strange sprintf behavior?

oseeliger has asked for the wisdom of the Perl Monks concerning the following question:

Hi,

I'm pretty confused about these 4 statements ( or I'm just a blockhead in understanding the evaluation order / precedence of operators and parentheses ).

The first and third statement print a simple arithmetic opertation (using differently placed parentheses) in a float value interpreted sprintf instruction. That is ok.

But why does the second and fourth statement print different rounded values? I have no clue and don't know how to find out.

Please, could someone advise?

perl -e 'print sprintf( "%f",     66.5 * ( ( 100 + 19 ) / 100 ) ), "\n
+79.135000"'
perl -e 'print sprintf( "%.2f",   66.5 * ( ( 100 + 19 ) / 100 ) ), " <
+-- wrong, expected 79.14\n"'

perl -e 'print sprintf( "%f",     66.5 * (   100 + 19 ) / 100   ), "\n
+79.135000"'
perl -e 'print sprintf( "%.2f",   66.5 * (   100 + 19 ) / 100   ), " <
+-- ok = 79.14\n"'
[download]

Comment on strange sprintf behavior? Download Code

Replies are listed 'Best First'.
Re: strange sprintf behavior? by Anonymous Monk on Oct 15, 2010 at 12:59 UTC
`print sprintf( "%f", 66.5 * ( ( 100 + 19 ) / 100 ) ), "\n79.135000 +\n"; print sprintf( "%.2f", 66.5 * ( ( 100 + 19 ) / 100 ) ), " <-- wrong, + expected 79.14\n"; print sprintf( "%f", 66.5 * ( 100 + 19 ) / 100 ), "\n79.135000 +\n"; print sprintf( "%.2f", 66.5 * ( 100 + 19 ) / 100 ), " <-- ok = 7 +9.14\n"; __END__ 79.135000 79.135000 79.13 <-- wrong, expected 79.14 79.135000 79.135000 79.14 <-- ok = 79.14` [download] What Every Computer Scientist Should Know About Floating-Point Arithmetic Why does printf() do this to big numbers? printf question sprintf rounding convention sprintf values	[reply] [d/l]
Re: strange sprintf behavior? by TomDLux (Vicar) on Oct 15, 2010 at 14:27 UTC
Your decimal brain tells you something divided by a hundred can be exactly represented, but in binary, 1/5, 1/10, 1/100 are infinitely repeating fractions. As Occam said: Entia non sunt multiplicanda praeter necessitatem.	[reply]
Re: strange sprintf behavior? by ig (Vicar) on Oct 15, 2010 at 16:01 UTC
`print sprintf( "%.30f", 66.5 * ( ( 100 + 19 ) / 100 ) ), "\n79.135 +000"; print sprintf( "%.2f", 66.5 * ( ( 100 + 19 ) / 100 ) ), " <-- wrong, + expected 79.14\n"; ; print sprintf( "%.30f", 66.5 * ( 100 + 19 ) / 100 ), "\n79.135 +000"; print sprintf( "%.2f", 66.5 * ( 100 + 19 ) / 100 ), " <-- ok = 7 +9.14\n";` [download] produces: `79.134999999999990905052982270718 79.13500079.13 <-- wrong, expected 79.14 79.135000000000005115907697472721 79.13500079.14 <-- ok = 79.14` [download] Which may make the difference more obvious. update: deleted incorrect example about integer Vs floating point calculations.	[reply] [d/l] [select]
Re^2: strange sprintf behavior? by oseeliger (Initiate) on Oct 18, 2010 at 06:19 UTC
Thank you all very much. It's now clearer and I think I understood it. The links are very informative.	[reply]