Mentoring session 14 Dec 2011 Arun and Eric

[04:45 PM] ericgj:	 satchmorun: hi I;m back	
[04:45 PM] satchmorun:	 hi	
[04:45 PM] ericgj:	 where do you want to start?	
[04:46 PM] satchmorun:	 i don't have any specific questions at the moment, so maybe we can start with anything that jumped out at you?	
[04:47 PM] ericgj:	 sorry, I'm back	
[04:47 PM] satchmorun:	 no prob	
[04:47 PM] satchmorun:	 i don't have any specific questions at the moment, so maybe we can start with anything that jumped out at you?	
[04:47 PM] ericgj:	 ok	
[04:47 PM] ericgj:	 first thing that jumped out at me was,	
[04:47 PM] ericgj:	 this code looks like how I'd write it :)	
[04:48 PM] satchmorun:	 :)	
[04:48 PM] ericgj:	 not that I've attempted either problem, so some of my impression is superficial	
[04:48 PM] ericgj:	 but generally really like how you organize the code and the documentation	
[04:49 PM] satchmorun:	 thank you	
[04:49 PM] satchmorun:	 i actually went through about 3 iterations of the stacker challenge before i got to a program organization that felt reasonably clean	
[04:50 PM] ericgj:	 that one definitely I could definitely tell you put serious work into	
[04:50 PM] satchmorun:	 the social network one is basically pretty procedural code wrapped up in a class	
[04:50 PM] ericgj:	 on the social network one, do you feel like your graph data structure was useful?	
[04:50 PM] ericgj:	 or would you have done it differently if you tried it again?	
[04:51 PM] satchmorun:	 that's a good question	
[04:51 PM] ericgj:	 it seems like that may be what makes it more procedural and harder to read	
[04:51 PM] ericgj:	 e.g. compute_rest_of_graph	
[04:52 PM] satchmorun:	 right	
[04:52 PM] ericgj:	 if you wrapped some of that in a class	
[04:53 PM] satchmorun:	 i'm trying to think about how i could do that, and i'm not sure	
[04:57 PM] ericgj:	 yeah, I'm not too much help not having studied the problem as you have.	
[04:58 PM] satchmorun:	 it might be that the problem is just kind of procedural in nature	
[04:59 PM] satchmorun:	 because in order to determine the connection from 2nd-order and up, you need to already know the first-order connections	
[05:00 PM] satchmorun:	 so (1) figure out all the first-order connections and (2) use those to iteratively calculate the remaining connections	
[05:00 PM] ericgj:	 yes. I was thinking if you parse the first-order connections into some structure you could easily ask,	
[05:00 PM] ericgj:	 "what's the distance from person A to person B?"	
[05:00 PM] ericgj:	 shortest distance	
[05:02 PM] ericgj:	 then take all combinations of pairs and process them	
[05:04 PM] ericgj:	 I'm no cs major, but it seems like there should be some classic algorithms out there to help	
[05:05 PM] satchmorun:	 me neither, but i do know there are a bunch of famous shortest path algorithms	
[05:06 PM] ericgj:	 right	
[05:06 PM] ericgj:	 the other thing that struck me about this one,	
[05:06 PM] ericgj:	 more of a style thing,	
[05:07 PM] ericgj:	 is that everything is done through the entry point (load_from_file)	
[05:08 PM] ericgj:	 not that it matters here, but I often see code where as little as possible is done in the entry point, initialize or whatever	
[05:09 PM] satchmorun:	 yeah	
[05:09 PM] ericgj:	 the heavy computations often being done in the last possible moment	
[05:09 PM] ericgj:	 so e.g. compute_first_order_connections.compute_rest_of_graph	
[05:09 PM] ericgj:	 called by your dump_graph or whatever	
[05:10 PM] satchmorun:	 the initialize method itself doesn't do very much, i just packed the logic into the load_from_file because it was, well, convenient	
[05:11 PM] satchmorun:	 but i see what you're saying - because if this were real, you'd want to be able to incrementally add tweets	
[05:11 PM] satchmorun:	 and wait to calculate anything until necessary	
[05:12 PM] ericgj:	 practically I guess it's just a question of whether you expect to do anything else with the input,	
[05:12 PM] ericgj:	 - right, or I guess that case too	
[05:12 PM] ericgj:	 but it completely works for the problem at hand	
[05:13 PM] ericgj:	 anyway, anything you feel could be improved on this one?	
[05:15 PM] satchmorun:	 well, i don't really think its exemplar code in the sense of clean, maintanable, clear	
[05:15 PM] satchmorun:	 the comments help a bit with the clarity	
[05:15 PM] satchmorun:	 as it is, i don't know what i'd try to improve on it without rewriting it	
[05:16 PM] satchmorun:	 and experimenting with different approaches	
[05:17 PM] satchmorun:	 there are some small readabilty things where certain small computations could be factored out into methods	
[05:17 PM] satchmorun:	 but ultimately, if i were going to try to improve it, i'd just start fresh, given how small a program it is	
[05:18 PM] ericgj:	 yeah	
[05:19 PM] ericgj:	 well I found it relatively very readable, definitely the comments helped	
[05:19 PM] ericgj:	 shall we move to the stacker problem?	
[05:20 PM] satchmorun:	 sure	
[05:20 PM] satchmorun:	 that's the way more interesting one anyway :)	
[05:21 PM] ericgj:	 could you explain what you mean by 'the main control structures could actually be written in "Stacker" as opposed to in Ruby'	
[05:21 PM] satchmorun:	 yeah	
[05:22 PM] satchmorun:	 so it turns out that real Forths are usually implemented in assembler, with a set of low-level primitive, from which control structures can actually be defined in Forth itself	
[05:23 PM] satchmorun:	 this article explains it better than i can: http://www.yosefk.com/blog/my-history-with-forth-stack-machines.html	
[05:23 PM] ericgj:	 so what would be the primitives here?	
[05:24 PM] satchmorun:	 it would be something like:	
[05:24 PM] satchmorun:	 a 'compile' primitive that switches the system into compilation mode	
[05:24 PM] satchmorun:	 a 'branch' primitive for marking the start of conditional branches	
[05:25 PM] satchmorun:	 a 'jump' for jumping to various points in the stack	
[05:25 PM] satchmorun:	 there would also need to be another stack for storing compiled-words-in-progress	
[05:27 PM] ericgj:	 interesting	
[05:28 PM] satchmorun:	 i had a lot of fun learning about Forth while doing this challenge	
[05:28 PM] satchmorun:	 it's one of those mind-bending languages	
[05:30 PM] ericgj:	 so in terms of your code, does this mean your Builtins would be expanded to define PROCEDURE, etc.?	
[05:31 PM] satchmorun:	 PROCEDURE would be the one control-flow word that would need to be implemented in the machine	
[05:31 PM] satchmorun:	 and then everything else could be defined with it	
[05:31 PM] satchmorun:	 PROCEDURE IF ... /PROCEDURE	
[05:31 PM] satchmorun:	 etc.	
[05:32 PM] satchmorun:	 if you go to the link i mentioned and search for 'conditional primitives', you'll see one example of how it can be done	
[05:33 PM] satchmorun:	 of course that's code from a C implementation, so this would be somewhat different	
[05:33 PM] satchmorun:	 because you wouldn't need certain kinds of pointers when you can just use Ruby arrays	
[05:37 PM] ericgj:	 interesting, so in terms of your code can you say how you would change it to say redefine ADD in terms of stack operation primitives?	
[05:37 PM] ericgj:	 (am I understanding right?)	
[05:38 PM] satchmorun:	 it's kind of like that	
[05:39 PM] satchmorun:	 but you wouldn't do that for ADD, as things like math would be available as primitives, much like they are here	
[05:39 PM] ericgj:	 ah, right	
[05:40 PM] satchmorun:	 but for, say, TIMES	
[05:40 PM] ericgj:	 TIMES, right just about to say	
[05:40 PM] satchmorun:	 you'd pop the top of the stack	
[05:40 PM] satchmorun:	 push it on the return stack	
[05:41 PM] satchmorun:	 set a pointer	
[05:41 PM] satchmorun:	 keep reading and executing words	
[05:41 PM] satchmorun:	 and then when you hit /TIMES, which would be a separate word	
[05:41 PM] satchmorun:	 you'd push 1 SUBTRACT onto the return stack (decrement the count)	
[05:42 PM] satchmorun:	 and jump to the previous marker	
[05:42 PM] satchmorun:	 if it the result of the decrement is 0, do nothing and keep going	
[05:42 PM] satchmorun:	 (or something like that)	
[05:43 PM] satchmorun:	 but it makes the internals a bit more complex than it is now	
[05:43 PM] ericgj:	 gotcha, I understand now	
[05:43 PM] ericgj:	 but an interesting exercise for sure	
[05:43 PM] satchmorun:	 agreed	
[05:44 PM] ericgj:	 regarding your current implementation,	
[05:44 PM] ericgj:	 I liked it a lot,	
[05:44 PM] satchmorun:	 thank you	
[05:45 PM] ericgj:	 but one thing that struck me was your mixins were pretty coupled to your interpreter	
[05:45 PM] satchmorun:	 yeah	
[05:45 PM] ericgj:	 what I've been doing (I don't know if others do this),	
[05:46 PM] ericgj:	 when I need coupled classes like this, I often define them within the class	
[05:46 PM] satchmorun:	 interesting	
[05:46 PM] satchmorun:	 except for the statemachine, the control flow modules were mainly just a convenient namespace	
[05:46 PM] ericgj:	 namespaced within the class	
[05:47 PM] ericgj:	 even if they are separate files - but usually don't separate the files unless they are huge modules	
[05:48 PM] ericgj:	 that way everything it in front of you and you don't have to remember where this instance variable was defined	
[05:48 PM] ericgj:	 et.c	
[05:48 PM] satchmorun:	 right	
[05:48 PM] satchmorun:	 that makes a lot of sense here, because those modules are kind of acting like little classes	
[05:49 PM] satchmorun:	 and all the @times_stack kind of instance variables are a bit of a smell	
[05:49 PM] satchmorun:	 i could just have a @current_compiler instance variable	
[05:50 PM] satchmorun:	 and set it to Times.new, Procedure.new, etc.	
[05:50 PM] satchmorun:	 instead of enter_procedure	
[05:51 PM] satchmorun:	 the one issue would be that those modules need to manipulate the @state instance variable	
[05:51 PM] satchmorun:	 for things to work as they do now	
[05:52 PM] ericgj:	 Times.new(self).compile ?	
[05:53 PM] satchmorun:	 that would work, but something feels weird about passing the outer class instance into the inner class	
[05:53 PM] satchmorun:	 but maybe that feeling is just silly	
[05:53 PM] satchmorun:	 though i think what this is really pointing to is that this stuff doesn't really belong in the Interpreter class	
[05:54 PM] satchmorun:	 and should be either part of the Dictionary	
[05:54 PM] satchmorun:	 or subclasses of a Compiler class	
[05:54 PM] satchmorun:	 i started futzing with the last option	
[05:55 PM] satchmorun:	 at one point, but i got frustrated by having to pass a ton of state between classes	
[05:56 PM] satchmorun:	 i may have just not hit on the right design though	
[05:56 PM] ericgj:	 right, I see what you mean	
[05:57 PM] ericgj:	 but if you go to the more Forth-y implementation, I guess you wouldn't have different compilers	
[05:57 PM] satchmorun:	 right	
[05:58 PM] satchmorun:	 that's a project i definitely plan to play with at some point	
[05:58 PM] satchmorun:	 mainly because i think Forth is really interesting	
[05:59 PM] ericgj:	 yeah, now you have me interested	
[05:59 PM] ericgj:	 I may try this problem after all	
[05:59 PM] ericgj:	 :)	
[05:59 PM] satchmorun:	 :)	
[06:00 PM] ericgj:	 while we have a few minutes left, do you have any questions about the core course?	
[06:00 PM] satchmorun:	 i do	
[06:00 PM] satchmorun:	 i actually had more than i do now, but i read the last post on the rbp blog, which cleared some things up	
[06:01 PM] ericgj:	 that's good	
[06:01 PM] satchmorun:	 my primary question is: what kind of time commitment is expected for those 3 weeks	
[06:01 PM] ericgj:	 we could do a lot better...	
[06:01 PM] satchmorun:	 and related: what are the exact dates again?	
[06:02 PM] ericgj:	 time commitment: generally 10 - 20 hrs a week is what people are finding.	
[06:03 PM] satchmorun:	 (i remember it starting on the 16th, but just wanted to confirm)	
[06:04 PM] ericgj:	 hold on, looking it up...	
[06:04 PM] satchmorun:	 10-20 is doable - 20 would be a bit uncomfortable, but i expect the discomfort would be worth it	
[06:06 PM] satchmorun:	 and another scheduling-style question: are there times where everyone is in the same digital place at the same time, or is it mostly asynchronous?	
[06:06 PM] ericgj:	 it's mostly if not all asynchronous	
[06:07 PM] ericgj:	 sorry, to clarify: the instructors do set office hours	
[06:07 PM] ericgj:	 when they can be reached for sure on IRC	
[06:08 PM] ericgj:	 but mostly it's done through the university-web assignment and review system	
[06:08 PM] satchmorun:	 got it	
[06:08 PM] ericgj:	 and email to some extent	
[06:09 PM] satchmorun:	 sounds good	
[06:10 PM] ericgj:	 there is always collaborative aspects to the assignments	
[06:11 PM] ericgj:	 so you'll be reading your fellow students' code and vice-versa	
[06:11 PM] satchmorun:	 makes sense	
[06:12 PM] ericgj:	 it's a great experience, and getting better every time the course is given from what I can tell.	
[06:12 PM] satchmorun:	 i'm definitely looking forward to it	
[06:12 PM] ericgj:	 if you can fit in the time...that's always the tricky part!	
[06:12 PM] satchmorun:	 there's a lot about the overall approach that i really like	
[06:12 PM] satchmorun:	 that will definitely be the tricky part	
[06:13 PM] ericgj:	 anyway, do you have any idea what your personal project for the course will be?	
[06:13 PM] satchmorun:	 but, within reason, time can always be made	
[06:14 PM] satchmorun:	 one idea that i've been toying with the last few days (because i could use it at the moment)	
[06:14 PM] ericgj:	 always a good reason...	
[06:15 PM] satchmorun:	 is a good way to use Mustache templates in rails	
[06:15 PM] satchmorun:	 there are a couple of solutions out there	
[06:16 PM] satchmorun:	 but they don't work the way i'd like and don't quite flex the right ways for me	
[06:16 PM] satchmorun:	 but i'm not sure if something like that would be in scope for what this course is looking for	
[06:16 PM] satchmorun:	 or is it? that's actually a useful question to ask	
[06:16 PM] ericgj:	 I think it would, sounds really useful	
[06:17 PM] satchmorun:	 so what kinds of things end up being personal projects?	
[06:17 PM] ericgj:	 lots of folks have done rails related projects	
[06:18 PM] ericgj:	 but many aren't.	
[06:18 PM] satchmorun:	 ah, and i see there's actually a student projects page on the rmu site	
[06:19 PM] ericgj:	 you can see a list http://university.rubymendicant.com/resources/student_projects.html	
[06:19 PM] satchmorun:	 yeah, that's just what's on my mind right now	
[06:19 PM] ericgj:	 oh, btw the course starts on Jan 9	
[06:19 PM] ericgj:	 runs to the 30th	
[06:19 PM] satchmorun:	 did that change? or was that always the way it was?	
[06:20 PM] ericgj:	 I think it usually does start the second week of the month	
[06:20 PM] satchmorun:	 it's not a problem, but i have the number 12 and 16 in my mind	
[06:20 PM] satchmorun:	 and 16 made more sense as it's a monday	
[06:20 PM] ericgj:	 yeah, I confirmed with Jordan, who would definitely know	
[06:21 PM] satchmorun:	 thanks, i appreciate it	
[06:21 PM] ericgj:	 practical projects that would make others lives easier, no matter the technology, are greatly encouraged	
[06:22 PM] ericgj:	 so the Mustache thing sounds perfect	
[06:22 PM] satchmorun:	 excellent	
[06:22 PM] satchmorun:	 i do plan to spend a little time brainstorming to see what else i can come up with	
[06:22 PM] ericgj:	 but Greg will give you feedback about it when the course starts	
[06:23 PM] satchmorun:	 sounds good	
[06:23 PM] ericgj:	 anyway, gotta run, great talking with you	
[06:23 PM] satchmorun:	 you too	
[06:23 PM] ericgj:	 nice code	
[06:23 PM] satchmorun:	 it was interesting to chat	
[06:23 PM] satchmorun:	 :) thanks	
[06:23 PM] ericgj:	 I feel like I learned a lot ! :)	
[06:23 PM] satchmorun:	 :)	
[06:24 PM] ericgj:	 good luck with the course and hope to work with you when you "graduate" !	
[06:24 PM] satchmorun:	 take care, i'm sure i'll be seeing you around	
[06:24 PM] ericgj:	 yep	
[06:24 PM] satchmorun:	 thanks again for taking the time