The Whiteboard

6:00 PM

if I was more talented I'd make an image that was something like "not sure if serious or trolling - wait definitely trolling" :)

user41796

@enderland Well, I just realized why they don't let the newer generation bots vote. Look at what happened to gnat's voting patterns

enderland

plot twist: Stack Exchange really is in the business of perfecting AI and most users are experiments from them

user41796

Hopefully they revised their 3 rules after the earlier AI versions.

enderland

"making the internet a better place" sounds suspiciously like this XKCD...

user41796

Yep, MFA.

enderland

6:07 PM

well, it's about programming and is asking a specific question, so if that is off topic here, where should I ask it? — ChrisBint 42 mins ago

hrm. should having to rewrite many of your test mocks be a sign of a bad test design or just... a large refactor?

I guess the fact that I'm changing the things that were previously mocked differently seems... the cause, so that's not "too" bad I guess?

It just feels wrong. hah. though it's helpful for refactoring to make sure all the logic stays the same to have to fix tests...

user41796

Just sounds like you're refactoring the code base is all.

enderland

I guess by definition my tests have to fail. since the underlying calls I was mocking changed...

user41796

Y U BREAK UR CODE?

enderland

experimenting with replacing some of the stuff I wrote with an API

Aaron Hall

IT tip of the day: nothing solves problems with indirection like more indirection

user41796

6:21 PM

Now you have { undefined } problems!

Telastyn

More indirection, more planning, more process, more people, more money, more abstractions... not the solutions to problems.

enderland

@Telastyn but job security :(

user41796

You just fled a job with lots of security

enderland

:-)

and am quite happy about it, tbh

there. more clear what I meant :)

Telastyn

sorry, I had a meeting this morning where I found out that senior leadership thinks our date is unrealistic, so they're going to go through a big long estimation process (again) to come up with a new date.

since the process worked so well the last time.

Jimmy Hoffa

6:36 PM

humbug. Do more work, create more code, construct more crap, all so that we can have an easy way to do less, which is actually a hard way to do more, instead of just not doing it all to begin with. Where'd I put that glass..

Defaults: Because nobody knows what they're doing, including the people designing the defaults apparently!

Telastyn

heheh

Jimmy Hoffa

mumble...grumble...

Telastyn

configuration is evil - always.

Jimmy Hoffa

@Telastyn which is why it should be utterly stringy as hell. It should be oversimplified beyond all belief because anything else will turn it into the spawn of satan. True Facts.

Anytime someone thinks their configuration system needs a feature added to it; there needs to be a Configuration Leprechaun that just leaps outta the corner of the dimension with a left jab, and disappears back to the corner of the dimension.

enderland

just add a few more user options, can't be too hard right?

Jimmy Hoffa

6:42 PM

> I want my configuration system to allow inheritance of pop smack pop ouch!

enderland

though I guess, in some systems, if your answer to that is "not too hard" it means you designed it well

Jimmy Hoffa

A) I didn't design this system, B) adding a few options is one thing, it's when you want some defaults for those options which are occasionally one way and occasionally another but you want it to just automatically be occasionally one or another without any work to change the setting saying which way it goes that makes me want to bite people. Full cannibal mode; f'real.

Aaron Hall

Speaking of not too hard, I felt like my multiple inheritance demo last night was not too hard, so that says a lot for Python, I think... :)

Jimmy Hoffa

@AaronHall MI is easy in some languages, but always solves a problem that would be better solved a different way

enderland

@AaronHall speaking of python, is there any data structure in python that matches this format for catching multiple exceptions?

Telastyn

6:46 PM

the best configuration system I've used was a POCO that got serialized out to json. Pretty much all the code just used the typed POCO and didn't care where it came from.

enderland

I had incorrectly assumed it would be [e1, e2] instead of (e1, e2)

Aaron Hall

I don't think you can solve this better than with multiple inheritance: gist.github.com/aaronchall/ae173ef9343f30d10a28

You mean suppresses exceptions like pass does?

enderland

no, I mean more like why the datastructure for (e1,e2) is not [e1, e2]

Aaron Hall

tuples vs lists?

enderland

yeah. is there a significant meaning for why the "except" would take a tuple instead of a list?

Aaron Hall

6:48 PM

Why can't you use a list?

enderland

or is that... relatively arbitrary

@AaronHall it seems to require the tuple for that

[e1, e2] does not catch both exception types (perhaps neither?)

Aaron Hall

I know that unassigned lists are transformed to tuples as a minor optimization (since you won't be mutating them)

enderland

o_O

Aaron Hall

so one would think using brackets should be ok

probably the parser only allows tuples as an optimization for the exceptions?

enderland

I'm not sure. I guess this is less trivial than I thought, I'll work up a MVWE

Aaron Hall

6:50 PM

Maybe you could get that changed, try the Python mailing list

enderland

I'm curious more than anything

Keen

@enderland The docs says the syntax there may be odd due to backwards compatibility with older syntax. Maybe a list there triggers older syntax parsing?

Jimmy Hoffa

@Telastyn I wrote one of these that just uses XML from the app.config to serialize to POCO. You define the configsection to be DynamicConfigurationSection and in the heading of the section you put type="Some.Type" and the DynamicConfigurationSection deserializes it to POCO. In code it's just DynamicConfigurationSection.Get<Type>("name")

it's still more overhead than simple JSON but it fits nicely with the .NET approach.

Telastyn

nod

enderland

@Keen not sure. I'm writing an SO question :)

@Keen I think that refers to Exception, e vs Exception as e

0

Q: Why does handling multiple exceptions require a tuple and not a list?

Consider the following example: def main_list(error_type): try: if error_type == 'runtime': raise RuntimeError("list error") if error_type == 'valueerror': raise ValueError("list error") except [RuntimeError, ValueError] as e: print str(e...

Keen

7:00 PM

@enderland Right, the old syntax 'Exception, e' is blamed for why we need to use a tuple there. I bet that a [Exception, e] would also trigger the parser to think you're using the old syntax.

@enderland thx, i'll be watching that.

bah, i miss being able to edit chat messages for longer periods.

user55340

@Keen too bad we can't make you an honorary blue for the room.

Jimmy Hoffa

7:19 PM

@MichaelT do blues get longer chat message edit times?

oh blues, yes, of course

thought you meant italics

Owners don't get a longer period than non-owners, do they?

user55340

@JimmyHoffa nope. We just get to move and pin.

amon

@enderland The answers to that are astonishingly unsatisfactory. I strongly assume someone just implemented the except clause as a proof of concept with tuples, then forgot to properly design the features. Now hordes of SOsians and mailing list dwellers are trying to do weird post-hoc justifications about tuple efficiency. Some language design grows organically, as in, covered in *bleep*

@AaronHall, would you enjoy submitting a PEP to end this madness?

Jimmy Hoffa

@MichaelT how do you cleanly solve this without multiple inheritance. (No time to actually read that and figure out what it's talking about and work up a solution m'self right now)

I've faith non-MI solutions are better in effectively every scenario, if you take the time to find a good non-MI solution.

Ixrec

my personal impression is that whenever you think you want MI, what really happened is you painted yourself into a corner design-wise and are looking for an easy way out instead of addressing the core issue

Jimmy Hoffa

^^

or in the parlance my colleagues have gotten used to hearing from me; you need to simply Try Harder

MI is an x/y solution

Ixrec

7:34 PM

and half the time the answer is "just use composition, it's not even that hard"

amon

I think the MI problem maps somewhat cleanly to MI, but using MI is by no means necessary. The alternative solution is to create a system of roles, where each soldier has exactly one role. The role can conform to an “Archer”, “Horseman”, or “Flying Rain of Fire” interface, the last of which subsumes the other two. The special commands can only be given after retrieving a view onto that soldier through that role. This solution is only multiple subtyping, not multiple inheritance.

Ixrec

oh it was the codeless code thing, I hadn't even followed the link yet

user55340

@JimmyHoffa that's part of the problem. MI is one of the distasteful approaches... Though not an option at all in Java at the time of writing.

Ixrec

Java even allows the thing that's sort of like MI but actually not terrible

user55340

The idea is soldier as a base class. Archer a d horseman as subclasses and then... Yea a mess with that last one.

user55340

7:37 PM

@Ixrec default methods go a ways at resolving this.

user41796

> Oh hey! This requirement I hadn't mentioned and completely derails what you've already been doing is one we can't budge on. Sorry about that!

Ixrec

^ half of our potentially decent questions do this =(

user41796

I'm just irked because it was a last minute curve ball during what was supposed to be the final presentation.

Aaron Hall

7:55 PM

@amon I don't think I'll be proposing a change. See my answer for why.

Jimmy Hoffa

@MichaelT Ok, my turn, let me look at this and see what I can solve...

Duga

It might be worthwhile looking at the types of questions that are on topic at programmers.stackexchange.com. — R Sahu 31 secs ago

Jimmy Hoffa

So if I understand correctly:
Soldiers can FightToDeath
Archers can FightToDeath, and ShootFoe
Horseman can FightToDeath, and TrampleEnemy
RainOfFire can FightToDeath, and ShootFoe, and TrampleEnemy

@MichaelT is this a correct assessment of the situation?

there's pieces of knowledge as well (rank, horse name) etc

enderland

@amon was it just the one answer that was lacking? I only see two total - one of them being Aarons

user55340

@JimmyHoffa rof can also charge.

amon

8:01 PM

@AaronHall Your quite extensive answer does a poor job of answering the question, but contains one interesting thought “we appear to be doing type checking for the special case of a tuple. It would certainly make the code slower to add another type to check for”. But why do we have to check for specific types, rather than just testing for iterability? Until shown something like a design document or a quote from the designer, I'll just assume this feature simply wasn't designed at all.

@enderland There's a self-deleted answer which I commented on. I'm dissatisfied with all answers, but least dissatisfied with Aaron's (see above)

enderland

@amon I need 10k rep on Stack Overflow! haha

durron597

@enderland I'd bounty you 500 if they hadn't added 25k permissions

I am sort of interested in site analytics, not enough to answer any crappy java questions though

enderland

HA

amon

I also don't buy the “allowing anything except tuples would be slow” argument. It would be truly trivial to optimize a tuple literal to a more efficient type matching implementation rather than actually allocating some container. Part of the problem making this inelegant is Python's extreme late binding. We don't know that an identifier refers to a type until it is evaluated, in contrast to Java where we need special syntax for multi-exception handlers.

Jimmy Hoffa

SoldierRole { FightToDeath(), Rank }
ArcherRole { ShootFoe() }
HorsemanRole { TrampleEnemy(), HorseName }
RainOfFireRole { LeadCharge() }

Army {
  Dictionary<Name,  SoldierRole> Soldiers
  Dictionary<Name, ArcherRole> Archers
  Dictionary<Name, HorseManRole> Horsemen
  Dictionary<Name, RainOfFireRole> RainOfFireMembers

  AddRainOfFireMember(string name) {
    Soldiers.Add(name, new SoldierRole())
    Archers.Add(name, new ArcherRole())
    Horsemen.Add(name, new HorsemanRole())
    RainOfFireMembers.Add(name, new RainOfFireRole())

there's one simple compositional approach

enderland

8:08 PM

@amon tuples are also only marginally better at memory usage than a list and seriously who is going to have something like Exception (e1, e2,... e99999) anyways :)

amon

I actually remember one case where I wanted to have non-hardcoded exception types in a catch clause, though that was in Java.

durron597

@amon +1, I don't really understand the why any better than i did before reading your [Aaron Hall's] answer

Ixrec

it sounds like the answer is there is no why

Jimmy Hoffa

^^ @AaronHall as we've said before; you'd really benefit in your communication/instruction towards other developers from learning more languages and getting out of your Python bubble

I've found lots of your answers to be a bit confusing and I think that's somewhat why, you should try to broaden your perspective a bit

Ixrec

finally read Aaron's answer, I have to admit I'm also confused; there seems to be an argument against making your exception tuple/list a variable that can be changed by other parts of the code at runtime (which, tbh, would be insane anyway), but I don't see anything that explains why a literal list would be any worse than a literal tuple

enderland

8:20 PM

@Keen that's how I feel on SO chat. :(

user55340

@enderland but why would you ever go there.

BarryTheHatchet

@Keen did something change?

Keen

@BarryTheHatchet I used to have a diamond.

BarryTheHatchet

ooh, you powered down

you don't even seem to have a SFF account any more??

not so Keen on SFF?

Aaron Hall

Thanks guys, still trying to improve it, but I'm at work too...

Jimmy Hoffa

8:27 PM

@Keen oo, drama? drama drama? We get none of that here; the most dramatic event we have is when we look in our mugs and can see the bottom. Spill! What fun!

BarryTheHatchet

missed msgs dammit

Keen

@BarryTheHatchet rand was elected moderator there, so I had my account removed.

user55340

@Keen and broke the election page too. Bravo.

Jimmy Hoffa

So I guess we're a refugee camp now. @BarryTheHatchet was a refugee escaping the travails of TL, but one refugee doesn't make a set. @Keen if you come by way of escape hatch route from SFF then we've got a set which is technically an entire demographic now.

BarryTheHatchet

@Keen I don't understand why an election result causes the staff to remove your account. Does SFF follow a different model?

enderland

8:28 PM

sigh

BarryTheHatchet

like mods are participants in the hunger games

that could be fun?

Keen

@BarryTheHatchet ...i requested the account deletion.

user114359

Whoever voted to migrate this to SO... please don't. — Snowman 40 secs ago

BarryTheHatchet

@Keen oh right I see

that's one heck of a protest vote.

Keen

i don't like to half-ass things. i whole-ass them instead.

BarryTheHatchet

8:30 PM

u ass

enderland

:( no one has answered my last Scifi question... alas, it's probably way too hard

user55340

@BarryTheHatchet the past activities of the newly elected mod caused some drama in the election and even spill over on MSE.

BarryTheHatchet

ok well welcome to the Whiteboard where the biggest problem we have is that people keep asking questions on Stack Exchange

enjoy :D

also feel free to suggest new nicks that I can use throughout the year.

Keen

@enderland looks it up That'll be a hard one to answer, if you include his participation in destroying the second Death Star.

BarryTheHatchet

@enderland meh, it's a bit of a "research for me plz" question :/

enderland

8:32 PM

@Keen nah, just the direct kills

user114359

@enderland link? I can't be arsed to find it myself

enderland

@BarryTheHatchet that's like 95% of scifi questions

Keen

1

Q: How many confirmed kills does Wedge Antilles have?

I have been rereading some of my favorite Legends Star Wars books and it occurs to me... Wedge Antilles has a lot of kills. It seems that he consistently is getting multiple kills in every engagement in each of the books and he is in a lot of books. I understand that many of these are in non-Dis...

star-wars star-wars-legends

BarryTheHatchet

one might suggest that if you want to know the answer, watch the movies and count 'em

cos you're basically asking someone else to do that for you

@enderland sadly so

Jimmy Hoffa

@BarryTheHatchet If the rest of SO is The Hunger Games, and this is the sanctuary for escapees, that makes this the rebellion headquarters. I guess that makes me the crazy bald lady; @GlenH7's scotch participation makes him Woody Harrelson; so @MichaelT must be Phillip Seymour Hoffman.

enderland

8:32 PM

@BarryTheHatchet it's the EU books that I'm more interested in ;)

Keen

@BarryTheHatchet Worse than that, it's asking about the many books that include Wedge as well.

BarryTheHatchet

@JimmyHoffa can I be the hot one?

@Keen enderland is "whole ass" too :D

Keen

And I bet he had kills pre-ANH too.

enderland

@Keen yeah, question is how many of them are documented

user114359

@Keen I don't even know what character that is

user55340

8:33 PM

@Keen ohh - canon fights! Legends vs Disney.

enderland

you star wars heathen @Snowman!

BarryTheHatchet

@Snowman SPOILER he's Darth Maul's mother

Keen

@Snowman He's like a B-level Star Wars character. One of the handful of people to survive the attack on the Death Star in ANH, the Empire's attack on Hoth in ESB, and the attack on the second Death Star in RotJ.

BarryTheHatchet

(and had the good sense to stay away from TFA)

user55340

Rogue squadron IIRC.

Jimmy Hoffa

8:34 PM

@Keen honestly though - I'm always curious how people come upon this little corner of SE, especially given the growth it's seen over the past year or so- what did link you here?

enderland

@JimmyHoffa he has been here quite a bit

Keen

@JimmyHoffa After leaving SFF.SE, I was looking for other chat rooms. Picked this one since I'm a budding programmer.

enderland

@MichaelT yeah. he is the founder of Rogue Squadron, not to be confused with Rouge Squadron

user114359

@enderland Rouge Squadron must be where the men of Rogue Squadron go to "unwind" after a tough battle

BarryTheHatchet

@JimmyHoffa I can't remember how WB came to my attention tbh

it's not like I frequented Programmers

heck, I still don't...

user114359

8:37 PM

And I haven't read any of the Star Wars books or seen anything other than the movies. So I don't know all of those obscure details

BarryTheHatchet

@Keen Unless you're a budding software designer, your interest in this chat room is off-topic.

Jimmy Hoffa

@Keen ah, just dropped into P.SE and looked for it's room then?

Ixrec

all the obscure details I know are from browsing SFF

user55340

I will note we do advertise it in close reasons, help center and meta faqs.

user114359

@BarryTheHatchet lolwut? We also talk about coffee and whiskey on a regular basis.

Jimmy Hoffa

8:38 PM

@MichaelT true, but then it would probably be well known how he got here as we typically mention to each other to watch for a drop-in and link the source Q/A in those events.

user114359

Not to mention @JimmyHoffa's rants on various random topics.

BarryTheHatchet

@Snowman cough s/programmer[s]/software design[er]/ :D

enderland

@BarryTheHatchet you probably mistakenly sent a "migrate to programmers" comment

Jimmy Hoffa

@BarryTheHatchet maybe we should change the name of the site?

BarryTheHatchet

@enderland heh nah I've been aware of Programmers.SE since the early days; I just don't know how my Whiteboarding came about

probably via a chat flag or someone oneboxed a particularly stupid message into the Lounge

enderland

8:39 PM

@BarryTheHatchet I know, but I suspect there was a comment chain of argument about it

BarryTheHatchet

@JimmyHoffa hey, good idea! can't believe I never thought of that

Jimmy Hoffa

@BarryTheHatchet you showed up around same time the TL had a couple of shitstorms broadcast all over via MSE and chat flags; You could have gotten linked around MSE or other places

BarryTheHatchet

and I'm the expert at changing names!

@JimmyHoffa lol you make it sound like the Lounge wasn't constantly a shitstorm broadcast all over via MSE and chat flags

anyway the upshot is that I'm here now and I hate you all <3

so what have I missed? been away a lot lately.

user114359

wtf another name change? I finally saw the Supergirl avatar.

BarryTheHatchet

I see

Keen

8:44 PM

@JimmyHoffa Yup.

user114359

:28084085 what's disgusting? I have no idea what you're talking about :-)

BarryTheHatchet

@Snowman To be fair I did give you like two weeks' warning

@Snowman huh? not sure what you're referring to!

haw haw

so I'm back on Docker work this week. finally going to finish that whole thing

in fact, hah, that's why I came in here! /cc @JimmyHoffa

enderland

heh I've been working a ton with Docker the past month ;)

BarryTheHatchet

I had a chat-like question to ask y'all, then kept the tab open because it was an ongoing project. and just sort of forgot to leave

@enderland I saw! Did you ever get that routing problem solved?

enderland

@BarryTheHatchet ... not really - the answer basically is (I think) that docker really doesn't like talking between networks

BarryTheHatchet

8:47 PM

@enderland well that was stated in your question

enderland

I think that's basically the answer, though

BarryTheHatchet

not a very good answer ;p

Jimmy Hoffa

Speaking of Whisky; @GlenH7 you ever finish that bottle of Connamara?

BarryTheHatchet

it's a con, mara!

user41796

@JimmyHoffa Long gone

Keen

9:03 PM

@enderland Looks like we need to ask Antoine Pitrou: github.com/python/cpython/blame/…

Jimmy Hoffa

@GlenH7 Where's it rate on your scale?

What're you working on now then?

psr

@MichaelT ericlippert.com/2015/04/27/wizards-and-warriors-part-one

BarryTheHatchet

@JimmyHoffa An aged GlenH7, obviously

user41796

@BarryTheHatchet "Forgot to leave?" I'm reasonably certain there's several of us in here who could help with that... :-P

user55340

@psr good and related read! And non-trivial.

BarryTheHatchet

9:07 PM

@GlenH7 :(

user41796

@JimmyHoffa Definitely on the positive side, was a very enjoyable bottle

user41796

@BarryTheHatchet No need for frowning faces, just letting you know that we're here to help.

psr

@MichaelT Makes the excellent point that you really shouldn't assume your business rules will happen to match the static typing rules of your language.

BarryTheHatchet

I'd get screamed at for that level of sarcasm

user41796

@BarryTheHatchet Nah, I would hope not

Jimmy Hoffa

9:14 PM

@GlenH7 to be fair, I think that's basically the reason all of us are here.

user41796

Probably more truth to that than I care to admit

user114359

@JimmyHoffa you can check out any time you like but you can never leave!

Jimmy Hoffa

room topic changed to The Whiteboard: General Discussion for programmers.stackexchange.com Pants optional when telecommuting. [coffee-day] [horror-stories] [xkcd.com/1305/] [my-code-is-compiling] [scotch] [the-other-monitor]

enderland

it'd be nice if that showed the before and after

user41796

@JimmyHoffa meh, that doesn't work as you would hope

Jimmy Hoffa

9:17 PM

room topic changed to The Whiteboard: General Discussion for programmers.stackexchange.com Pants optional when telecommuting. [coffee-day] [horror-stories] [xkcd.com/1305/] [my-code-is-compiling] [scotch] [the-other-monitor]

@GlenH7 not really; no. Fuck it though. It's up to others to figure it out. Not my job.

user41796

How do you really feel about that?

user114359

room topic changed to The Whiteboard: General Discussion for programmers.stackexchange.com Pants optional when telecommuting. [coffee-day] [horror-stories] [xkcd.com/1305] [my-code-is-compiling] [scotch] [the-other-monitor]

user114359

@JimmyHoffa I give up too

enderland

what are you trying to do?

user41796

They're trying to ~~break tag links~~ link to the comic instead of the prog's tag page.

user114359

9:22 PM

clearly

enderland

yeah. clearly. ...

user55340

Almost works on mobile... But that's mobile.

Jimmy Hoffa

room topic changed to The Whiteboard: General Discussion for programmers.stackexchange.com Pants optional when telecommuting. [coffee-day] [horror-stories] [my-code-is-compiling] [scotch] [the-other-monitor] [xkcd-1305]

now it should work..

@GlenH7 need your approval on my tag wiki...

user114359

@JimmyHoffa I approved it

user114359

probably needs one more

amon

9:26 PM

there.

enderland

wow that is depressing

probably doesn't help the acoustic song I'm listening to is "Never Meant to Belong"

user55340

You know some day the big Y is going to go "WTF is this?"

user41796

@enderland Life frequently is. passes over a glass of scotch

user41796

@MichaelT That's exactly what I was thinking

enderland

the big Y?

user55340

9:28 PM

Say not his name so it's harder to find that message.

enderland

hahahaha

gotcha

user114359

@MichaelT I assume that is the Greek letter Y?

user41796

It will expire off in a week anyway, right? No Q associated with it

enderland

@GlenH7 hrmm, that's sad

user55340

@GlenH7 has wiki. Not sure.

user41796

9:30 PM

@Snowman It is the Alpha and the Omega. All is encompassed.

user41796

@MichaelT Then we can declare: Science!

user55340

@Snowman Υ♦️

enderland

@GlenH7 or irony

Phil Lello

What's the right site to discuss handling technical interviews where the interviewer got the answers wrong?

user114359

@PhilLello none of the above. The subject of the question may be on-topic, but not as an interview question

user114359

9:32 PM

19

Q: Why do interview questions make poor Programmers.SE questions?

Occasionally a question that is asked at an interview comes up - "how would you design Amazon Web Services" or "how would you write Google" or some such question. These questions find their way to Programmers.StackExchange in hopes of getting an answer for the next time they are asked on an inte...

discussion close-reasons questions faq-proposed

enderland

44

Q: How to tell a interviewer that he is wrong on a technical question

Yesterday I went for a job interview at a product based company. In one of the interviews the interviewer asked me a question that I answered correctly (I have cross checked it) but he kept on asking me the same question again and again. First I thought maybe he just wants me to get confused and...

interviewing professionalism

Ixrec

the part about interview etiquette is for The Workplace, the actual technical question itself goes to one of the technical SEs like SO or Programmers or a few dozen other places depending on what it was

enderland

yuck who wrote that answer with the syntax it had

Phil Lello

That may in fact be the question I would pose.... I've researched a few network stack topics that came up; turns out I was right and the interviewer kept pushing for the wrong answer!

user114359

is there a meta post somewhere describing how rep is grandfathered in on deleted questions? I know if it is deleted before a certain point you lose any rep changes, I'm just not sure what that threshold is

enderland

9:38 PM

Don't overestimate interviewers, I'm sure there are plenty would would argue with you from a point of being wrong too

user41796

@PhilLello The correct answer is to call that a bullet dodged and to move on

user114359

@PhilLello that may be a positive sign: do you want to work somewhere where people are clueless?

user114359

what @GlenH7 said

Jimmy Hoffa

@PhilLello TWP and the answer is to quit

@PhilLello you didn't want to work there anyway then.

amon

Oh, so we're back to the “cultural fit” discussion now? Though that might really be applicable here, since there's a difference between friendly fallibility to foster learning and ossified “truths” that won't be questioned.

Phil Lello

9:46 PM

I think it would be a tad unfair to say clueless, I think it's more a case of dated knowledge (like interrupts vs polling in device drivers). I'm hopefully at a stage in my career where the person above me doesn't need to know how to do my job, just that I know how to do it.

Aaron Hall

ok, ye naysayers, say ye nay no more: stackoverflow.com/questions/35851782/…

user55340

Heh. Physics is writing a bunch of canonical dups for relativity questions.

Aaron Hall

man, IE likes to choke on 5000 lines of code in a mercurial web view.

user41796

@AaronHall you forgot the semi-colon in the TL;DR

Aaron Hall

It's a comma splice.

Totally legit.

user55340

9:48 PM

@GlenH7 Python is anti semicolon.

user41796

passes himself a glass of scotch

Aaron Hall

Apparently grammarians will have to deal with "scare quotes" too... sigh.

Jimmy Hoffa

@GlenH7 I should do this... What's in your bottle right now?

I'm still working on a cheap bottle of Calvados I got a while back. Getting an itch for a proper bottle of Whiskey again. Hmm...

Aaron Hall

Now it's an "Executive Summary"

amon

@AaronHall Your answer is truly impressive and tries to teach how to fish. Reading bytecode or even the interpreter source is a convenient skill to answer many questions about implementation behaviour. Though I still don't see an answer to the “why” question.

Which leads me to believe that @enderland just picked the wrong site. Should have asked on Programmers, tagged with history, received -7 score, almost get closed, then receive a +3'ed answer after 2 days.

Aaron Hall

9:56 PM

why, you want more history?

we used to raise strings

amon

language design question vs. implementation details

Aaron Hall

so we didn't catch error types

we caught strings

amon

@AaronHall we still do in Perl. Better libraries declare their exceptions to match a certain regex. *shudders*

Aaron Hall

so there wasn't even a tuple in the beginning

amon

“In the beginning, there was Guido. And Guido said ‘let there be tuples’. And the tuples were good.”

user41796

9:59 PM

@JimmyHoffa Current bottle I have is from the pacific NW, but I can't remember the name. Wheat based whiskey, rather tasty. Fairly smooth with a good number of notes to it.

user41796

The marketing blurb on the back has a lot of references to fly fishing. Ha! Just found it - Washington Wheat.

Aaron Hall

I'm frustrated with my inability to satisfy your question, but perhaps the only one who could is Guido or his crew.

psr

@GlenH7 Flagging as inappropriate. This discussion is turning into soft pour porn.

Aaron Hall

drinks whiskey from his stackoverflow mug

user41796

@psr You ought to be able to find Dry Fly distillery products easily enough IIRC

user41796

10:05 PM

Their Washington Wheat was good enough that I'd definitely entertain trying their other products

enderland

@AaronHall that explains why Exception (RealException, 'hahahahaha') as e: works in 2.7 ;-)

Jimmy Hoffa

The sound of music - Do Re Mi

►

enderland

@amon I could always repost!

Ixrec

random: is Java the only major language with checked exceptions? (namely, exceptions that you must somehow explicitly mention in every function they might pass through)

Jimmy Hoffa

@GlenH7 interesting. I've been meaning to try small batch local stuff- Breckenridge Bourbon seemed ok in the mixed drinks I've had it, though if I recall @Shog9 decried it as terrible boot-swill...

user41796

10:10 PM

@Ixrec offhand, I want to say yes.

user41796

@JimmyHoffa I'm pretty flexible in that regards. I value the experience of learning about different types of drinks and am willing to suffer through the occasional duff bottle as a result.

user41796

For example, I'll never buy aeroplano tequila again but I don't regret having bought a bottle.

Jimmy Hoffa

@Ixrec for certain definitions of "major"- though I have a tickle in the back of my mind saying Obj-C might also...

user41796

But I'm snobby in that I don't think tequila should have oak notes like a chardonnay

user41796

@JimmyHoffa I think you're right on obj-C, but I haven't done much (any) work in that language.

user114359

10:13 PM

checked exceptions are like STDs. Once one method has them, they spread to the rest of your application like an infection that just won't go away.

Jimmy Hoffa

@GlenH7 same, I have a duff bottle of Mead that's been sitting untouched since I bought it and tried it on new years. Elderflower and Hibbiscus and Malt used in it's making. The malt really made it quite too awkward a flavor, c'est la vie.

Ixrec

@Snowman though I have never used them myself this is the impression I have of them, and I'm glad they aren't in any of the languages I use

user41796

@JimmyHoffa Time to turn it into sangria?

Aaron Hall

3 down 4 up

Jimmy Hoffa

@GlenH7 time to throw it away. Opened over 3 months ago now... yeah.

Aaron Hall

10:16 PM

that's why I don't usually answer new questions

I'm not the fastest gun in the west.

Ixrec

I'm seeing no evidence of checked exceptions in the Apple docs on ObjC, so maybe that's a no

Jimmy Hoffa

@Ixrec yeah, mayhaps I'm thinking of something else

amon

I still believe checked exceptions are a mostly good idea, and that a unified exception hierarchy is a good idea. I do however notice that less-than-stellar programmers will invariable drift to the root exception type for the method annotations which defeats the purpose of both ideas. The only solution would be throws-annotation inference. Or (my preferred approach) no exceptional control flow at all, just returning Either types.

user41796

@JimmyHoffa Yeah, pitch it if it's too far oxidized for even sangria

Jimmy Hoffa

@amon to be fair, I think Haskell has checked exceptions the right way. Frankly, I think ADT's are just bloody awesome and every language should have them, but they're so incompatible with a liskov subtyping type system

Ixrec

10:20 PM

I still don't understand the motivation behind checked exceptions tbh; they're already not ignored by default when they get thrown, and having to maintain an explicit list of all possible exceptions sounds to me like a lot of wasted bookkeeping effort since 90% of the time you can't do much with the exception besides log it

Jimmy Hoffa

therein lies the rub, putting both type systems in the same language causes too many weird things that people have difficulty with. And nobody's familiar with the HM type system so it's kinda DOA so much.

@Ixrec it's about compile time enforcement, not runtime enforcement

amon

@JimmyHoffa Haskell has unchecked exceptions – I can error () wherever I want without having to change the function type. ADTs don't rule out subtyping.

Ixrec

but compile-time enforcement of what exactly? is having an up to date list of all exception types which could ever reach my main() function somehow really valuable?

Jimmy Hoffa

@amon you're not wrong - but to be certain Either is indistinguishable from a checked exception. As for subtyping, composition isn't really the same thing - nor is multiple interfaces (which multiple type classes are)

Ixrec

monads like Maybe and Either are totally fine

enderland

10:22 PM

... it's spreading. first it was only Jimmy going on about monads. then it was amon too. now even Ixrec has adopted it into his vernacular :o

Ixrec

explicit nullability is great

amon

checked exception are a kind of type-checked documentation. It can be really challenging to figure out all the important error modes of a function, just being told what they are is extremely helpful. The problem lies in the law of demeter violation: my function signature inherits all throws annotations from the used functions, which leaks an implementation detail.

Jimmy Hoffa

How do you make an ADT sub-type another such that it automatically receives every piece composed to that other thing composed to it - without just composing that other thing. A x B = C<-- this is subtyping, taking 2 things to create a unique distinct thing. ADT's a sum types. A + B = AB <-- the B is composed to the A, the B hasn't been taken away

Ixrec

@amon "leaking implementation details" is a much more elegant phrasing of my complaint

but what would it mean to have a checked exception which does not "leak" those details? that problem seemed to me like it was a fundamental part of the idea

unless we've diluted the term enough that it now includes things like Either

amon

@Ixrec Java's specific design of checked exceptions has significant problems, but they are a great tool for API design. They can be useful if I decide to only declare exceptions of my own semantically useful exception hierarchy rather than passing through the exceptions of whatever library I'm using.

Unfortunately, this takes tremendous effort. While I like creating convenient and correct libraries, this feature is in the way for most people. Common Java wisdom is to derive all your exceptions from RuntimeException which doesn't have to be declared, and just be done with it.

user114359

10:31 PM

This almost sounds like a good question for the main site, if it weren't POB: do checked exceptions leak implementation details, violating LoD?

user114359

Or at the very least, it is a yes/no question

Ixrec

my question is slightly different albeit even more POB: Is it possible to have a form of checked exceptions that are actually useful without infecting all my callers with bookkeeping obligations?

user114359

11

Q: Why is "Is it possible to..." a poorly worded question?

A lot of questions try to cast a net into sea of possible designs with the wording "is it possible to...". These questions often get closed for one reason or another. What steps should I take to try to ask a better question?

discussion closed-questions faq-proposed

Ixrec

perhaps if you were only required to declare those exceptions which originate within the current package

user114359

I think I have a way to word the question to be on-topic and not POB/broad

Aaron Hall

10:35 PM

Asking "Is it possible" means you don't know if it is possible, which means you probably do not know enough to ask a question at the level that we expect.

BarryTheHatchet

whether*

@AaronHall Is it possible to write an on-topic question on Programmers.SE?

user114359

@BarryTheHatchet there are at least three questions per day we don't close

BarryTheHatchet

:D

lol

Aaron Hall

On Stackoverflow, asking "How do I" allows us to better tell your competence and properly allocate effort.

On Programmers, we expect higher level questions. I have not achieved zen for Programmers yet.

amon

@Ixrec yes, if we introduce the notion of functions that are transparent to throws annotations (should be the default), have a throws annotation interference system, and properly integrate the annotations into the type system so that I can consider them in generics etc.

BarryTheHatchet

10:40 PM

now that is elitist

Ixrec

@amon I'm struggling to work out what half of that means

amon

The important part is making these annotations optional, but still available. A compiler warning to urge you to spell out a throws annotation is more valuable than forcing you to do so, because most people will try to ignore them (which is OK).

But even if I don't have to declare the exceptions, a method will still throw a set of exceptions. Fortunately, this set can be inferred by the compiler.

psr

@amon unless the method is open for sub-typing, and sub-types can throw other exceptions

amon

The point about the type system means that currently lambdas and function interfaces like Function<>, Consumer<>, Supplier<> all assume no throw. They literally cannot be composed with methods that have a throws annotation. But if they could take a throws annotation as generics parameter, this would change.

user55340

@RobertHarvey you just love talking civility about bounties offered in alien languages I bet.

Ixrec

10:47 PM

tbh I still don't get why we need complex type hierarchies of exceptions in the first place; in the code I'm normally maintaining pretty much every exception indicates a programming bug that simply has to be fixed and has no sensible recovery action even if I bothered to catch it

so all my effort goes into giving the exceptions sufficiently detailed messages that I don't waste my future self's time

user114359

@Ixrec because the type of the exception indicates the error. It makes it easy to differentiate e.g. a null pointer and a divide by zero. This tells you what broke, which is even more important if the line that broke has multiple expressions.

amon

@psr of course throws annotations must obey Liskov. IIRC subtyping doesn't make inference impossible, just more costly (see also my complaints about Scala compilation speed…)

user114359

What would a generic exception tell you about float z = 3 / object.getValue()? Was object null? Did getValue() return 0.0f? This is why we have an exception hierarchy.

psr

@amon Impossible if you need to dynamically link outside your compilation unit.

Ixrec

@Snowman so...the type of the exception is valuable primarily because it'll show up in the error message? (I'm cool with that, though obviously that doesn't need to be a type to get that effect)

psr

10:50 PM

Well, impossible until run-time I guess.

amon

@Snowman having exception types is only useful if my control flow depends on the exception type. Reasonable for business exceptions, not for stuff like null pointers. The error message which can detail the exact reason is far more valuable for debugging

@psr good point, thank you.

Ixrec

@Snowman in principle there's no reason an "untyped"/generic exception couldn't have a competent error message like "Tried to divide 3 by 0" or "Could not invoke getValue() method of undefined" (and those are pretty close to what JS actually does throw)

amon

@psr actually, that's irrelevant. Changing the checked exception breaks binary compatibility, just like widening the return type of a method.

aagh, now I see how that won't work with inference. Reality sucks.

Ixrec

it may be that I simply don't have any "business exceptions" in my app, since I don't write code for payment systems or network synchronization or any other messy error-prone operations (at least, error-prone in that sense that they depend on lots of arbitrary rules or 3rd party systems)

Transcript for