00:00:00 --- log: started forth/18.12.29 00:07:46 --- quit: pierpal (Quit: Poof) 00:08:07 --- join: pierpal (~pierpal@95.239.223.85) joined #forth 00:21:00 --- quit: smokeink (Remote host closed the connection) 00:39:46 --- join: [1]MrMobius (~default@c-73-134-82-217.hsd1.va.comcast.net) joined #forth 00:41:00 --- join: [2]MrMobius (~default@c-73-134-82-217.hsd1.va.comcast.net) joined #forth 00:42:19 --- join: [3]MrMobius (~default@c-73-134-82-217.hsd1.va.comcast.net) joined #forth 00:42:30 --- quit: MrMobius (Ping timeout: 245 seconds) 00:44:41 --- quit: [1]MrMobius (Ping timeout: 272 seconds) 00:44:44 --- join: MrMobius (~default@c-73-134-82-217.hsd1.va.comcast.net) joined #forth 00:45:22 --- quit: [2]MrMobius (Ping timeout: 246 seconds) 00:47:47 --- quit: [3]MrMobius (Ping timeout: 268 seconds) 01:06:34 --- join: proteusguy (~proteus-g@49.230.15.216) joined #forth 01:06:35 --- mode: ChanServ set +v proteusguy 01:11:07 --- quit: gravicappa (Ping timeout: 250 seconds) 01:34:04 --- quit: pierpal (Read error: Connection reset by peer) 01:38:53 --- join: pierpal (~pierpal@95.239.223.85) joined #forth 01:45:38 Heck yeah, just realized I have 13K bytes after HERE to play with, and can write around 2 or 3 16K Forth programs for my operating system 01:45:59 Would not be possible to get more than 1K of space after HERE if I used TI's operating system 01:47:02 tabemann: Cool stuff. Let me know how it goes. 01:47:09 Time to implement anything to my heart's content 01:47:30 implement an x86_54 compatability layer. 01:47:36 Could do with one of those for the z80. 01:47:38 I just tried the example Forth program on Wikipedia https://en.wikipedia.org/wiki/Forth_language#A_complete_RC4_cipher_program and it works perfect 01:47:44 How many bytes does that take? 01:47:56 You mean x86-64? 01:48:06 Uh, that too. 01:48:34 Pull requests always welcome :). I'll try to implement more of the standard for now 01:48:38 Get CATCH and THROW to work 01:49:37 What do you mean by a compatibility layer? 01:50:42 Nothing, don't worry. Just making a stupid joke 01:55:59 My entire OS fits in 10861 bytes, heh. 01:57:03 Well I should focus on getting my OS to be usable outside of an emulator, there's no real terminal implemented yet but the way input is structured I can add a new input "device" 01:57:09 And everything should work fine. 01:59:26 i found this useful for fm/mod https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/divmodnote.pdf 02:03:07 --- quit: ashirase (Ping timeout: 250 seconds) 02:03:54 I still don't know if I'd consider a Forth an operating system 02:04:15 Unless you had a Forth kernel and forth userspace 02:04:25 WilhelmVonWeiner: It isn't really, imo. 02:04:39 Operating systems are bloat 02:04:58 --- join: gravicappa (~gravicapp@85.26.234.223) joined #forth 02:05:07 Operating systems aren't bloat. 02:05:58 They are so. 02:07:26 I might consider a multiprogammed, garbage-collecting Forth an OS? 02:08:30 --- join: ashirase (~ashirase@modemcable098.166-22-96.mc.videotron.ca) joined #forth 02:08:59 multitasking. 02:10:07 It's better to think of them as Forth /systems/ 02:10:24 I think an OS has to fundamentally be able to not let the user shoot themselves in the foot easily. 02:10:42 That's giving UNIX far too much credit 02:10:46 For instance it's difficult to crash TI's operating system because there's only so much you can do 02:10:59 Except if you write asm code and flash it 02:11:11 Haha, true. 02:13:58 I think the difference between a normal OS and Forth has to be the kernelspace/userspace distinction. Where in a given OS you usually interface with hardware through the kernel, and with Forth you're just in the kernel 02:14:27 Ah, right. I've been thinking of how to implement a safe/limited version of Forth 02:14:38 Or some other language, but it may be counterproductive. 02:14:56 The more layers you add the worse performance is 02:15:04 In that case I think you would write a kernel kernel in Forth 02:17:33 Userspace could be considered blocks 02:18:26 Huh looks like there's a port to interface with an MD5 device on my calculator 02:18:27 http://wikiti.brandonw.net/index.php?title=83Plus:Ports:18 02:21:45 --- quit: proteusguy (Ping timeout: 250 seconds) 02:22:13 --- quit: gravicappa (Ping timeout: 272 seconds) 02:22:36 --- quit: pierpal (Ping timeout: 250 seconds) 02:24:18 --- join: proteusguy (~proteus-g@49.230.15.216) joined #forth 02:24:18 --- mode: ChanServ set +v proteusguy 02:25:15 --- join: pierpal (~pierpal@95.239.223.85) joined #forth 02:28:45 --- quit: pierpal (Read error: Connection reset by peer) 02:29:06 --- join: pierpal (~pierpal@95.239.223.85) joined #forth 02:42:36 blocks are just memory 03:13:24 --- join: gravicappa (~gravicapp@85.26.234.223) joined #forth 03:34:51 --- join: dddddd (~dddddd@unaffiliated/dddddd) joined #forth 03:52:28 --- join: rain1 (~My_user_n@unaffiliated/rain1) joined #forth 03:52:30 --- quit: rain2 (Ping timeout: 245 seconds) 04:22:32 --- quit: proteusguy (Ping timeout: 272 seconds) 04:24:22 --- join: proteusguy (~proteus-g@49.230.15.216) joined #forth 04:24:22 --- mode: ChanServ set +v proteusguy 05:03:23 --- join: smokeink (~smokeink@42-200-116-35.static.imsbiz.com) joined #forth 05:35:30 proteusguy: verilog 05:36:17 what's the difference between COMPILE and POSTPONE? 05:36:24 i must be something important 05:40:13 --- quit: dys (Ping timeout: 250 seconds) 05:41:35 compile always compiles the following word. postpone is aware of whether the word is immediate and has different behavior 05:41:49 postpone usually does what you actually want 05:43:18 i must be daft, i think i still didn't internalize interpretation vs compilation vs execution semantics 05:43:26 like "POSTPONE IF" - what you probably wanted there wasn't to compile the immediate word IF into the target word, but actually execute IF so that it compiles a jump into the target word 05:44:59 ok let me explore this better 05:45:15 it is confusing at first and i think it's because there isn't (or at least i'm not aware of) better language to express it 05:45:36 both POSTPONE and COMPILE are only used in IMMEDIATE words or words that are called by IMMEDATE words, right? 05:46:00 they don't have to be, but yes that's where you'll generally find them most useful 05:46:06 ok 05:46:31 is their behavior the same for non-IMMEDIATE words? 05:48:30 --- join: Kumool (~Khwerz@adsl-64-237-238-189.prtc.net) joined #forth 05:48:58 when executed from a non-imm? yes 05:50:39 no, i mean when taking non-IMMEDIATE words as argument 05:51:38 COMPILE is deprecated 05:51:44 It's been replaced by POSTPONE 05:52:30 POSTPONE: "Skip leading space delimiters. Parse name delimited by a space. Find name. Append the compilation semantics of name to the current definition. An ambiguous condition exists if name is not found. " 05:53:19 In the end I think implementations should at least conform to CORE standard words, otherwise portable programs would be a nightmare, but extension words can be implementation defined 06:02:49 corecode, think of it in levels of deferred action 06:03:23 at level 3, the word is executed when the parser encounters it, even while in compile mode (IMMEDIATE words) 06:03:51 at level 2, the word is compiled into a target word when the parser encounters it 06:04:12 and at level 1, the word has been compiled and will be executed when the target word is executed 06:04:29 COMPILE parses the next word and always treats it like level 2 06:04:45 POSTPONE parses the next word and bumps it down one level 06:05:32 sorry, still don't get it 06:05:54 i should find an implementation of POSTPONE 06:06:10 hold on, let me switch to my laptop where it'll be easier for me to type examples 06:06:21 --- quit: gravicappa (Read error: Connection reset by peer) 06:07:19 --- join: gravicappa (~gravicapp@85.26.233.92) joined #forth 06:08:38 i guess the first question is: do you understand how IF works? 06:10:47 let me try to code it up to see 06:21:40 oh man, there is also ' 06:21:54 yup 06:21:58 and ['] 06:23:18 imo this is the hardest part of understanding forth - the line between interpretation and compilation has been very blurred, and a lot of the word names don't help make that any more palatable 06:26:31 https://gist.github.com/62d40cad720ae0c38362cca9ae4ed429 06:27:36 close, except you want a SWAP before ! 06:27:51 ah yes 06:28:11 also this assumes 0BRANCH wants an address and not an offset. but that's beside the point 06:28:38 but in this example, you understand that when IF is parsed, it compiles 0BRANCH into the target word 06:30:19 now what if you wanted to write a word that extends the behavior of IF? like, maybe you wanted to count the number of times IF conditions are used in your program for some reason. you might write : MY-IF COUNT @ 1+ COUNT ! IF ; 06:30:23 IMMEDIATE 06:30:29 but that doesn't work, because IF is immediate, right? 06:30:55 which means it'll try to compile 0BRANCH and push HERE and compile a 0 into the definition of MY-IF 06:31:16 and you can't use : MY-IF COUNT @ 1+ COUNT ! COMPILE IF ; IMMEDIATE 06:31:40 because that would just compile the word IF into the word being compile when you use MY-IF, which isn't what you want, either 06:32:05 you basically want to negate the IMMEDIATE off of IF, so that it executes when your MY-IF executes 06:32:36 POSTPONE is "smart" and knows how to do that when you give it an immediate word. : MY-IF COUNT @ 1+ COUNT ! POSTPONE IF ; IMMEDIATE 06:33:29 --- quit: Kumool (Ping timeout: 240 seconds) 06:35:39 --- quit: smokeink (Ping timeout: 244 seconds) 06:35:48 --- join: Kumool (~Khwerz@adsl-64-237-234-130.prtc.net) joined #forth 06:38:20 hmm 06:38:24 yea that makes sense 06:38:29 but how do i implement that 06:38:58 i compile the call to if into my-if 06:39:07 right 06:39:17 so it is literally just 06:39:41 : POSTPONE FIND , ; ? 06:39:48 IMMEDIATE ofc 06:40:02 not quite, because it has to do something different for non-immediate words 06:40:15 if the word isn't immediate, then it needs to behave like COMPILE 06:40:27 now i need to figure out what compile would do :) 06:40:39 Hmm, how do I implement things like : where I have to read forward? 06:40:56 WORD 06:41:19 or sometimes PARSE 06:43:35 hi john_cephalopoda 06:44:07 i have : : CREATE ] ;CODE doCOLON: ... END-CODE 06:44:07 06:44:47 oh yeah, i forgot CREATE parses. i did it differently in mine 06:49:28 --- join: dys (~dys@tmo-103-128.customers.d1-online.com) joined #forth 07:00:03 how? 07:02:58 Hi corecode 07:05:23 Wait, you're using the data space at HERE for the dictionary? 07:09:30 yea, where else? 07:09:49 is there a dedicated code space? 07:11:51 corecode: noticed on your gist profile that you're an Emacs user too :) 07:12:38 corecode: : POSTPONE FIND , ; ? doesn't quite cut it, you have to manipulate the pointer returned by FIND to point to the code field address 07:12:56 This is my POSTPONE 07:13:01 : POSTPONE WORD FIND >CFA , ; IMMEDIATE 07:13:44 Where WORD reads the next word in the input stream, FIND finds the dictionary entry, >CFA converts that pointer to a code field pointer (by skipping the word name and header 07:18:34 my FIND would always return the code field address 07:18:40 why would it ever return something else? 07:22:49 --- quit: Zarutian (Read error: Connection reset by peer) 07:23:06 --- join: Zarutian (~zarutian@173-133-17-89.fiber.hringdu.is) joined #forth 07:26:15 Well how do you check if a word is immediate? 07:26:36 Must be in the headers 07:27:33 aha good point 07:27:38 thanks 07:27:39 And so if you FIND the word and get its code field address it's not gonna be easy finding the length and flags byte 07:28:01 Yeah it's not really standard but this is where I think it's better not to be 07:28:07 that explains tabemann's layout with the length byte after the name 07:28:38 Right, you can change the design to be like that if you want 07:29:11 --- join: kumul (~kumool@adsl-64-237-234-130.prtc.net) joined #forth 07:29:22 But I prefer the length byte right after the link pointer so we can just stop comparing if the lengths don't match 07:30:26 I thought the usual setup was (where | denotes cell boundary): | prev word link | xt | name length&flags | name ... |, where prev word link points to the name length&flags cell of the previous word in dictionary 07:31:20 but that is with dictionaries that split name and code parts 07:31:27 It shouldn't matter as far as Forth is concerned 07:31:45 No wonder why the standard wants FIND to return a code field pointer 07:32:15 what's xt? 07:32:45 in the case I described, each dict entry points to the name of the previous to speed up comparisons iirc 07:33:09 xt stands for eXecution Token, usually just code address 07:33:54 Ah, I see. 07:34:06 I don't do it like that but it's equivalent. 07:34:20 --- join: learning_ (~learning@108.47.9.90) joined #forth 07:34:32 Of course, a linked list is a natural way to do things 07:34:43 so with xt it is an ITC? 07:34:49 --- quit: learning_ (Remote host closed the connection) 07:35:01 in FIG Forth, IIRC, the header is just | prev word link | name length&flags | name ... | code 07:35:32 Yep that's mine 07:35:35 corecode: not nescisarly, it could just be an address to control-flow-branch to 07:36:14 siraben: I even seen one where the name is bracketed by its length 07:36:36 One shouldn't rely on layout quirks if one wants to write portable code 07:36:57 Zarutian: how so? 07:37:43 re portability: well I target a version of a dual stack machine from the "Stack Machines The New Wave" book 07:38:12 siraben: prev word link | name length&flags | name ... | name length | code ... | 07:38:44 Why would you do that? 07:39:13 Interesting nonetheless 07:40:07 because given an xt which lets say just points to the code section you can decrement it by one cell, get the length, decrement further that amount and get a pointer to the name length&flag cell 07:40:58 so getting the name of the word doesnt involve looking through the whole dictionary for it 07:41:47 Cool, that makes sense 07:42:06 Makes decompiling faster 07:44:21 In my forth system i get to control what words are exposed to the user 07:44:24 saw it in a forth that can quine itself out, so to speak. (That it is, print it entire state as an macro assembler file) 07:44:53 A lot of things aren't written in forth style for speed and making it hidden 07:45:14 Well that should be easy if you can read itself 07:45:18 You can read yourself 07:45:46 So I guess that makes some sort of kernel/user space separation in my system 07:48:27 siraben: say, does the TI you are targetting have any kind of kernel/user state bit or even simplistic MemoryM(anagement|apping)Unit that gets disabled at interrupts (and where its mmio control registers can be blocked out when the MMU is active)? 07:50:12 There is some memory protection 07:50:21 Eg the first 4000h bytes 07:50:29 Can't be written to 07:50:48 I can swap out some RAM pages with flash pages 07:51:26 And there's only some flash pages (privileged) that can write to protected pets 07:51:31 Ports 07:51:46 Some of which control flash memory management 07:52:09 There's also different modes for interrupts 07:52:26 I can make it not respond to the user pressing ON for instance 07:56:33 hmm... just wondering if 'process' seperation could be done on it basically 07:57:40 There is a project that implements a Unix like OS onto it 07:57:48 With multitasking, processes and so on 07:58:26 https://knightos.org/ 07:58:40 I haven't read their code for those parts but I should 08:01:20 --- quit: nighty- (Quit: Disappears in a puff of smoke) 08:09:47 corecode> how? 08:10:01 in mine i combined : and create so there's only one word, : 08:10:32 you use : to declare a new word, and then you say what it is. like : x var ; : page 4096 const ; 08:10:52 if you want to compile stuff, you use "does" 08:11:04 : 2dup does dup dup ; 08:11:25 and higher-order words just use the word "does" again 08:12:03 : initvar does , does @ ; : z 123 initvar ; 08:14:43 zy]x[yz: Huh, doesn't 2DUP do ( a b -- a b a b )? 08:14:59 sorry, dup2 08:15:21 whatever. all of the forth words are stupid and awful 08:15:31 What would you propose instead? 08:15:49 i don't know 08:16:05 i cleaned up : and create, though! 08:18:20 i remember someone once had an example of a generic stack-shuffling word, where you could write something like #1212 and it would compile something that would do it. i'm starting to have the opinion that that's way better than dup, swap, nip, tuck, etc. 08:23:24 Hmm, not sure. 08:23:34 First off all the efficiency would probably suffer. 08:23:51 Also, how would you drop elements with that? 08:26:13 And finally this syntax stops working when you go more than 9 stack elements down because #12 could mean ( ...3 2 1 -- ...3 2 1 12 ) or ( 2 1 -- 2 1 1 2 ) 08:28:28 yeah i don't remember the exact syntax. whatever it is, it can be designed to work 08:29:13 zy]x[yz: Retro has reorder, you pass it two strings and it shuffles the stack like so. 08:29:21 maybe that's what I'm thinking of 08:29:26 i.e. 'abcd 'dbca reorder 08:29:34 would reverse the order of the top four items 08:30:31 and i don't buy that performance would suffer. at worst it would compile the words you'd write manually, anyway. at best it would actually generate machine code to do it 08:34:35 At best your stack would be 16 registers and `reorder` owuld generate an optimised sequence of register exchanges 08:34:44 16 deep, circular 08:36:08 Do computers even have got so many registers? 08:36:36 yes 08:36:39 Yeah 08:37:01 but I'm imagining a stack processor designed for it 08:37:26 unfortunately a lot of architectures don't put their registers in addressable memory, though, so they can't overlap with the stack like he's describing 08:37:30 x86 has ~7 easily usable 32-bit registers (EAX, EBX, ECX, EDX, ESI, EDI, ESP and EBP. 08:37:59 --- quit: gravicappa (Read error: Connection reset by peer) 08:38:03 amd64 doubled those 08:38:36 (also that's 8, not 7) 08:38:57 Right, 0-indexed. 08:39:44 --- join: gravicappa (~gravicapp@85.26.165.169) joined #forth 08:39:47 "It's seven! 0, 1, 2, 3, 4, 5, 6 and 7!" said the programmer :Ã¾ 08:41:35 --- join: PoppaVic (~PoppaVic@unaffiliated/poppavic) joined #forth 08:42:04 do they still have different functions or did they ever get more orthaganol? 08:42:07 (sp) 08:43:13 what, the amd64 registers? all of the classic registers retain their old usage. the new r8-r15 registers are just general purpose 08:44:19 MrMobius: All those x86 registers I mentioned are multi-purpose, except for ESP. 08:44:29 ah ok 08:44:54 well, some of them have special meaning with some instructions 08:45:08 i remember thinking wtf when i read about 8086 registers. seems like it wasnt very multi purpose back then 08:45:15 When you want to move a register to the address of ESP, like MOV [ESP], EAX, 08:45:17 ecx, esi, and edi, in particular 08:45:19 then it won't work. 08:45:27 thats what i mean 08:45:38 you cant do memory offsets and indexing type stuff with all registers? 08:45:53 what? yes you can 08:46:02 POP and PUSH will increment or decrement ESP. CALL will also use ESP. 08:46:43 can you do MOV [x],EAX where x is any register? 08:46:51 yes 08:47:02 MrMobius: It works for every register, except or ESP. 08:47:08 *for ESP 08:47:13 no, you can also use esp 08:47:32 neato 08:48:11 a simple implementation of swap is to xchg %ebx, (%esp) (where %ebx is your TOS register) 08:48:26 zy]x[yz: Huh, I can't find a way to do that. 08:48:37 http://ref.x86asm.net/geek.html#modrm_byte_32_64 08:49:15 I can "MOV ESP, EAX", but having MOV [ESP], EAX" doesn't seem to work - at least not with MOV. 08:50:13 Of course I can do "PUSH EAX", which will basically do that, directly using MOV isn't possible though - as far as I know. 08:50:43 it has to be encoded using a SIB byte 08:51:27 http://ref.x86asm.net/geek.html#sib_byte_32_64 08:51:43 yep 08:51:48 The place where it should say "ESP" says "none" 08:51:54 that's the scaled index 08:52:18 Uh, [ESP] 08:52:37 Same for the "32-bit SIB Byte" 08:52:51 ESP would be the base, the scaled index would be 100 (none) 08:53:17 and ss would be 0 of course 08:53:37 Oh, huh, looks like I have overlooked that. Thanks. 08:54:40 x86 is a pretty crazy architecture. 08:54:49 Yes 08:54:53 yeah it's complicated. it's a huge pain to write an assembler for 08:56:22 i know because i've failed twice now 08:57:13 --- quit: dave0 (Quit: dave's not here) 08:57:58 zy]x[yz, what was your assembler for? 08:59:04 --- join: rdrop-exit (~markwilli@112.201.166.158) joined #forth 09:03:01 my forth 09:03:21 i wanted it to be self-hosting 09:04:24 zy]x[yz: I am working on an assembler right now. It actually builds runnable OS images already. 09:05:28 I think the most irksome thing is that most systems require the use of goddamned Linkers. After you manage to write the code in ANY language, you get to fight with the fucking linkers. 09:06:16 PoppaVic: Heh, I also had that problem. And then I decided to just use no linker at all. 09:07:00 I just wrote a x86 compiler that creates multiboot-executables for x86. 09:07:05 https://github.com/jmf/impexus 09:07:09 Yeah, I can remember years of no-linker-at-all. 09:07:26 And all that in less than 500 lines of code. 09:08:07 (Although some parts are still missing - ADD, MUL, DIV and SUB arer not implemented yet) 09:08:37 I suppose there is a way to write the proper file-magic header for linux, but it's something I don't care to get involved in. And, my targets for this would be samd21/51 or little avr's anyway.. Linkers, sheesh 09:08:41 the LOC metric doesn't make sense because you use way too much whitespace in your definitions john_cephalopoda 09:09:14 you can't use too much whitespace, but LOC was never a valid metric anyway 09:09:26 You definitely can use too much whitespace 09:10:10 In Forth, words might be a valid metric with well-factored code 09:12:31 Heh, code metrics are often relatively useless. 09:12:52 With enough dedication you can cheat every metric. 09:12:59 Well sure 09:13:39 I can try to find patterns and pull them together into words, even when that makes the code worse readibility-wise. 09:13:41 but you get a general idea of the maintainability, simplicity, and factored-ness of well-written code 09:13:44 Well, you'd like "codespace used vs dataspace-used", possibly "word-counts", and then you get into the speeds/bottlenecks. Lines? Not so much. 09:14:28 Well, the point of my line count measure was to show that my code is relatively small for what it can do already. 09:14:41 LOC is just a barstool comment - and yer buying the next round. 09:14:49 Otoh in Forth line count depends a lot on who writes it. 09:15:17 I put a lot of whitespace and linebreaks into my code to make clear where blocks are. 09:15:20 john_cephalopoda: http://forth.org/forth_style.html 09:15:37 yes, I've seen a LOT of forth over the years that seems to embrace the C "infinite line length" mentality.. I continue to write my vertical code 09:19:05 PoppaVic: I set an 80 character line width marker in my text editor and I'm keeping my code to the left of it. 09:19:41 There are not many things more annoying that editing code in nano that is longer than the terminal width. 09:20:41 I'm happy with 64x16 09:21:25 80x25 is the resolution of VGA text mode on x86. 09:22:03 Most terminals were 80x24 09:22:16 Also I believe that gopher recommends 80 characters width max. 09:22:38 john_cephalopoda: : 4C, DUP 2C, 16 RSHIFT 2C, ; 09:24:52 john_cephalopoda: yes, I've always used something like 75 or 78 as a wrap-point.. Unlike the numpties, I can dropkick into pico or mc-editor, or joe - I use GUI editors for ease on the eyes only. 09:25:02 --- part: kumul left #forth 09:25:21 I like 72. I think that's the recommended wrap for the OpenBSD mailing lists 09:26:06 72 would work, my setting is just predicated on readability and often a WORD will end up fitting out to 75 or 78, is all 09:26:27 PoppaVic: I use geany as my main editor and nano only if I don't intend to stay there for long or if there is no X available on the box. 09:26:38 I always think that to be the anthropic principle at heart PoppaVic 09:26:46 whatever that principle is but more general 09:27:00 with forth, I always liked th 16x64 ENFORCEMENT in blockfile[s], but I tend to think of it as 16x64 ;-) 09:27:03 Does the code just fit, or did we make it just fit? 09:27:21 Yes, geany is awfully comfortable. 09:33:07 I really like the feel of coding on 64x16 09:34:04 well, like I said above: I like to think of it as 16 chars of 64 lines. (and editor that could flip like that would be nice ;-) 09:34:34 lol 09:35:08 It's similar to the restraints I like on C code - I always get argued down - but, if I can't see the entire func in an 80x24 terminal, YER WRONG. - and less is mobetta 09:36:08 I think of it as coding on an index card 09:36:12 rdrop-exit: It would be even better if the editor had two panes: code and shadow - both 16 wide ;-) 09:36:45 :-) 09:37:14 THe majority of the space should be comments, or blank.. Or multiple SIMPLE transformers/translators/transliterators - trannies ;-) 09:37:18 My code is fairly horizontal 09:37:44 yeah, I don't do Horiz for control-flow 99% of the time. 09:38:17 hey guys 09:38:26 Hey tabemann! 09:38:56 Hi tabemann :) 09:39:13 * tabemann tends to opt for 80 columns 09:39:48 john_cephalopoda: can you share some kind of assembler code you used with your WIP assembler? 09:42:05 WilhelmVonWeiner: The init.fth file contains assembler code. 09:42:07 the core of my words tend to be only a few lines, but I often have a number of lines afterwards of ELSE foobar THEN ELSE foobar THEN etc with ELSE, foobar, and THEN all on their own lines 09:42:54 WilhelmVonWeiner: The actual code is between "HERE ASSEMBLY_START !" and "HERE ASSEMBLY_END !" 09:43:29 I very rarely nest ifs, or use else 09:44:14 I can't remember the last time I used else actually 09:45:46 * tabemann normally uses IF ELSE THEN rather than IF THEN 09:46:21 really 09:47:15 That's suprising 09:48:12 why? in a lot of my code ELSE is used to clean up from the non-execution of IF 09:48:14 When I feel tempted to use ELSE I usually refactor away the temptation 09:49:12 e.g. : FOOBAR DUP 0> IF . ELSE DROP THEN ; 09:49:39 and that's a simple example 09:51:04 I could code it as : FOOBAR DUP 0> IF DUP . THEN DROP ; 09:51:14 but that only works because it's a simple example 09:53:43 looks like yer ?DUP was fubar ;-) 09:53:55 LOL 09:54:15 you think you're funny dont ya 09:54:34 well, it's better than thinking yer a WIT - and being half-right ;-) 09:55:47 the problem with ?DUP is that it doesn't work for arbitrary conditions 09:56:10 I'm actually bemused above: in every case the examples are 'sane' in that they COULD be one-liners.. Mostly, I see code where they put too much code on the one-line to follow with a mental stack-comment 09:58:03 tabemann: I disagree. ?dup is dead-simple.. zero? nope. not-zero? dup it. YMMV 09:58:50 often one-liners are more cryptic, whereas breaking things up into multiple lines give them extra structure to help wrapping one's brain around them 09:59:05 PoppaVic: that assumes your condition is 0<> 10:00:49 Yes, it does. So, your case would require yer own spin on ?DUP 10:00:59 You can use factoring to break things up 10:01:59 until you have nice one-liners 10:05:28 You're FOOBAR seems a little contrived, do you have a more realistic example? 10:05:47 my code is rarely one-lines, more like two to five-liners - however I have many two liners where the first line is just : FOOBAR ( stack -- comment ) and the second line is the code, because the stack comment and the code are two long to fit on the same line together 10:06:50 I have a lot of those too 10:07:18 john_cephalopoda: the gopher spec recommends a max of 70, leaving 10 for client to use for displaying type information, etc 10:09:04 this one isn't exactly your idea of good forth code, but here is an example of an else clause being used to clean up: 10:09:05 : (FILTER) ( a-addr1 count1 a-addr2 xt -- count2 ) 10:09:05 0 10:09:06 [: 3 PICK 0> ;] 10:09:06 [: 4 ROLL 4 ROLL FETCH-ADVANCE DUP 5 PICK HIDE-6-BELOW-2 EXECUTE 10:09:06 SHOW-6-BELOW-1 10:09:07 [: ( a2 xt c2 a1 c1 x ) 5 ROLL TUCK ! CELL+ 3 ROLL 1+ 4 ROLL SWAP ;] 10:09:09 [: DROP 4 ROLL 4 ROLL 4 ROLL ;] CHOOSE ;] WHILE-LOOP 10:09:11 ROT ROT 2DROP ROT ROT 2DROP ; 10:09:28 eek! :) 10:09:37 CHOOSE is IF ELSE THEN 10:09:51 WHILE-LOOP is BEGIN WHILE REPEAT 10:10:22 PICK and ROLL 10:10:52 so how do you keep your UP? 10:11:14 UP? 10:11:28 is there something to get the address of the pointer that HERE will dereference? 10:11:42 something like @HERE or (HERE)? 10:11:43 oh 10:12:00 in attoforth it's not user accessible 10:13:22 oh and, if you haven't figured it out [: ;] form quotations 10:14:14 --- quit: gravicappa (Ping timeout: 246 seconds) 10:14:50 I'd say that's fairly cryptic 10:15:06 HIDE-6-BELOW-2 stuffs six values below the top two values on the stack into the return stack 10:15:46 SHOW-6-BELOW-1 takes six values from the return stack and places them below the top value on the parameter stack 10:15:53 What are all these stack gymnastics trying to accomplish? 10:16:42 that there is code for doing a FILTER against an array-list (there is another word, named FILTER, which calls this and places a-addr2 before count2 for convenience's sake) 10:17:14 the reason for hiding stuff on the return stack is so that the inner xt can operate as if it could see the outside world on the parameter stack 10:18:32 also, the gymnastics are because it is advancing a pointer pointing to the source array and decrementing a source count, while also advancing a pointer pointing to the destination array and incrementing a destination count 10:19:13 You must be taking quite a performance hit from all these gymnastics, is there no simpler way to accomplish what you need? 10:20:19 the thing is that it's got to keep track of five different things at once by its very nature 10:20:58 What do you need it for? 10:21:16 I mean, sure, I could stick these values into globals or task-locals or shit, but that wouldn't be the right way to do it 10:21:52 this words? it's a library word for implementing array filtering with an xt 10:21:56 *word 10:22:26 it's inspired by filter in Haskell and many other languages 10:23:28 The cost of all this abstraction must be very high, couldn't you strip it away and solve your problem in a more direct way? 10:23:53 https://forthworks.com/share/6f890e8fe946b99250d23cf7bc62a9ce is how I do this in my Forth 10:25:12 well the person could always just use COUNT-LOOP with a single destination index on the stack and read from one array on the stack (via PICK) or in a global and write into another array on the stack (via PICK) or in a global with the provided index 10:28:14 Just using PICK or ROLL would make my brain explode :) 10:29:11 Different strokes 10:31:14 I don't know how to program in Forth without both PICK and ROLL - I can see not using PICK, but that would require using ROLL, and ROLL probably has worse performance than PICK 10:33:37 --- join: gravicappa (~gravicapp@83.149.21.200) joined #forth 10:34:13 I can't fathom a scenario where I would need either PICK or ROLL. 10:37:27 you never need to keep track of more than three values on the stack whilst looping? 10:38:25 Keep track? No 10:40:49 --- quit: gravicappa (Remote host closed the connection) 10:42:13 more than 3 shufflers and you probably need to refactor your code 10:42:30 3-deep shufflers* 10:44:26 I never have multiple data stack operators back to back, as in your "drop 4 roll 4 roll 4 roll" example 10:45:09 --- quit: Kumool (Ping timeout: 240 seconds) 10:45:17 each stack operator is followed by work 10:45:38 that's because I'm restoring the stack to its original configuration at the start of the loop 10:46:52 could anybody link me to a single page html ANS FORTH spec? 10:47:08 i keep searching in the txt, and it is a bit tedious 10:47:20 tabemann: one forth optimization trick I read was to find which stack config needed the least manipulation inside the loop. One then put words to reconfig the stack for that right before the loop. 10:47:29 --- join: Monev (~Khwerz@adsl-64-237-239-208.prtc.net) joined #forth 10:50:01 in this case, all parts of the stack are being used at different points 10:51:23 corecode: I don't remember ever seeing any copies of the html copy as a single page 10:53:16 okay, I need to go - I'll bbl 10:55:01 Take care 10:57:01 It's almost 3am, I need to crash. Keep on Forthin' guys. 10:57:25 --- quit: rdrop-exit (Quit: Lost terminal) 11:05:38 --- quit: Monev (Ping timeout: 250 seconds) 11:05:57 corecode: Why do you need it on a single page? 11:06:25 corecode: I am using the pdf personally. Searching for words in it works more or less. 11:07:50 --- join: Monev (~Khwerz@adsl-64-237-238-109.prtc.net) joined #forth 11:08:50 corecode: "see Here"": Here dp @ ; OK", "see allot". 11:09:00 (GForth) 11:09:58 --- quit: Monev (Client Quit) 11:10:12 --- join: Kumool (~Khwerz@adsl-64-237-238-109.prtc.net) joined #forth 11:10:40 shouldn't that pdf be on forth.org, right at the top? 11:11:28 https://forthworks.com/forth/standards/DPANS/DPANS.pdf 11:11:36 https://forthworks.com/forth/standards/DPANS/DPANS.txt 11:11:37