https://medium.com/@bellmar/is-cobol-holding-you-hostage-with-math-5498c0eb428b
Face it: nobody likes fractions, not even computers.
When we talk about COBOL the first question on everyone's mind is always Why are we still using it in so many critical places? Banks are still running COBOL, close to 7% of the GDP is dependent on COBOL in the form of payments from the Centers for Medicare & Medicaid Services, The IRS famously still uses COBOL, airlines still use COBOL (Adam Fletcher dropped my favorite fun fact on this topic in his Systems We Love talk: the reservation number on your ticket used to be just a pointer), lots of critical infrastructure both in the private and public sector still runs on COBOL.
Why?
The traditional answer is deeply cynical. Organizations are lazy, incompetent, stupid. They are cheap: unwilling to invest the money needed upfront to rewrite the whole system in something modern. Overall we assume that the reason so much of civil society runs on COBOL is a combination of inertia and shortsightedness. And certainly there is a little truth there. Rewriting a mass of spaghetti code is no small task. It is expensive. It is difficult. And if the existing software seems to be working fine there might be little incentive to invest in the project.
But back when I was working with the IRS the old COBOL developers used to tell me: "We tried to rewrite the code in Java and Java couldn't do the calculations right."
[Ed note: The referenced article is extremely readable and clearly explains the differences between floating-point and fixed-point math, as well as providing an example and explanation that clearly shows the tradeoffs.]
(Score: 1) by nitehawk214 on Monday September 16 2019, @05:16PM (16 children)
You do realize that C has floating point datatypes and suffer the same flaws, right? And that no precise calculations should use them, right?
"Don't you ever miss the days when you used to be nostalgic?" -Loiosh
(Score: 2) by FatPhil on Monday September 16 2019, @05:52PM (13 children)
Have you ever read any Kahan?
So if the penaly fees were 300*1.27^(17/365), how much *exactly* do I owe?
No, exactly, not that approximation.
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 2) by DannyB on Monday September 16 2019, @06:17PM (3 children)
Unrelated to financial calculations. Please elaborate on that. I tend to think of standardized FP implementations are giving the same results.
I kind of thought that the whole point of IEEE standards were binary interchangeable formats, and reproducible results. Something sadly lacking once long ago when ever language, indeed every implementation of every compiler had its own weird FP implementation.
Thanks.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 2) by FatPhil on Tuesday September 17 2019, @07:47AM (2 children)
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 2) by DannyB on Tuesday September 17 2019, @02:12PM (1 child)
Can you kindly point me to that link please?
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 2) by DannyB on Tuesday September 17 2019, @02:12PM
Nevermind, I believe I found it.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 2) by DannyB on Monday September 16 2019, @06:44PM (8 children)
Obviously not. Can you elaborate on what Kahan is?
Googling for Java vs C floating point. I find that Java FP, depends on platform implementation, and a runtime option. Based on these, Java may do intermediate> arithmetic of 32-bit floats, using higher precision before converting back to 32-bit floats. If C / C++ don't do this, that could account for some discrepancy.
I don't find a whole lot of relevant anything that pops up in Google. So I'm sincerely interested in what Java gets wrong in floating point.
I do notice a number of examples of values that are not exactly representable in floating point. The famous 0.1 is but one of many examples. In fact, I suspect there are a huge number of exact decimal numbers not exactly representable in binary, and vice versa.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 2, Informative) by Anonymous Coward on Monday September 16 2019, @07:25PM (1 child)
Not vice versa. Any fraction that can be represented in binary can be represented in decimal.
The numbers that have terminating representation in binary are those that can be expressed as the sum of fractions where the denominator is a power of two. e.g. 0.140625 is the sum of 1/8 and 1/64. Or 5/16 + 9/128 + 23/256 = 0.47265625.
The numbers that have terminating representation in decimal on the other hand are those that can be expressed as the sum of fractions where the denominator is a power of two times a power of five (the factors of ten). So there are a lot of denominators that are available in base ten but not base two.
Of course rational numbers are rational, and irrational numbers are irrational, regardless (or irregardless ;) ) of the radix used. The numbers that don't have terminating representations have repeating representations instead. You can see this in the simple fractions that have repeating representations in decimal : they're the ones where the denominator is relatively prime to 2^n*5^m. 1/3, 1/6, 1/7, 1/9...
(Score: 3, Informative) by DannyB on Monday September 16 2019, @08:02PM
Thank you. That makes sense.
Terminating binary denominator must be power of 2.
Terminating decimal denominator must be power of 2 times power of 5.
This is excellent point to refute another reply that I trotted out the often used example of 0.1 not having an exact binary representation.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 3, Informative) by FatPhil on Tuesday September 17 2019, @07:20AM (5 children)
Kahan is one of the guys who wrote the IEEE754 floating point standards. Hasn't been a research academic for a long time, but over the decades has published a whole bunch of reports on what can go wrong with FP, including at the language-implementation level, such as How JAVA's Floating-Point Hurts Everyone Everywhere [berkeley.edu] (PDF file) whose title should be self-explanatory.
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 2) by DannyB on Tuesday September 17 2019, @02:23PM (4 children)
That is interesting.
It is from 1998, over twenty years ago. Java has changed a lot -- but I don't know how much in relation to this topic.
There are a few rebuttals I could make -- but, I don't use floating point much in Java. But I've never had a problem with it.
Since I mostly work with BigDecimal (which is really an unlimited precision integer underneath), I don't have the inexactness of floating point approximations of money values.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 2) by FatPhil on Tuesday September 17 2019, @02:51PM (3 children)
Having a specific type for currency values is ideal (as long as you can trust the implementer of that black box). It can also help with dimensional correctness too, if you have strict typing.
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 3, Interesting) by DannyB on Tuesday September 17 2019, @03:32PM (2 children)
Microsoft, amusingly, has a type for currency. It is basically your standard 2s compliment 64 bit integer with an implied decimal point four places from the right. So it counts in hundredths or a cent, or ten-thousandths of a dollar -- or whatever unit of currency in use.
I would like to see more languages that can attach dimensional correctness to scalar values, which in turn can be carried into more complex types such as rational values, complex numbers, etc. It would be a compile time error to have incorrect dimensions.
But when I muse about higher level and more abstract languages, I am often taken to task for it because programming should always be only about working at the bits and bytes and cpu cycles.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 2) by FatPhil on Wednesday September 18 2019, @08:19AM (1 child)
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 2) by DannyB on Wednesday September 18 2019, @01:58PM
Microsoft has had its currency type since at least the 1990's. It is in all their languages, as far as I know, but started in VB, I think. At the time it probably seemed like a good idea. And it probably stopped many programmers from using floating point instead.
People today are educated enough to repeat what they are taught but not to question what they are taught.
(Score: 3, Insightful) by iWantToKeepAnon on Tuesday September 17 2019, @02:36PM (1 child)
Nostalgia isn't what it used to be. :/
"Happy families are all alike; every unhappy family is unhappy in its own way." -- Anna Karenina by Leo Tolstoy
(Score: 1) by nitehawk214 on Wednesday September 18 2019, @02:27PM
Wtf is a happy family? Something that only happens in books? :/
"Don't you ever miss the days when you used to be nostalgic?" -Loiosh