Welcome to The Stack Overflow Podcast episode #119 recorded Thursday, October 12, 2017 at the Stack Overflow HQ in NYC. Today’s motley crew includes VP and GM of Stack Overflow Jay Hanlon, CTO David Fullerton, News Editor Ilana Yitzhaki, me, Producer Jess. Joining us are very special guests Michael Ludden, Director of Product at IBM Developer labs for IBM Watson, and SO Content Marketing Manager Rachel Ferrigno.
Special thanks to this week’s sponsors: Oracle and Digital Ocean.
Stack Overflow appreciates Oracle’s support for this podcast and for our community. Learn more about all the ways Oracle supports open source, java, and developers like you at oracle.com/developers. …
JS is horribly designed, and I'm constantly blown away by some of the decision choices when I (rarely) use it, but C++ is a lot more painful to work with.
Ehh, as long as I'm doing simpler tasks it's fine, but when I work with more complicated stuff, the error's are nearly impossible to figure out. Simple problems result in 100's of lines of errors.
I'm jaded from too many times spending 100+ minutes trying to figure out a problem that ended up being solved by removing a const
And also, the standard library is a little bit of a mess. The fact that boost exists speaks volumes about the way modern C++ evolves
The hating it part corresponds exactly with the time when C++11 was the flavor of the month and either a library would be drenched in boost, smart pointers and weird templates or was so archaic it wouldn't build under C++11
@Mego like if they would just take C++ and then put in Crystal's macro system I'd be so happy
idk templates are messed up and being seriously abused
@quartata But in all honesty, the real reason is probably mostly because I use it at work on a two person team (my manager and me) and he is waaaaaaay better at it than I am, so he ends up writing crazy template stuff that I only half understand. Which is fine when it works most of the time, but then when something goes wrong I spend way too long trying to figure it out before going to him for help.
Like literally just this morning he described something he did in templates that I can't even begin to understand
Not saying your manager is that bad but I wish people would dial it back a few notches when engineering complicated systems with templates or other hellish things like reflection
@quartata Yeah, he's certainly not anywhere near that. He's definitely not toxic like that. I'd be quicker to blame my relative noob-ness with C++ than to blame him for making things too complicated
Metatemplates are just inherently difficult to document and not very human readable. It depends on the scenario but if you're trying to balance doing things at compile time vs making code that another person has to work on I know which one I'll pick every time
The answer is the metatemplates because I'm a performance obsessed bastard but at least I'd shed a few tears for those I know will come after me
@Poke The product itself is kinda hard to describe, but the main case for our templates is hierarchical objects in a tree-like structure with polymorphism between the various different types.
So for each object, you can iterate through all of it's children (and even add more children dynamically) even though each child might be another collection or a parameter or an object interface
I'm doing an introductionory course on logic and was wondering what it means by 'It is really sufficient that A such that B.' and what's the difference between sufficient and really sufficient
The Lambda Calculus
I can't believe this doesn't exist yet.
Factiod:
The lambda calculus (specifically, the untyped lambda calculus) is, according to Wikipedia, "a formal system in mathematical logic for expressing computation based on function abstraction and application using variable bindin...
@ConorO'Brien There is a "newer" version which has just a single screen under all the keys together. Unfortunately this means that the pictures are "deep" down.
Wonder if it would be cheaper/feasible/cheap to use low-res diode grids instead of full-colour screens on each key. Monochrome should be enough for a keyboard.
ideally, all challenges (from low-rep users at the least) would go through the review queue with at least N (probably 4 or 5) votes to send the challenge to main
@ATaco That too. Or people who post in sandbox, ignore all comments and leave it there for like 2 hours, post it, get downvoted, and say the sandbox is dumb
The sandbox does have its deficiencies and some people who don't like it legitimately have reasons but I think I've seen the odd one or two who does ^
I sometimes use it but I have enough experience here to know if one of my challenges is definitely going to be well-received. I don't think I've been wrong in guessing before :P Not sure though
@EricTressler yeah, I stopped participating there as well for the same reason. it's a help site, not a "why didn't you already know this" site. At least, that's my most memorable experience
@ConorO'Brien mine was different; basically I posted a good answer to something, and someone decided it wasn't a valid argument, and it left a sour taste in my mouth. It only happened twice, but it was extremely negative
@MagicOctopusUrn I have a friend who literally thinks SO is a help site. I tried my best to dissuade him from posting a question, but he nonetheless did, and was ticked that "those people didn't help me for ~~~~"
@EricTressler I've lost whatever high regard I held for the SE system :P it's nice when you have a popular question that gets a good answer... but other than that, it's mediocre at best
Note: This is about the standard 9x9 sudoku puzzle. The solution only has to support solved, legal puzzles. So a solution doesn't need to support empty cells and can rely on the properties of a solved sudoku puzzle.
I was wondering this, but I couldn't think of an answer that I was content with....
@DJMcMayhem if you've coded it right, it takes exactly the same curve. Since your naïve algorithm has F(n) = F(n-1) + F(n-1) we have T(F(n)) = T(F(n-1)) + T(F(n-2)) therefore T(F(n)) is proportional to F(n)
@orlp I would generally draw the line at "produces original research, even rarely"... but I haven't thought about it very much. It's not that useful to draw the line, I guess.
@Neil Well yeah, that makes sense. But the timing is inherently noisy because I'm measuring seconds not operations, and I'm running it on a real machine that has an os and other programs running too
The Lambda Calculus
I can't believe this doesn't exist yet.
Factiod:
The lambda calculus (specifically, the untyped lambda calculus) is, according to Wikipedia, "a formal system in mathematical logic for expressing computation based on function abstraction and application using variable bindin...
