[Glass] Is there a way to print a Float with all decimals but without exponential notation?

Hi guys.

I am exporting some data to a CSV file and I am finding some differences with Pharo. For my use-case, I would like to:

- Print all decimals of the float

- do not round

- do not show exponential notation

#greaseString (or doing self printOn: strm base: 10.) rounds the string. For example:

19896.800000000003 greaseString 

'19896.8'

while in Pharo I get

--
Mariano
http://marianopeck.wordpress.com

On 03/03/2014 05:46 AM, Mariano Martinez Peck wrote:
> Hi guys.
>
> I am exporting some data to a CSV file and I am finding some differences
> with Pharo. For my use-case, I would like to:
>
> - Print all decimals of the float
> - do not round
> - do not show exponential notation
>
> #greaseString (or doing self printOn: strm base: 10.) rounds the string.
> For example:
>
> 19896.800000000003 greaseString

> -> '19896.8'
>
> while in Pharo I get (what I want)
>
> 19896.800000000003 greaseString
> -> '19896.800000000003'
>
>
> #asStringUsingFormat: allows me to NOT use exponential notion..but I am
> forced to define the first and second parameter as well :(
>
> #asString shows me all decimals but it uses the exponential notion:
>
> 19896.800000000003 asString
> -> '1.9896800000000003E+04'
>
>
> So what I need is a kind of #asString but not using exponential notion.
> Is that possible?

In order to unambiguously identify an arbitrary 64-bit Float, you need
17 significant decimal digits. So it's a bit clunky, but this should
work for most cases:

| num |
num := 19896.800000000003.
^num asStringUsingFormat: {0. 17 - (num truncated printString size). false}

--> '19896.800000000003'

This assumes that your number is >= 1.0. If your number is ever
negative, or if it's between 0 and 1, the calculation of the number of
digits to the right of the decimal place would need a bit more
complexity, since the leading character would be $- or $0, neither of
which count as a "significant decimal digit".

Hi Martin,

Yes, that looks a bit clunky and probably slow? I am writing reports
where I print lots of floats, so this may be important.
Also, yes, my numbers can be negative and even between 0 and 1.0. Also,
not only Float but SmallDouble as well....

Wish there could be a simple method (primitive) to print all type of
64-bits floats without the exponential notion...

For the moment I guess I will stay with the exponential :(

Hi Mariano,

Clunky it definitely is. Though the *real* algorithm would be a lot more complicated.

But how slow is slow? How many is "lots" and how often do you need to print them? When it comes to performance, assumptions are very risky. I recommend trying it and testing what the performance is, rather than assuming it will be too slow. You can probably do that kind of test in half an hour or so.

SmallDouble and Float should be completely polymorphic (at least plenty polymorphic for this job), so you shouldn't have to worry about the difference. The handling of negative and 0 - 1 won't add much complexity.