perl --do-not-sometimes-use-scientific-notation-ever-even-for-really-small-numbers

[jidanni]

Or call this plea for a new flag or variable:
$ perl --use-one-format-when-printing-numbers-that-doesn't change at some arbitrary point.

https://stackoverflow.com/questions/75773212/where-in-the-perl-documentation-does-it-mention-0-0001-0-00001-printing-differe
https://stackoverflow.com/questions/75797326/how-to-force-perl-to-print-small-numbers-in-the-same-format-as-other-numbers
https://stackoverflow.com/questions/75785937/workarounds-to-print-00001-in-perl
Thanks!

[iabyn]

On Fri, Mar 24, 2023 at 01:39:44AM -0700, 積丹尼 Dan Jacobson wrote: Or call this plea for a new flag or variable: $ perl --use-one-format-when-printing-numbers-that-doesn't change at some arbitrary point.

So, what exact behaviour would you like for this flag/whatever when perl converts a num to a string in the absence of a format specifier? Would you expect this flag/whatever to be global in behaviour, lexical scoped, or dynamically scoped, or what?

…

-- "I do not resent criticism, even when, for the sake of emphasis, it parts for the time with reality". -- Winston Churchill, House of Commons, 22nd Jan 1941.

[sisyphus]

Yeah - I sympathize.
I can't find the documentation that covers this - it may not exist ... I don't know,.
But it seems that scientific notation kicks in whenever the "non-scientific notation" contains only zeros as the first 5 (or more) leading digits:

D:\>perl -e "print 1e-4;"
0.0001
D:\>perl -le "print 1e-5;"
1e-05

It's the same in python3, BTW:

$ python3 -c "print(1e-4)"
0.0001
$ python3 -c "print(1e-5)"
1e-05

Is there some standard that specifies what a function named print should do when it prints out a floating point value ?

The mpfr library defaults to printing out values using only scientific notation, and such that they "survive the round trip". Consequently my Math::MPFR module's overloading of print() should do the same, but Perl itself is not guaranteed to do either.

Python3 and raku (I think) follow the same rule as perl wrt when "scientific notation" kicks in but they always produces values such that "the round trip is survived" using the minimum number of significant digits.
Perl's print() function no longer cares about "survival of the round trip" - though there used to be a (now discontinued) special perlvar that could be set to cater for that .

I would like to see perl follow the same rules as python3 (and raku) - and I've written a Math::Ryu module that does that.
But, IIUC, even that is not what you're seeking - because it switches form "non-scientific notation" to scientific notation at 1e-5.

It's all so much more messy than it ought to be, IMO :-(

Cheers,
Rob

PS If you don't know what I mean by "surviving the round triip", it's basically the notion that, (in perl)"$x" == $x is always true (except for NaN values).
It's not always the case - please ask if I need to elaborate.

[jidanni]

I just want a way to force perl to print numbers in a consistent way...
that doesn't get weird near zero. Just use the same format that it uses
for all other numbers... and also use that format when near zero.

Maybe even some environment variable, command line switch, whatever, anything.

[sisyphus]

I just want a way to force perl to print numbers in a consistent way

I think you'll need to write your own function.
(Put that function in a module so that you can access it in your scripts by simply loading that module.

For something that will produce the same values as a perl but in a consistent format I can use Math::MPFR's Rmpfr_out_str() function.
It seems like overkill but, TBH, I can't think of another straightforward way of doing it.

D:\>perl -MMath::MPFR=":mpfr" -le "Rmpfr_out_str(Math::MPFR->new(0.1), 10, 15, 0);"
1.00000000000000e-1
D:\>perl -MMath::MPFR=":mpfr" -le "Rmpfr_out_str(Math::MPFR->new(0.01), 10, 15, 0);"
1.00000000000000e-2
D:\>perl -MMath::MPFR=":mpfr" -le "Rmpfr_out_str(Math::MPFR->new(0.014), 10, 15, 0);"
1.40000000000000e-2
D:\>perl -MMath::MPFR=":mpfr" -le "Rmpfr_out_str(Math::MPFR->new(sqrt 2), 10, 15, 0);"
1.41421356237310e0
D:\>perl -MMath::MPFR=":mpfr" -le "Rmpfr_out_str(Math::MPFR->new(sqrt 3), 10, 15, 0);"
1.73205080756888e0
D:\>perl -MMath::MPFR=":mpfr" -le "Rmpfr_out_str(Math::MPFR->new(2 ** -1074), 10, 15, 0);"
4.94065645841247e-324

All of those values are exactly equivalent to the values that perl's print() outputs when nvtype is double.
And, unlike those values provided by perl, they are all formatted identically.

If you like to post a question about this on perlmonks, I'm happy to work through it with you. (This is not the place for that.)
Perhaps stack overflow is also a good place to ask. (I don't follow that forum.)

Cheers,
Rob

UPDATE: Duh ... printf "%.14e", $val will, I believe, serve the purpose just as well as the use of Math::MPFR (that I demonstrated above).

[jidanni]

I guess I'll just stick with adding one and then chopping the one off and putting a zero there!
https://stackoverflow.com/questions/75797326/how-to-force-perl-to-print-small-numbers-in-the-same-format-as-other-numbers/75831019#75831019

I wonder how it got into this mess in the first place. I'm imagining the grandfathers had some underlying hardware that printed things great. So then they just decided to use that. And then later the hardware changed and made this unreliable way of printing things...

[sisyphus]

https://stackoverflow.com/questions/75797326/how-to-force-perl-to-print-small-numbers-in-the-same-format-as-other-numbers/75831019#75831019

Do you really want 1e-300 to be displayed as:
0.000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001
??

Cheers,
Rob

[jidanni]

Do you really want 1e-300 to be..

As a matter of fact yes! No problem. Bring them on. I'm not scared. That's the risk with the future
--no-scientific-notation-even-if-means-more-zeros-than-in-your-bank-account flag.

Anyway, there should be a way to adjust these undocumented arbitrary limits without needing printf:

$ perl -wle 'for(-5..20){print 0+"1e$_"}'
1e-05
0.0001
0.001
0.01
0.1
1
10
100
1000
10000
100000
1000000
10000000
100000000
1000000000
10000000000
100000000000
1000000000000
10000000000000
100000000000000
1000000000000000
10000000000000000
100000000000000000
1000000000000000000
10000000000000000000
1e+20

[sisyphus]

Fair enough.
No objections from me.

I think this ticket can therefore be closed.

Cheers,
Rob