2024-06-03 01:11:57 <lf94        > I'm trying really hard
2024-06-03 01:12:00 <lf94        > to turn this into the comments:
2024-06-03 01:12:02 <lf94        > enum: Color Green Red Yellow ;enum
2024-06-03 01:12:04 <lf94        > \ 0 constant Color.Green
2024-06-03 01:12:06 <lf94        > \ 1 constant Color.Red
2024-06-03 01:12:08 <lf94        > \ 2 constant Color.Yellow
2024-06-03 01:12:10 <lf94        > \ 3 constant Color.#
2024-06-03 01:41:56 <lf94        > I give up for today
2024-06-03 01:45:10 <zelgomer    > me too, bud
2024-06-03 01:45:12 <zelgomer    > #metoo
2024-06-03 08:44:34 <veltas      > lf94: You'll probably need to parse the words, put them in a buffer somewhere and EVALUATE them as 'constant <name>.<suffix>'
2024-06-03 08:49:44 <skvery      > A simple state machine is a case statement inside an infinite loop for any language.
2024-06-03 08:55:51 <veltas      > lf94: Conventionally you'd just call CONSTANT in your word and write out the prefixed names, no normal forth defining words add a prefix automatically
2024-06-03 08:56:16 <veltas      > Like FIELD: for instance just defines a word with the given name
2024-06-03 08:58:00 <veltas      > e.g.  : ENUM:  DUP CONSTANT 1+ ;   0  ENUM: COLOR.GREEN  ENUM: COLOR.RED  ENUM: COLOR.YELLOW  ENUM: COLOR.#  DROP
2024-06-03 08:59:47 <veltas      > Better names in context of the program would be:  GREEN RED YELLOW #COLORS
2024-06-03 09:00:03 <veltas      > Because, duh, GREEN is a color!
2024-06-03 09:09:14 <veltas      > KipIngram: Remember in chuck's POL paper he talks about routines vs subroutines
2024-06-03 09:09:26 <veltas      > A state machine is easy to get right with routines
2024-06-03 09:18:35 <veltas      > The second-best is a loop executing a variable's xt
2024-06-03 09:19:38 <veltas      > So e.g.  VARIABLE NEXT-STATE    : MACHINE  FIRST NEXT-STATE !  BEGIN NEXT-STATE @ EXECUTE AGAIN ;
2024-06-03 09:21:23 <veltas      > ['] FIRST  I should say
2024-06-03 09:21:33 <veltas      > The switch statement is the compromise that came from languages that don't have vectorised execution, but in Forth and even C that's not a problem, so you shouldn't be using a switch
2024-06-03 09:27:33 <veltas      > Could use defer for this too
2024-06-03 09:28:36 <veltas      > DEFER NEXT-STATE  : MACHINE  ['] FIRST IS NEXT-STATE  BEGIN NEXT-STATE AGAIN ;
2024-06-03 09:28:53 <veltas      > I don't really know if that's legit but I guess it would work on most forths
2024-06-03 09:55:41 <skvery      > For the way my brain works a switch statement is clearer.  One can always optimise later if you run into performance issues.  Coding a machine with:  StartMachine, addState is possible but for me way more complex.
2024-06-03 09:58:25 <skvery      > Coding a traffic light is a very nice way to learn how a state machine works.
2024-06-03 10:11:03 <veltas      > In the problem I think you can encode all the state in one field.  It might be easier to split them up anyway.
2024-06-03 10:11:22 <veltas      > I definitely would represent traffic with one field, with a bit assigned to north traffic and east traffic each
2024-06-03 10:12:20 <veltas      > And in the transition function you then check:  TRAFFIC EAST-TRAFFIC AND  and  TRAFFIC EAST-TRAFFIC =  in the relevant conditions
2024-06-03 10:13:27 <veltas      > A lot of these kinds of groups of boolean ideas work a lot easier if you just use bitwise logic, which the machine is good at anyway, and for which there's good support in all Forths
2024-06-03 10:14:17 <veltas      > Combining colors into one field means you can do a CASE statement on the combined colors
2024-06-03 10:14:51 <veltas      > e.g. 0 CONSTANT GREEN  1 CONSTANT RED  2 CONSTANT YELLOW  : COLORS ( c1 c2 -- c)  2 LSHIFT OR ;
2024-06-03 10:17:46 <veltas      > : TRANSITION ( c t -- c)  TUCK SWAP CASE  GREEN RED COLORS OF  EAST-TRAFFIC AND IF  CLEAR-NS  YELLOW OR  THEN ENDOF  .... etc
2024-06-03 10:36:17 <veltas      > lf94, skvery: https://termbin.com/3t29
2024-06-03 10:36:51 <veltas      > Well LIGHTS is wrong because there's no word to get next traffic state
2024-06-03 10:37:26 <veltas      > But you get idea
2024-06-03 10:41:48 <rendar      > wow this is an incredible work https://www.youtube.com/watch?v=0PclgBd6_Zs
2024-06-03 12:10:21 <veltas      > Don't write C++ like Rust, don't write C like C++, don't like Forth like C.
2024-06-03 12:10:31 <veltas      > don't write* Forht
2024-06-03 12:11:35 <olle        > Yea, don't write forth haha
2024-06-03 12:11:44 <olle        > jk
2024-06-03 12:12:05 <olle        > btw thanks for the tip earlier about simplifying some expressions by factoring out words, it worked
2024-06-03 12:12:14 <olle        > But I still have to implement dictionary chaining for my tiny DSL
2024-06-03 12:14:16 <veltas      > olle: Maybe WORDLIST / FIND will do?
2024-06-03 12:15:03 <olle        > Don't think so, some parts of the DSL evaluate to a string, other parts actually do the calculation
2024-06-03 12:15:22 <olle        > So + is sometimes and operation, other times returns a string
2024-06-03 12:15:25 <olle        > an*
2024-06-03 12:16:00 <veltas      > Allow parsing and do like S" maybe
2024-06-03 12:19:39 <olle        > Yea, but then I need to filter the inside of the string :) If it's raw SQL.
2024-06-03 12:19:55 <olle        > The benefit of DSL would be to allow only certain operation, so there's already a safe subset of SQL, PHP, etc
2024-06-03 12:20:28 <veltas      > I mean S" for parsing a string literal, not to encapsulate the expression
2024-06-03 12:21:38 <olle        > Aha
2024-06-03 12:23:06 <veltas      > >some parts of the DSL evaluate to a string, other parts actually do the calculation
2024-06-03 12:23:14 <veltas      > Does this mean more like constant folding?
2024-06-03 12:23:34 <veltas      > So + might literally do plus 'right now', and in some cases + is passed to SQL query?
2024-06-03 12:23:52 <olle        > Yep
2024-06-03 12:24:19 <veltas      > There's a syntax for this in Forth, [ and ] allow you to disable compilation temporarily
2024-06-03 12:24:21 <olle        > The part of the DSL that happens inside PHP evaluates + as an expression right now, but SQL evaluates to a string that's passed to the database.
2024-06-03 12:24:28 <olle        > Oh yea!
2024-06-03 12:24:52 <olle        > Hm hm
2024-06-03 12:25:29 <veltas      > So if you want to fold a constant you can do:   : MY-FIELD [ PLACE OFFSET + ] LITERAL ;
2024-06-03 12:25:50 <veltas      > You can reuse [ and ] words and define new behaviour like swapping the top wordlist
2024-06-03 12:25:59 <olle        > I can look into it, have to do "normal" work right now :(
2024-06-03 12:26:21 <veltas      > Terrible, having to actually be productive :P
2024-06-03 12:26:27 <olle        > I writing my own tiny Forth environment inside PHP btw, because that's the server host language.
2024-06-03 12:26:31 <olle        > It sucks :(
2024-06-03 12:26:57 <olle        > I did the `:` word yesterday, was not that hard.
2024-06-03 12:28:08 <veltas      > As fun as this is, if someone did this at work I would probably complain at them :P
2024-06-03 12:28:39 <olle        > Yea this project, to make a report generating DSL, is not green lit
2024-06-03 12:28:48 <veltas      > Like 'what the f*** did you do this for'
2024-06-03 12:29:02 <olle        > Hehehe
2024-06-03 12:29:19 <olle        > There is one big benefit tho - release a new report WITHOUT needed a new deployment of the software
2024-06-03 12:29:33 <olle        > And customize reports per customer. Dunno if needed/wanted yet.
2024-06-03 12:29:40 <veltas      > Probably cheaper ways to achieve that
2024-06-03 12:30:42 <veltas      > Also the cost you've associated with 'deployment' suggests process issues
2024-06-03 12:30:51 <olle        > Sever process issues, yes
2024-06-03 12:31:09 <olle        > Cheaper why, a DSL is a pretty simple thing to do
2024-06-03 12:31:23 <olle        > 
ACTION lunch

2024-06-03 12:33:55 <veltas      > If I saw the queries people actually use I would probably be able to define a simple subset that doesn't need full expressions
2024-06-03 12:37:31 <veltas      > Even if expressions are needed I would probably parse to AST in javascript and send something dumb back to PHP in a format it understands but won't be a security risk
2024-06-03 12:44:34 <olle        > The DSL scripts should be saved in db tho, to be able to have a per-customer system
2024-06-03 12:45:00 <olle        > And they might need to be changed by non-customers, by changing the db values
2024-06-03 12:46:45 <veltas      > Well the more sophisticated this sounds, the more it seems like you should use an existing language and just tighten the security another way
2024-06-03 12:47:16 <veltas      > Because the aim here is to allow queries with some security consideration, otherwise we'd just let them send raw queries
2024-06-03 12:47:35 <veltas      > There are probably better ways to control access
2024-06-03 12:51:22 <olle        > It's not only queries, it's also CSS, PHP, potentially JavaScript
2024-06-03 12:51:43 <olle        > And multiple export formats - HTML, CSV, something called SIE4 but that might be harder to do in the DSL
2024-06-03 15:48:18 <KipIngram   > This is quite good coverage if you're interest in quantum stuff:
2024-06-03 15:48:20 <KipIngram   > https://www.youtube.com/playlist?list=PLRN3HroZGu2mCtdalEmZAM2nr1xBWAtUn
2024-06-03 15:56:27 <olle        > Bah, modern physics
2024-06-03 15:56:46 <olle        > PBS is good too btw
2024-06-03 15:56:50 <olle        > And Crash Course
2024-06-03 16:37:34 <thrig       > could be worse, could be post-modern physics
2024-06-03 16:38:17 <olle        > Hehehe
2024-06-03 16:38:34 <olle        > It's already confusing enough eh
2024-06-03 16:59:59 <KipIngram   > Oh, PBS Spacetime in particular is great.  I've enjoyed it a lot since they replaced the guy who yelled all the time with the current guy - he's got such a pleasant, soothing voice - totally stark contrast to the old guy.
2024-06-03 17:06:32 <KipIngram   > They sometimes push the Many Worlds interpretation a bit harder than I care for.  But it's easy enough to overlook.
2024-06-03 17:06:53 <KipIngram   > I like Sabine Hossenfelder too.
2024-06-03 17:08:50 <KipIngram   > The link I posted above is a bit different, though - it's not what I'd call "popular science communication."  It's an actual univerisity caliber course.  In depth.
2024-06-03 17:09:15 <KipIngram   > I particularly like how well he's tying the development to the experimental history of the field.
2024-06-03 17:42:03 <rendar      > my Forth is at 1378 words!
2024-06-03 17:58:45 <KipIngram   > :-)
2024-06-03 17:59:14 <KipIngram   > How many primitives?
2024-06-03 18:31:11 <rendar      > 1378
2024-06-03 18:31:12 <rendar      > lol
2024-06-03 18:39:13 <zelgomer    > that sounds like a lot
2024-06-03 18:45:42 <veltas      > gforth 0.7.3 has 1840 findable words by default
2024-06-03 18:46:29 <veltas      > I think a few hundred is more normal for a simple forth, but 1378 isn't crazy depending on what it's for
2024-06-03 18:46:54 <zelgomer    > /why/
2024-06-03 18:47:12 <veltas      > Well what if it was a full OS
2024-06-03 18:47:17 <veltas      > That would be more reasonable then right
2024-06-03 18:47:46 <zelgomer    > but we're not talking about a full os, we're talking about core primitives
2024-06-03 18:48:05 <veltas      > I'm assuming they've just programmed them all in assembly or some high level language
2024-06-03 18:48:13 <veltas      > So because they're allergic to colon words
2024-06-03 18:48:49 <veltas      > Probably get better performance that way
2024-06-03 18:49:36 <zelgomer    > just counted mine, currently at 151 and could probably stand to lose a few
2024-06-03 18:50:21 <thrig       > ANS has about 350 words, and Common LISP about 950
2024-06-03 18:50:59 <veltas      > thrig: If you write an ANS Forth you usually need more words than the standard ones to factor it at all
2024-06-03 18:51:09 <veltas      > Sorry didn't mean to @ you
2024-06-03 19:08:43 <veltas      > Re the 'I2C is simple' claim https://stackoverflow.com/a/61623107
2024-06-03 19:09:38 <veltas      > There is no agreement on whether I2C addresses should be written shifted or unshifted
2024-06-03 19:10:25 <veltas      > 'Annoying detail no. 2134 of I2C-in-practice'
2024-06-03 19:13:12 <zelgomer    > ah yes, even i've encountered that one
2024-06-03 19:13:33 <zelgomer    > addresses in the data sheet are often half the address linux wants (or twice? i forget now)
2024-06-03 23:45:38 <lf94        > w o r d z
2024-06-03 23:55:25 <veltas      > : microsoft ;