search

brick / math

Arbitrary-precision arithmetic library for PHP
MIT License
1.78k stars 75 forks source link

RFC: add `getPrecision` method. #58

Open bendavies opened 3 years ago

bendavies commented 3 years ago

Hi,

Similar to getScale, would you be on favor of adding a method getPrecision? I want to perform app side validation on BigDecimal before passing to Postgres numeric fields, and getPrecision would make that easier.

Thanks!

BenMorel commented 3 years ago

Hi, what would be the difference between getScale() and getPrecision()?

bendavies commented 3 years ago

hi @BenMorel

The PostgreSQL manual will say it better than me:

We use the following terms below: The precision of a numeric is the total count of significant digits in the whole number, that is, the number of digits to both sides of the decimal point. The scale of a numeric is the count of decimal digits in the fractional part, to the right of the decimal point. So the number 23.5141 has a precision of 6 and a scale of 4. Integers can be considered to have a scale of zero.

bendavies commented 2 years ago

@BenMorel any interest? thanks!

alexviar commented 1 year ago

@BenMorel

Working with precision is generally better than working with scale

Check this example:

(0.03/360)*8*30 = 0.02

Woking with scale of 10: 0.03/360 = 0.0000833333(3) 0.0000833333*8=0.0006666664(0) 0.0006666664*30=0.0199999920(0)

❌ 0.0199999920 == 0.02

Working with precision of 10: 0.03/360 = 0.00008333333333(3) 0.00008333333333*8=0.0006666666666(40) 0.0006666666666*30=0.01999999999(80)=0.02

✔0.02==0.02

bendavies commented 1 year ago

@alexviar i'm not sure what your example shows but it's not precision of 10

Working with precision of 10: 0.03/360 = 0.00008333333333(3)

0.00008333333333(3) has as precision of 16

alexviar commented 1 year ago

@alexviar i'm not sure what your example shows but it's not precision of 10

Working with precision of 10: 0.03/360 = 0.00008333333333(3)

0.00008333333333(3) has as precision of 16

It has a scale of 14 but a precision of 10. The precision refers to how many significant digits it has, not how many decimal places it has. https://www.logicbig.com/quick-info/programming/precision-and-scale.html

For example Big.js uses scale whereas Decimal.js uses precision

alexviar commented 1 year ago

@alexviar i'm not sure what your example shows but it's not precision of 10

Working with precision of 10: 0.03/360 = 0.00008333333333(3)

0.00008333333333(3) has as precision of 16

I did put in parentheses digits that was rounded, and underlined digits are the Periodo of decimal

bendavies commented 1 year ago

ah right thanks, i forgot that leading zeros are not significant. but also for storage purposes (postgres), 0 <= scale <= precision so, It has a scale of 14 but a precision of 10. doesn't work, in practise.

anyway, the point of this of was to simple add a getPrecision() method, similar to getScale, nothing mnore

BenMorel commented 1 year ago

Hi @bendavies, sorry for the late reply. I'm not against adding a getPrecision() method if it is useful to you, so please feel free to open a PR along with a few tests, in particular to clarify what happens for leading zeros.

@alexviar I'm not sure what the decimal period has to do with the precision? Anyway, you make me think that this would be a great addition to BigRational: a way to output 10/3 as 3.3(3) for example. Would you want to work on this?

BenMorel commented 1 year ago

I opened issue #74 to track the BigRational as decimal number with period separately.

alexviar commented 1 year ago

@alexviar I'm not sure what the decimal period has to do with the precision?

I think I didn't explain myself well. I did put rounded decimals (and decimal period) in parentheses for illustrative purposes only. Precision is given by working with n significant digits instead of n decimal place:

Operation Rounded to 10 significant digits Rounded to 10 decimal places
0.03 0.03000000000 0.0300000000
360 360.0000000 360.0000000000
8 8.000000000 8.0000000000
30 30.00000000 8.0000000000
0.03/360 0.00008333333333 0.0000833333
(0.03/360)*8 0.0006666666666 0.0006666666
(0.03/360)*8*30 0.02000000000 0.0200000000

Now, if we take the rounded values to do subsequent operations, we have:

Operation Rounded to 10 significant digits Rounded to 10 decimal places
0.03 0.03000000000 0.0300000000
360 360.0000000 360.0000000000
8 8.000000000 8.0000000000
30 30.00000000 8.0000000000
0.03/360 0.00008333333333 0.0000833333
(0.03/360)*8 0.0006666666666 0.0006666664
(0.03/360)*8*30 0.02000000000 0.0199999920

Currently it can be circunvent by using BigRational, but this workaround doesn't work with irrational opreations like square root for example (sqrt(2)*sqrt(2) Never mind, I had read an article with an example like this, but I can not found it)