00:00:00 --- log: started forth/08.04.16 00:53:21 --- join: tane (n=tane@p57903B4F.dip.t-dialin.net) joined #forth 01:01:32 --- join: Maki (n=Maki@adsl-224-84.eunet.yu) joined #forth 01:04:39 --- quit: tane`weq (Read error: 110 (Connection timed out)) 01:45:23 --- quit: ecraven (Read error: 110 (Connection timed out)) 03:07:18 --- quit: Malfermi1aKodo (Read error: 104 (Connection reset by peer)) 03:11:43 --- quit: nighty^ ("Disappears in a puff of smoke") 03:12:39 --- join: MalfermitaKodo (n=kansu@xdsl-78-34-147-245.netcologne.de) joined #forth 03:29:06 --- quit: ramkrsna (Remote closed the connection) 03:35:41 --- join: ramkrsna (n=ramkrsna@unaffiliated/ramkrsna) joined #forth 03:43:04 --- join: ecraven (n=nex@140.78.42.119) joined #forth 05:23:55 --- join: nighty^ (n=nighty@p5025-adsau17honb13-acca.tokyo.ocn.ne.jp) joined #forth 06:31:26 --- quit: ramkrsna (Remote closed the connection) 07:10:02 --- quit: ecraven ("bbl") 08:22:32 --- join: tane`weq (n=tane@p579039DE.dip.t-dialin.net) joined #forth 08:32:55 --- quit: tane (Read error: 110 (Connection timed out)) 08:53:18 --- join: luinfana (n=tony@c-69-253-227-21.hsd1.nj.comcast.net) joined #forth 09:02:49 back with another quick question if anyone's lurking around... 09:03:50 hi 09:03:56 o hai 09:04:13 ugh 09:04:29 was wondering if someone could explain how this little function works 09:04:34 haha, gotta love lolspeek 09:04:35 what's your query? 09:04:42 it goes like this 09:04:52 : find.divisors 09:04:59 0 # divisors ! 09:05:20 dup 1 do dup i mod 0= if 1 # divisors +! i swap then loop drop; 09:05:37 I assume that's #divisors, not # divisors 09:05:41 so what that does is find the divisors of a given number, if #divisors is already declared as a variable of course 09:05:45 right, sorry 09:06:07 significant difference 09:06:12 so I understand what it does and all, just not quite how it works...what's the purpose of #divisors +! ? 09:06:30 +! increments. 09:06:58 okay, but why is it incrementing the divisors variable...? 09:07:12 so 1 #divisors +! adds 1 to #divisors 09:07:17 right 09:07:26 it's counting them. 09:07:36 but shouldn't #divisors just be holding the numbers it finds? 09:08:01 I guess I'm confused as to where the division actually happens...there's MOD but that should only return the remainder, not the result of division 09:08:14 it's a variable. It holds one number. 09:08:26 OK 09:08:50 MOD is the division operator here. If there's no remainder, it's a divisor. 09:09:01 --- join: JasonWoof (n=JasonWoo@unaffiliated/herkamire) joined #forth 09:09:01 --- mode: ChanServ set +o JasonWoof 09:09:21 the list of divisors winds up on the stack. The count is in #divisors. 09:09:36 understood...but how is it actually dividing, say 6, by 1, 2, 3, 4, 5...etc 09:09:47 with mod. 09:10:01 really? I didn't think mod worked that way 09:10:06 I thought it would have to be /mod 09:10:07 or something 09:10:13 how do you think it works? 09:10:44 given two numbers, divides them (quietly) and returns the remainder only 09:11:06 right. And that's what you want here. The remainder. 09:11:06 right? maybe I'm completely off 09:11:20 right, I definitely want it to check if the remainder is 0 or not 09:11:29 so... 09:11:41 maybe it'd help if I showed you the whole program...it's only a few lines 09:12:05 one sec 09:13:55 http://retroforth.com/paste/?id=522 09:14:25 it does a search through a given range of numbers, looking for "perfect numbers" 09:15:21 ok. 09:15:47 I need to do a line-by-line analysis of this. So the first function "integers" just increments the top of the stack on every loop, right? 09:16:12 and then find.divisors finds the divisors of the current number, and puts them on the stack 09:16:15 no. 09:16:18 oh ok 09:16:36 enlighten me 09:17:07 bench test integers at the prompt. See what it does. See what it leaves on the stack. Write a stack diagram for it. 09:18:17 yeah, I did that before...you can give it a number like 10, and it counts up to 10 (prints 1 2 3 4 5 6 7 9 10) and puts all of the numbers on the stack 09:18:29 haha I don't know how to count, forgot the 8 09:18:30 lol 09:18:50 --- join: forther (n=forther@c-67-180-150-67.hsd1.ca.comcast.net) joined #forth 09:18:51 ok. So how do you get from that, to saying it increments the top of the stack each time through? 09:19:34 ummmmm 09:19:40 not sure 09:19:45 hi 09:20:04 guess I was looking at my comments from earlier, and wasn't thinking about it 09:20:27 Analysis requires thought. 09:20:33 indeed, indeed. 09:20:51 hi forther 09:21:09 so now the stack has this string of consecutive numbers onit 09:22:22 with the highest one on the top... 09:22:34 where in that code is 'integers' used? 09:22:49 did you see the code I pasted? 09:22:58 http://retroforth.com/paste/?id=522 09:23:00 yes. Where is it used? 09:23:13 oh, at the top of the program 09:23:25 before find.divisors and all that 09:23:25 read my question. I see where it is defined. But where is it used? 09:23:37 ahhhh OK, gotcha 09:24:01 --- join: JasonWoo1 (n=JasonWoo@c-65-96-165-155.hsd1.ma.comcast.net) joined #forth 09:24:07 --- quit: JasonWoof (Read error: 104 (Connection reset by peer)) 09:24:08 it.......isn't? 09:24:42 correct 09:24:54 so...why would it be there in the first place if nothing calls it? 09:25:04 hell of a good question. 09:25:06 that's so weird, I didn't notice that before 09:25:09 haha 09:25:35 okay, so I'm going to delete that 09:25:45 and just worry about find.divisors 09:26:47 what does the <> signify? 09:26:57 down at #divisors @ 1 <> if 09:27:21 look it up in the Standard. 09:27:23 didn't find it in the Forth manuals I've been reading 09:27:46 :/ can you just tell me, or is it that complicated? 09:28:09 you need to have the references available, and know how to use them 09:28:38 yeah, you're probably right 09:28:46 http://quartus.net/dpans 09:29:14 thanks! 09:29:44 section F is an index. 09:30:52 ahhhhh not equal 09:30:58 OK 09:31:00 makes sense 09:33:28 then I think the only other thing I'm not getting is the next line, #divisors @ 2 - for + next...and I'm not seeing "for" in the Standard index 09:34:08 for/next are nonstandard. What Forth is this written for? 09:34:18 that's what I thought 09:34:20 gforth 09:34:43 then look at the gforth manual. 09:34:49 so I assume gforth includes those commands, right...they're in the words list 09:35:00 yep 09:35:30 where on earth did that code come from? 09:36:16 hah, a couple of different places...mostly written by crc 09:36:23 it could do with some factoring. 09:36:37 --- nick: JasonWoo1 -> JasonWoof 09:36:44 --- mode: ChanServ set +o JasonWoof 09:37:07 in 2nd grade? 09:37:47 I was under the impression that crc was a good programmer 09:38:02 really, it's that bad? 09:38:04 works for me 09:38:13 I just need to understand it like it's the back of my hand 09:38:21 maybe he was in a hurry. 09:38:28 he did mention that it could be done "more efficiently" 09:38:39 luinfana, is this homework or something? 09:39:29 it's not efficiency, it's the stack usage 09:39:41 no, it's a large project I'm doing on Forth in general. This is just one little part of it. 09:40:08 what's the project? 09:41:56 discuss the history, implementations, syntax of the language and give specific examples in code...perfect numbers seems to be a pretty standard thing 09:42:37 So homework. Ok. That code you're chewing on is badly factored, and does what it does in the clumsiest way. 09:43:09 ah, I still wouldn't call it homework... :) 09:43:21 it's for school, though. 09:43:24 hm, and you said crc was a good programmer? 09:43:28 for college ^_^ 09:43:50 and I'm probably not supposed to ask that sort of stuff here, I gather 09:45:02 look, I can't account for who wrote it and why, I can only tell you what's true about it. 09:45:15 FWIW, crc specifically indicated that his solution was not well-factored: Apr 13 21:28:52 for the record, a working (though not well-factored) solution to luinfana's problem is at http://retroforth.com/paste/?id=519 09:45:39 oh, of course...I mentioned he said it could be improved 09:45:56 woh, I guess there are some large perfect numbers, like 191561942608236107294793378084303638130997321548169216 09:46:11 yeah, the fifth one is already more than 33 million 09:46:22 this only needs to find the first four and I'm happy 09:47:35 was in crc's idea to keep all divisors on the stack? 09:47:57 so, unless someone wants to rehash the entire problem in better-working code (and I'm not asking you to), I think I'm sticking with this 09:48:25 no, I provided the find.divisors part...found it somewhere else 09:48:50 he changed it to have that #divisors +! part though 09:48:51 you're making a key error, trying to give an overview of forth by starting with a badly-written example of it. 09:49:08 mmm 09:49:11 that's not good 09:49:39 so what's the alternative to using the stack? 09:49:57 the find.divisors leaving an untold and possibly very large number of stack items behind is _very_ _bady_ 09:50:02 bad 09:50:17 you can't not use the stack 09:50:20 everything uses the stack 09:50:23 so a partial solution would be if it "cleaned up after itself?" 09:50:41 words should take 0 09:50:42 right, it leaves thousands of items on the stack 09:51:00 take 0-3 arguments and return 0-2 arguments 09:51:03 as a general rule 09:51:20 words should never need to be cleaned up after 09:51:35 I...don't know enough to understand what that 0-2 and 0-3 mean... 09:51:38 they should provide you with the number or effect you asked for 09:51:52 sure, that would definitely be best 09:52:03 it would make a lot more sense if it gave you the sum of the divisors, not a slew of stack items 09:52:27 : perfect? dup divisor-sum = ; 09:52:27 it does eventually sum the divisors...it has to to determine whether the number is perfect or not 09:53:20 so why collect them first? Sum them as you go. 09:53:39 true... 09:54:39 seriously though, if I'm being a pain or I'm not supposed to ask these kind of questions here, this'll do...I'll leave you in peace 09:54:50 --- join: ygrek (i=user@gateway/tor/x-0de99eeeb2285aa6) joined #forth 09:55:07 understood that this is a pretty terrible solution 09:55:28 variable v-divisor-sum 09:55:42 ok 09:56:16 and how to sum them up? 09:56:29 : divisor-sum ( x -- y ) 0 v-divisor-sum ! dup 1 do dup i mod 0 = if i v-divisor-sum +! then loop drop v-divisor-sum @ ; 09:56:53 something like that 09:57:03 oh wow, OK 09:57:05 thanks 09:57:08 np 09:57:13 hope it works, I didn't test it 09:57:19 yeah let me try 09:58:07 : print-perfects 2000 2 do i perfect? if i . then loop ; 09:58:51 Try this: http://retroforth.com/paste/?id=523 09:58:54 or better yet, do it without the variable 09:59:36 OK 09:59:39 will do 09:59:55 : divisor-sum ( x -- y ) 0 over 1 do over i mod 0 = if i + then loop nip ; 10:00:11 idiomatically, I might even say i mod 0= i and + but I don't expect you to follow that at this stage 10:00:13 If you don't have nip, it's just: : nip swap drop ; 10:00:40 Quartus`: nice :) 10:01:56 : divisor-sum ( x -- y ) 0 over 1 do over i mod 0= i and + loop nip ; 10:02:22 that's a confusing optomization 10:02:27 very nice 10:02:37 not to me. 10:02:39 I like it, but for intro material, stick to the one above with the "if" in it 10:04:10 I'm on gforth, so I've got nip 10:04:50 Quartus' pasebin solution works great 10:05:24 so work on understanding that one. 10:05:56 divisor-sum ( x -- y ) 0 over 1 do over i mod 0= i and + loop nip ; looks interesting 10:06:41 it's the same, but with an alternate idiom instead of if/then 10:07:22 luinfana: if you're interested in learning about the trick, it's cool, but for demo/intro purposes, stick to the one with the "if" (Quartus`'s paste looks great) 10:07:48 OK, yeah I like that solution 10:07:58 thanks alot, this is much more compact 10:08:09 it's better forth. 10:08:14 :) 10:08:15 Is it really a trick? I would call it one of the things that makes Forth unique. Kind of like your nth-execute: "trick" that I am so fond of. :) 10:08:55 sure, it's tricky 10:09:18 as for nth-execute... many languages have jump tables 10:10:39 I have always felt that "true" being defined as -1 was done on purpose, so that you could do things exactly like the "AND +" idiom in Quartus's solution. 10:11:01 sure 10:11:09 very cool 10:11:17 and so you don't need a "logical and" 10:11:54 right, the same one serves both needs 10:14:06 malyn: how'd you get nth-execute from me? from herkforth or fronds or did I describe it here? 10:15:58 I saw it here a while back. I didn't understand what was happening at first, but after doing the research (and testing things out) I understood both nth-execute: and quite a bit more about how the Forth VM operates. 10:16:42 how one particular sort does, anyway 10:17:03 Every now and then you drop interesting tidbits like that into #forth and I am always happy to stash them away in my notebook. :) 10:18:18 cool 10:20:23 Quartus`: why do you do "over 1" before the do loop? 10:23:23 look up DO, see what it requires 10:26:48 has something to do with how many times it loops, right? 10:26:58 look it up. 10:27:05 like the 2000 2 do at the bottom 10:27:12 yeah, I was just looking at it 10:28:39 bbiaw 10:31:41 Any preferences for naming of deferred words? 10:34:45 I would like to take this opportunity to cast one vote against the suffix "-I-am-a-deferred-word", on the grounds that it would be stupid. 10:35:06 madwork: "emit" 10:35:19 Yes, clearly you should replaces dashes with dots in that suffix. 10:35:37 gnomon: hehe :) 10:42:05 --- quit: JasonWoof ("Leaving.") 10:44:00 --- join: JasonWoof (n=JasonWoo@unaffiliated/herkamire) joined #forth 10:44:00 --- mode: ChanServ set +o JasonWoof 10:53:39 I think I've got Quartus` solution down, except for the over 1 ?do part...looked at the Standard definition as he suggested but I 10:53:46 I'm still not getting it 10:53:50 ... 10:55:21 I'm just too much of a Forth noob, I guess 10:56:53 luinfana, in which paste is ?do being used? 11:00:01 Also, could you explain your current understanding of what ?do actually does? If there's a misunderstanding there, that's the best place to start. 11:02:02 luinfana: in understanding confusing sequences of words, (especially for beginners) I recommend writing a stack comment after each word 11:02:21 (just copy/paste a copy of the code, replace all spaces with line breaks, and add the comments in) 11:07:41 yeah, funny you should say that, that's exactly what I'm doing right now 11:08:32 gnomon: current understanding of ?do? non-existent. I guess. I think I understand regular old do, though 11:09:36 Ah, if you're experimenting with it then you're 90% of the way there. Carry on. You'll have it in minutes. 11:10:04 :) well that sounds promising 11:17:23 oh...sorry, if you haven't figured it out already I'm referring to http://retroforth.com/paste/?id=523 11:26:08 Hm. 11:26:26 Am I missing something or does Forth not have structures? 11:37:42 Forth doesn't have structures in the same way that C doesn't have strings. 11:41:05 Hm. 11:41:41 alright...I still don't get it. If I type "0" in my Forth, followed by "over 1", I get a stack with three items: 0, some huge number (memory address?), and 1. Why is that? 11:42:05 my guess is this should NOT happen 11:42:11 luanfina, try putting 10 and 15 on the stack before executing over. 11:43:05 right right, I tried it with numbers on the stack and I see what it does, I just don't understand why using it the way the program dictates gives me a weird number 11:44:06 10 and 15 gives me 10 15 10 as it should 11:44:20 The way the program dictates..? The program expects there to be a number on the stack already, doesn't it? 11:44:39 it just slaps 0 onto it! 11:44:50 luinfana: read the documentation for OVER 11:44:57 There's some number N on the stack; divisor-sum starts by putting a 0 on the stack and then executing over to bring a copy of N back up to the top of the stack so that it can start working on it. 11:45:06 luinfana: it takes two operands. Weird things happen if you only give it one 11:45:17 some forths will crash in this case. I'm suprised gforth didn't 11:45:21 Weird things like nose demons. 11:45:37 ok...that sounds right 11:45:40 haha 11:45:48 nah, gforth is some pretty stable stuff 11:45:50 1 2 3 .s 11:45:52 1 2 3 OK 11:45:55 over .s 11:45:59 1 2 3 2 OK 11:46:05 yup 11:46:05 JasonWoof, the fact that gforth doesn't crash just means that it's not checking for stack underflow, no? It's just pulling from the memory region beyond the end of the stack. Could be anything there. 11:46:15 got it got it 11:46:28 gnomon: right, but iirc gforth generally does check for stack underflow 11:47:06 Ah, thus your puzzlement. Gotcha. 11:47:31 gnomon: I just popped open gforth, and typed "drop" which resulted in: *the terminal*:1: Stack underflow 11:47:54 right, generally looks like this: 11:47:57 Backtrace: 11:47:57 . 1714303587 11:47:57 *the terminal*:8: Stack underflow 11:50:31 ok, so gforth allows read-access to below the bottom of the stack 11:50:37 dup and over both work when the stack is empty 11:50:51 well... "work" in the sense that they put *something* on the stack 11:52:02 so when this program starts for the very first time, it's checking 2, right? And it incements that number all the way to 20,000. Each time, the number is put through divisor-sum, which loops that number of times (to get all the divisors) using mod to check for even division, adds up the divisors, compares the sum with the original number, does some boolean to separate perfect/nonperfect by those rules, and starts all over again 11:52:15 that's my interpretation 11:52:21 probably horribly wrong 11:53:50 pretty close 11:54:11 only off on one detail, the loops never reach the upper bound 11:54:20 ie the main loop goes from 2 to 19,999 11:54:37 ahhh 11:54:37 and the inner loop (in divisors-sum) doesn't check the number against itself 11:54:57 it can be an important detail 11:55:02 then...how is the number checked against itself? 11:55:10 the most common use of do/loop is to loop x times 11:55:32 example: 37 0 do ... loop 11:55:39 that will execute "..." 37 times 11:55:41 yup, got that down 11:56:01 "do" is usually preceded by "0" 11:56:27 but in this case we know 0 and 1 can't be perfect 11:56:27 so in my example above i is first 0, then 1 all the way up to 36, but is never 37 11:56:35 OK 11:56:38 right 11:56:54 but it's ?do in this case 11:57:05 which is slightly different? 11:57:23 ?do is like do, except it handles the weird case where you give it a range of zero length 11:57:29 eg: 1 1 ?do ... loop 11:57:33 that doesn't do anything 11:57:39 where as: 1 1 do ... loop 11:57:50 would loop for a very very long time 11:58:09 oh, that's all? Then why not make regular "do" handle that by default (I'm just thinking aloud, I'm sure there are lots of important reasons) 11:58:25 ok, yeah that pretty much answers that 11:58:34 just in case I guess. Quartus put it there, mine had "do" 11:58:45 ok 11:59:52 so where is the number actually compared to the sum of its divisors? In over i mod 0= if i + then? 12:00:10 no 12:00:17 in "perfect?" 12:00:19 didn't think so 12:00:27 so it's just the "dup" 12:00:37 whaaat? 12:00:47 divisors-sum ( n -- sum) 12:00:54 that means you pass a number, and you get back a sum 12:01:06 do you know what "dup" does? 12:01:25 divisor-sum is a predefined function? 12:02:02 no, it's half your source code 12:02:18 do I have the name wrong? am I going mad? 12:02:31 woah........ *luinfana is Confused* 12:02:46 no, it's my idiotic approach to Forth, not your explanations 12:03:07 tell me if you understand this: 12:03:17 : foo ( -- ) 1 2 + . ; 12:03:29 do all those symbols and words make sense to you? 12:03:35 if not which ones don't? 12:03:39 the function "foo" will add 1 and 2 when called 12:03:45 right? 12:03:52 right 12:04:02 do you understand the ( -- ) part too? 12:04:26 a comment, in the form of input -- output 12:04:32 cool 12:04:35 just making sure 12:04:36 ok 12:04:40 so from there... 12:05:09 in your source code you define 3 words correct (the version Quartus pasted) 12:05:23 divisor-sum, perfect?, and check 12:05:32 --- quit: ygrek (Remote closed the connection) 12:06:04 ahh, yes. good. I opened it up again, so we're on the same page now 12:06:09 ok great 12:06:28 --- join: ygrek (i=user@gateway/tor/x-d213e7a22cc2a85a) joined #forth 12:06:51 so "check" is responsible for kicking off the program with the loop range, and displaying perfect numbers. 12:06:55 so PERFECT? is passed a number, and returns true/false (true if the number passed is perfect, and false otherwise) 12:07:09 yes 12:07:12 sounds good 12:07:35 and divisor-sum... 12:07:48 PERFECT? makes a copy of the number (DUP) passes one copy to DIVISOR-SUM (which returns the sum of its divisors) 12:08:08 then PERFECT? checks of the copy of the original number matches the sum returned by DIVISOR-SUM 12:08:21 (with "=") 12:08:31 AHHHHH the EQUALS SIGN 12:08:33 follow that? 12:08:39 that was what I was missing 12:08:43 yes, perfectly clear now 12:08:51 wow 12:08:52 --- join: arke (n=arke@p57A76277.dip.t-dialin.net) joined #forth 12:08:52 --- mode: ChanServ set +o arke 12:09:47 ahh, that's what you were missing :) 12:10:06 ok cool, thanks so much...it didn't make any sense without that one piece of information 12:10:20 I was wondering... you'd gotten all the tricky parts 12:10:27 :) 12:10:30 * gnomon cheers a little bit for luinfana 12:10:43 * luinfana bows 12:10:44 you understand . too? (period) 12:11:07 I think so...pops off the top of the stack and displays it? 12:11:23 yeah. good. just checking :) 12:11:25 ok 12:12:52 I think you've got the whole thing pegged 12:13:37 yeah I think so too...and it only took me like 3 hours! Yay! 12:14:43 not bad for a new language 12:15:00 I did this in assembly last year...oddly enough I understood that much, much more 12:15:27 how did you learn it? 12:15:37 this just takes a far different approach 12:15:50 oh, an assembly class 12:16:08 * is a comp. sci. major 12:16:08 sounds like more than 3 hours :) 12:16:15 and how'd you learn forth? 12:16:21 brute. force. 12:16:23 :) 12:16:44 heh, I suspected that learning technique would account for the difference in ease of learning 12:16:53 definitely 12:17:09 forth is suprisingly simple 12:17:28 For some reason I have a hard time understanding forth memory management. 12:17:57 Deformati: what? it hardly manages memory at all 12:18:10 I also found it more difficult to find complete, well-commented Forth programs for reference (and maybe I just wasn't looking in the right places), but I started doing Forth with a really bad attitude toward the language...it's not so bad once you figure out its "quirkiness" I guess 12:18:42 hope you won't mind my calling it quirky 12:18:50 nope 12:18:55 it's different 12:19:04 Allot confuses me. 12:19:08 most of is here (if not all) find that to be a good thing 12:19:22 : allot here + >here ; 12:19:24 pretty simple 12:20:24 variable v-here : here v-here @ ; : >here v-here ! ; 12:21:24 : , here ! here 4 + >here ; 12:21:54 Deformati: If you're confused, you're probably trying to imagine something complex 12:21:58 There really isn't anything 12:22:07 Forth figures it runs in one block of memory 12:22:25 and that everything important (like itself, variables, code, etc) is in the first part 12:22:36 HERE points somewhere in the memory block, after all that stuff 12:22:48 Hm. 12:22:52 You mean there is no heap? 12:23:05 what do you mean "heap"? 12:23:10 it has many definitions 12:23:18 C's malloc 12:23:37 some do, some don't 12:23:48 that's nothing to do with ALLOT 12:23:57 Yeah, that clear some things up. 12:24:17 ALLOCATE is basically like malloc() 12:24:40 Oh. 12:24:48 Allot is just the current stack pointer? 12:24:55 no 12:25:02 no 12:25:13 ALLOT changes HERE 12:25:23 which points to the middle of the main memory block 12:25:50 HERE returns a memory pointer into the middle (somewhere) of the main memory block 12:26:07 the idea is that _below_ HERE is "used" and everything past there is up for grabs 12:26:23 the way you "grab" some, is to move HERE ahead (up) 12:26:43 you can do this with ALLOT, "," (comma), etc 12:27:50 Oh. 12:27:51 follow? 12:27:59 Somewhat. 12:28:09 ALLOT and HERE don't do anything else, besides what I just described 12:28:11