Re: [gentoo-dev] [PATCH] eapi7-ver.eclass: 'Early adopter' version of EAPI 7 version manip

[Kent Fredric]

On Fri,  8 Sep 2017 13:19:23 +0200
Michał Górny <mgorny@g.o> wrote:

> a. getting wider review and some real-life testing before
> the specification is set in stone, and

Any thoughts on a function that would represent a dotted-decimal style version
as a floating point string?

eg: 

  ver_float 0.1.0    -> 0.001
  ver_float 0.10.0   -> 0.010
  ver_float 0.100.0  -> 0.100

That's of course *the most* generic example I can offer, but seeing
this sort of transformation is commonly needed in the world of perl, I
thought maybe now is a good time to mention something.

Sadly, its just the sort of idea that if done wrong, would be no use.

For instance, sometimes you want:

   ver_float 0.1.0  -> 0.0010

Or

   ver_float 0.1.1  -> 0.001001

The two key things here is to know: 

1. How many digits each position represents
2. The maximum number of digits to represent

So, some ideas in that regard are:

   ver_float ${INPUT} ${PRECISION}  

Where the values per position are fixed, so:

    ver_float 0.1 3   -> 0.001
    ver_float 0.1 2   -> INVALD  # fidelity loss by truncation
    ver_float 0.10 2  -> 0.01    # permitted because there's no fidelity loss by truncation
    ver_float 0.100 1 -> 0.1     # permitted because there's no fidelity loss by truncation
    ver_float 0.100 2 -> 0.10
    ver_float 0.100 3 -> 0.100
    ver_float 0.101 1 -> INVALID # fidelity loss by truncation

    ver_float 0.1 5   -> 0.00100
    ver_float 0.1.1 5 -> INVALID, need 6 digits to represent 0.1.1 

    ver_float 0.1.1 6 -> 0.001001


Or say, 

    ver_float ${INPUT} ${PATTERN}

Where "pattern" is a string like 3-3-3 indicating how to map digits to numbers

     ver_float 0.1      3    -> 0.001
     ver_float 0.1      2    -> 0.01 
     ver_float 0.1      1    -> 0.1
     ver_float 0.1.1    3    -> # INVALID, no map for '.1'
     ver_float 0.1.1    3-3  -> 0.001001
     ver_float 0.1.10   3-3  -> 0.001010  
     ver_float 0.1.10   2-2  -> 0.0110
     ver_float 0.10.10  2-2  -> 0.1010
     ver_float 0.100.10 2-2  -> # INVALID, can't map "100" into 2 characters

Though I suspect you'd want both features ...

     ver_float ${INPUT} ${PATTERN} ${TRUNCATION_LENGTH}

Because we do need packages where 

     0.123.10 means 0.1231 

Either way, much of this is probably a time wasting bad idea.

But I thought I'd just 

( •_•)

( •_•)>⌐■-■

(⌐■_■)

Float it by you.