• Booger (unregistered) in reply to Some Jerk
    Some Jerk:
    Whiner:
    faoileag:
    a lot of people *don't know how floating point numbers are handled by a computer*.
    But why should we have to know anything? All that geeky stuff is supposed to be abstracted away! Computers should be designed so that the proudest most ignorant lamer can do just as well as a CS grad, if not better! This is really just another failure of everybody-but-me.

    In my experience... ignorant programmers tend to do much better than CS grads... but then, I suppose I could incite a riot if I started ragging on college programmers. I'll do my best to keep my prejudices to myself.

    Insert the old bit about I've seen shit Grads and I've seen quality peole without any qualification, I've also seen the opposite.

    I'm one of the best, but I think I would have been even if I hadn't gone to Uni. That said, the piece of paper did help me land my first job....

  • milli (unregistered) in reply to Some Jerk
    Some Jerk:
    Cbuttius:
    Even when the amount is noticeable, if you are dealing with an amount like $10,000 you probably don't care about 10 cents above or below that amount so even float would be accurate enough.
    That general line of thinking has been exploited to generate tens of millions of dollars worth of deposits by capturing the rounded difference of banking transactions. Even $0.001-$0.0049 at a time managed to generate such a massive amount of money in a short period of time when acrewed over millions and millions of transactions... which can take place easily inside of a few days with regard to banks.

    There is a second problem with your theory as well. The problem that occurs with a lack of precision becomes is something like a plague. It doesn't grow incrementally, but instead proportionally. As the number of people infected with a plague increases, so does the rate at which people are infected. The same goes for code that cannot handle numbers precisely. We may be dealing with a few cents difference now, but over time a few cents difference can throw the ledger off by more than the $10,000 that you rounded up or down to.

    and anyway, the issue with using numbers for calculation isn't so much the issue of rounding as the issue of inaccurate representation
  • (cs)
    nomous:
    Should just store them ass bit fields
    nomous:
    Should just store them as bit fields OOPS...
    That one bit you on the ass, eh? <g>
    One would think that the submitter's name would be at the bottom of the receipt, but well...
    But it is! It was either "Y. Li", or his seventeenth cousin (eleven times removed), "Could not load file or assembly 'ICU4NET, Version=1.0.3891.39017, Culture=neutral, PublicKeyToken=null' or one of its dependencies. An attempt was made to load a program with an incorrect format." (That cousin must be related to Nicholas If-Jesus-Christ-Had-Not-Died-For-Thee-Thou-Hadst-Been-Damned Barebon.)
  • ManxBug (unregistered) in reply to bhtooefr
    bhtooefr:
    And in 2006, VWs didn't have trim levels, they had packages on the Jetta.

    Not familiar with the packages on the 2.0Ts, but on the TDIs, it went something like this:

    Package 0 (no packages selected): Pleather, no trip computer, CD changer with MP3, ordinary HVAC Package 1: I think added sunroof and Sirius to Package 0 Package 2: Added real leather, trip computer, climate control to Package 1

    I'd guess that Package 3 on a 2.0T would be HIDs or something like that, on top of Package 2.

    Package 3 is probably a lot of high end options. 3 driver memory seat settings Navigation HIDs (as you said)

    Basically the current Tech package.

  • Lgp (unregistered) in reply to asdf

    At least in SVR2 (System V Release 2) the "You're not here. Go away!" message occurred when you tried to run a command needing access to another terminal (e.g., "talk") while you had no entry in the wtmp file. Which could happen, of course, if you happened to be on a system being developer as a secure UNIX system which didn't allow writing to that file. Which might also have been an Apple-based system. Ahem.

  • Craig (unregistered)

    I've gotten the "You don't exist, go away!" message many times over the years on Linux systems. A foolproof way seems to be to su to an account that was never meant to be logged into.

  • Cheong (unregistered)

    "If there was only an option between 5 hours and 136.102 years this would be perfect!" writes Alan.

    I would be more impressed if this option is shown on the battery side, which implies the battery can power the laptop without running out for 136.102 years.

    I swear if such battery exist, I'd urge my UPS provider to use it, and then replace all my UPS with the new model.

  • iMalc (unregistered)

    But if I don't exist, then how can I go away?!

  • Dave (unregistered) in reply to asdf
    asdf:
    "You Don't Exist. Go Away!" is much older than that, login command from linux's [...].

    Sheesh, first you say "much older" and then you mention Linux... this dates to at least Unix Version 7 from 1979 (at which point Linus was nine years old), and possibly before that.

  • Should Be Obvious (unregistered)

    The credit card slip is the "Customer Copy" or "Duplicate". The total was written by the customer for the customer's own records. The actual "Store Copy" likely has the tip amount and the signature and is probably with the store.

  • YBM (unregistered)

    I got such message (you don't exit go away) on SunOS 20 years ago when NSS (Name Service Switch) were broken in some way, I could authenticate anyway (so PAM or whatever SunOS used was ok) but almost any command run from the Shell complained.

    I guess you could get it on Linux/BSD/Modern UNIXES quite easily when messing with /etc/nsswitch.conf

  • dogmatic (unregistered)

    You don't know what an options package is? Maybe you should have someone else help with selling/buying a car.

  • Brendan (unregistered) in reply to Remy Porter
    Remy Porter:
    I used to train programmers, and my first lesson was: "I don't care what you call it- if you aren't going to use it in mathematical expressions, it isn't a number. PINs, phone numbers, zip codes, SSNs, etc.- they are all text."

    And then people would go and hand me programs that treated each one of them as an integer.

    Computers are incapable of storing "text". The closest options are bitmap pictures, vector data (e.g. describing pen movements), or (more common) a series of character identifiers.

    Are you sure you didn't want these PINs, phone numbers, zip codes, etc stored as "a series of (integer) character identifiers"?

    If yes, did you specify which standard the series of (integer) character identifiers should conform to (e.g. ASCII, EBCDIC, UTF-8, etc)?

    If no, then you didn't preclude storing the series of (integer) character identifiers in a compressed format. Compressed formats include storing "digit only" data as BCD, or using an even more efficient compression method like "compressed_identifier_sequence = digit1 * 1000 + digit2 * 100 + digit3 * 10 + digit4 * 1".

    Please don't blame your students when you make the mistake of thinking their "compressed series of character identifiers" is a number when it's not.

  • Michael (unregistered)

    I thought – everyone should be able to see such an awesome error message once in a while. So here it is – your very own fatal App :–)

    http://moapp.tumblr.com/post/29285592807/fatal-the-app

  • noob (unregistered) in reply to faoileag
    A WTF, sure, but a common one. I do not want to know how many sums (of money) are stored in double
    I unterstand why not use int/double for phone or things you wont do arithmetic operations with. But whats the problem with using double for money?
  • (cs) in reply to noob
    noob:
    A WTF, sure, but a common one. I do not want to know how many sums (of money) are stored in double
    I unterstand why not use int/double for phone or things you wont do arithmetic operations with. But whats the problem with using double for money?

    The "issue" of using a double for money is generally based on the fact that amounts are stored with fractional parts, often in hundreds, which cannot be represented exactly by a double.

    So if an amount is say $1.30 then because in reality it is stored in hexadecimal, there will be rounding errors.

    In reality you get about 15 significant figures of precision which means only when you are dealing with amounts in trillions will the pennies start to become insignificant.

    The issue I raised is that money is not calculated simply by adding and subtracting amounts in pennies but there are all kinds of compound interest being applied to money held in accounts, as well as tax deductions to be paid, which causes amounts to be rounded anyway.

    The issue only normally gets noticed when one tries to print the amount as a string and doesn't set a format string properly (formatted printing seems to have been managed rather badly since people had issues with printf. Yes, there were issues with printf but the solutions often lost the plot a bit). Anyway if you print the double representation of 1.30 "naturally" you will possibly get something that looks like 1.299999999975 which looks silly.

    It is a bigger issue if you multiply the amount by 100 then convert to an int as you could well get 129 through failure to adjust for rounding, but that would be your error, rather than initially storing the amount as a double.

  • Mike Dimmick (unregistered) in reply to noob
    noob:
    A WTF, sure, but a common one. I do not want to know how many sums (of money) are stored in double
    I unterstand why not use int/double for phone or things you wont do arithmetic operations with. But whats the problem with using double for money?

    'double' is binary floating point. That means it can only store numbers that are a sum of negative powers of two - one half, one quarter, one eighth, one sixteenth and so on to the number of bits that can be stored in the 'mantissa' part of the variable.

    Unfortunately one-tenth is not expressible precisely as any finite sum of negative powers of two. You can get close approximations but they're always a little off. As the quantities before the point get larger, the amount of precision left to represent the decimal part gets smaller and the representation gets further away from the correct value. The only 100th amounts that are accurately representible in binary floating point are 0.25, 0.50 and 0.75.

    If it helps, think of how one-third or one-seventh are represented in decimal notation. We have to add 'recurring' markers to indicate that we actually need infinite precision to represent these values.

  • Some Damn Yank (unregistered) in reply to Remy Porter
    Remy Porter:
    I used to train programmers, and my first lesson was: "I don't care what you call it- if you aren't going to use it in mathematical expressions, it isn't a number. PINs, phone numbers, zip codes, SSNs, etc.- they are all text."

    And then people would go and hand me programs that treated each one of them as an integer.

    I'm on a project right now to fix our SSN fields that suppress leading zeros (ex: 034-45-1234). That's right, they're stored as numbers, and by default we display numbers with suppressed leading zeros.

  • (cs)

    People where I work will constantly open perfectly good CSV data files in Excel and save them out from there, in the process mangling account numbers, phone numbers, package tracking numbers, etc., into exponential notation, and stripping leading zeroes from zip codes.

  • (cs) in reply to Remy Porter
    Remy Porter:
    I used to train programmers, and my first lesson was: "I don't care what you call it- if you aren't going to use it in mathematical expressions, it isn't a number. PINs, phone numbers, zip codes, SSNs, etc.- they are all text.
    Dex, the Yellow Pages for 14 states, stored their phone numbers in three numbered types (can't remember which kind though). One for each segment. Doing so DOES guarantee that you're not storing a non-numeric value. And, no, that wasn't their downfall. Mismanagement was.
  • (cs) in reply to Gurth
    Gurth:
    One of the WTFs here is not pressing the spacebar after ⇧⌘4.
    Or using the utility Grab (but your/my method is faster)
  • Merlin (unregistered) in reply to Bob
    Bob:
    Who the hell leaves an $8.02 tip?
    This is the comment I was looking for. It was the tip amount that made me say WTF, the rest it just the normal fair around here.
  • Ol' Bob (unregistered) in reply to Some Jerk
    Some Jerk:
    Larry:
    Bill P. Godfrey:
    Don't exist? Try thinking.
    I did once, but damn, that is so fucking painful. There must be another way.

    fart... and if you can smell it, then you can be reasonably certain that whomever caused the smell did in fact exist.

    Ah...but that only indicates that your internal non-beingness is in fact quite well integrated. Only years and years of very expensive therapy will enable you to understand, at last, that you do not exist, and that your belief in the smelliness of your own farts is in fact merely a manefestation of the internal consistenciness of your delusions of existence. If that's too complicated to understand, let's try a simple thought experiment. Be the ball. Stop thinking. Let things happen. Just...be...the ball. And remember - the Zen philosopher Basho once wrote "A flute with no holes is not a flute, and a donut with no hole is a Danish". He's a funny guy...

    (CAPTCHA: duis - pidgin Latin for "be the ball")

  • Bugmenot (unregistered) in reply to asdf

    The error in git comes from SSH. Surely that's also where the error in TFA comes from.

    Your theory about syslog makes no sense.

  • Richard A. O'Keefe (unregistered) in reply to Remy Porter

    One of my first jobs was to clean up a lot of RSTS/E Basic someone else had left behind. He stored part numbers as single precision floats. Nine-digit part numbers.

  • Phil Srobeighn (unregistered)

    So it's my first day on this site and I run across this. I wonder if Neil is going to dispute his Verizon bill this month with his original bill for the period?

Leave a comment on “You Don't Exist. Go Away!”

Log In or post as a guest

Replying to comment #:

Log Out

« Return to Article