
2022-08-02 07:01:50 <dave0       > this is forth!
2022-08-02 07:31:07 <dave0       > lisbeths: if the dictionary is a stack, defining a new word would push the xt onto the dictionary stack
2022-08-02 07:32:23 <dave0       > lisbeths: to find a word, you would have to scan the dictionary stack from most recently pushed all the way to the start of the stack
2022-08-02 07:33:33 <dave0       > find would treat the dictionary stack like an array
2022-08-02 07:35:04 <dave0       > note it preserves forth's requirement that new words are found in preference to old words
2022-08-02 10:27:50 <KipIngram   > lisbeths: The only real constraint on how the dictionary is implemented is that *sometimes* you want to guarantee that if you re-define a word you will find the latest definition (at least within your search path).
2022-08-02 10:28:08 <KipIngram   > Beyond that, you can really use whatever data structure you want.
2022-08-02 10:28:42 <KipIngram   > It's also possible to use a hashing type algorithm; I suspect that GForth does exactly that, though it may be used "on top" of a linked list, as an "accelerator."
2022-08-02 10:28:50 <lisbeths    > I have lately fallen in love with ocaml/haskell style curried syntax
2022-08-02 10:29:04 <lisbeths    > forth syntax can be curried
2022-08-02 10:29:10 <KipIngram   > Falling in love is fun.
2022-08-02 10:29:10 <lispmacs[work]> there would there be other considerations, though...
2022-08-02 10:30:08 <lisbeths    > Imagine building a set of words that were turing complete such that you wouldn't need to input numbers ever again
2022-08-02 10:30:23 <lispmacs[work]> like, you typically want the location of the executable code itself not changing, so you don't have to update all the definitions that point to executable code
2022-08-02 10:30:57 <lisbeths    > It is fast to rewrite an array compared to walking a long linked list over and over
2022-08-02 10:30:59 <KipIngram   > Yeah, but that's not really to do with the *dictionary* is it?  The dictionary is just how you find the cfa's.
2022-08-02 10:31:05 <KipIngram   > And yes - the CFAs shouldn't change.
2022-08-02 10:31:23 <lisbeths    > do you know what a combinator is?
2022-08-02 10:31:40 <KipIngram   > Not formally, no.  Maybe in some vague fuzzy way.
2022-08-02 10:31:48 <lisbeths    > swap is a combinator
2022-08-02 10:31:56 <KipIngram   > Ok.
2022-08-02 10:32:02 <KipIngram   > A "rearranger"?
2022-08-02 10:32:03 <lispmacs[work]> Also, since the dictionary is (usually) only used during compilation, one has to ask if it makes sense to have to add a bunch of extra code to manage balancing or resizing the dictionary
2022-08-02 10:32:09 <KipIngram   > Something that makes permutations?
2022-08-02 10:32:24 <lispmacs[work]> something to consider for a Forth for MCs
2022-08-02 10:32:29 <KipIngram   > lispmacs[work]: Indeed - totally agree.
2022-08-02 10:32:38 <KipIngram   > Most of the time compile time isn't a terribly important thing.
2022-08-02 10:32:42 <lisbeths    > combinators either act postfixedly infixddly or prefixedly. they take arguments and leave behind some transformed output in their place in the sytnax tree
2022-08-02 10:32:48 <KipIngram   > I guess it could be in some situations.
2022-08-02 10:33:07 <KipIngram   > My impression is that when Chuck compiled his chip design system he frequently rebuilt the entire thing.
2022-08-02 10:33:14 <lisbeths    > when most people talk about combinators they are talking about pure combinators meaning immutable stateless words
2022-08-02 10:33:28 <KipIngram   > He did a bunch of stuff to accelerate compile time, but it was probably a unique need.
2022-08-02 10:33:42 <lisbeths    > chuck is interested in low power chips right now so he needs a purely reversible forth this I can prove
2022-08-02 10:33:44 <thrig       > pretty sure ycombinators are where hackers go to talk about news
2022-08-02 10:33:59 <lisbeths    > the y combinator is part of the lambda calculus
2022-08-02 10:34:03 <KipIngram   > Yes, I know about the theoretical power considerations of reversibility.
2022-08-02 10:34:10 <KipIngram   > That's fascinating stuff.
2022-08-02 10:34:23 <lisbeths    > the stirling and kestrel combinators when combined with swap nip tuck etc are turing complete
2022-08-02 10:34:43 <KipIngram   > Ok.  You're way out of my knowledge domain here; I've heard all these terms before, but I'm not able to actually "work with them" effectively.
2022-08-02 10:34:53 <lisbeths    > combinators are one of the few computer systems which may gain the benefit of being able to be purelu reversible
2022-08-02 10:35:12 <lisbeths    > thats because I deal with high level abstract programming not low level microcontroller programming
2022-08-02 10:35:18 <KipIngram   > My training is really in hardware.  I've dipped into software quite a lot over the years, but I'm a "hack" on that front.
2022-08-02 10:35:37 <lisbeths    > In cloud computing the hardware is kept from us
2022-08-02 10:35:43 <KipIngram   > Right.
2022-08-02 10:35:56 <lisbeths    > the closest we can get is we can rent a virtual machine with an fpga
2022-08-02 10:36:15 <KipIngram   > The thing is, though, current power consumption levels are nowhere NEAR that theoretical limit.
2022-08-02 10:36:23 <KipIngram   > We're not even close, at least not in the mainstream.
2022-08-02 10:36:27 <lisbeths    > I study in part highly portable systems like forth and lisp. Systemd with few parts that a human can memorize
2022-08-02 10:36:56 <KipIngram   > Yeah - the reason I like Forth is because it's simpe enough for one person to deploy from the ground up, all by themselves.
2022-08-02 10:37:00 <dlowe       > That doesn't describe systemd at all ;)
2022-08-02 10:37:08 <lisbeths    > systems
2022-08-02 10:37:19 <KipIngram   > Oh - typo.
2022-08-02 10:37:48 <lisbeths    > a language like forth should in theory be easier to audit for aerospace naval and military industrial purposes
2022-08-02 10:38:00 <thrig       > systemd might blow that minimal forth memory requirement, a little
2022-08-02 10:38:28 <lisbeths    > Forth doesnt just produce small filesizes it produces source code files with less text
2022-08-02 10:38:31 <KipIngram   > That doesn't mean reversible computing isn't worth studying - it's always worth understanding the limits of what's possible.
2022-08-02 10:39:08 <lisbeths    > In the future we may be forced to use purely reversible forths because the quantum computers may get so good and cheap
2022-08-02 10:40:18 <dlowe       > I suspect when quantum computers become cheap we will be wrong on every count about what it'll be like
2022-08-02 10:40:51 <dlowe       > except that everything ever encrypted over previous decades will be laid bare.  That should be a reckoning.
2022-08-02 10:41:26 <lisbeths    > we can store an environment in a single pointer in the stack. It can be a balanced binary search tree with all my program data
2022-08-02 10:41:53 <lisbeths    > I can design forth words that take this pointer as their one argument and leave this pointer as their one output
2022-08-02 10:42:17 <lisbeths    > This is one of many ways to encode combinators using forth
2022-08-02 10:43:11 <lisbeths    > combinators allow for very high level abstractions without using much system resources, but they are slow to program with
2022-08-02 10:43:49 <KipIngram   > Sure - quantum computing has to be reversible.  I have my doubts about quantum computers "displacing" regular computing; I think it will find its place in particular applications, where it makes sense.
2022-08-02 10:44:10 <KipIngram   > I suppose it it turns out to be TRULY possible to make quantum computers dirt cheap that might change, but it seems unlikely to me.
2022-08-02 10:44:28 <KipIngram   > Dangerous thing to make predictions about.
2022-08-02 10:45:21 <KipIngram   > That sounds interesting, but it's just so far from the type of thing I play with that I don't think I really have anything worth saying about it.
2022-08-02 10:46:09 <KipIngram   > I clicked at some point to the idea that you CAN regard Forth as a functional language if you want to, just by saying that each word takes "the state of the system" as its input and produces a new system state as its output.
2022-08-02 10:46:28 <KipIngram   > That just has quite a lot of inefficiency in it, though, so it's a very different type of system.
2022-08-02 10:48:11 <KipIngram   > I think the biggest beneficiaries of quantum computing will be materials science, chemical engineering, drug design, and so on.  Stuff that profits fron excellent modeling of molecular systems.
2022-08-02 10:48:34 <KipIngram   > And the payoffs in those areas might turn out to be enormous.
2022-08-02 10:49:08 <KipIngram   > I don't expect to ever use a quantum computer to watch a movie or check my email, though.