Concerning FORTRAN, I think it was removed for a long time, it was just an example of how the NumberParser could be used.

Concerning the ExtendedNumberParser, I find it useful to have it, because it can parse inputs like:
+1.4

-1.2e+3
.005

127.e6

A lot of languages/applications around the world would accept these as valid Number.

Smalltalk syntax is much stricter...

OK, it's just a DSL, and it could as well be implemented by other means...

Whether such DSL implementation belongs to NumberParser hierarchy or not, is a choice that was guided by 2 things

- NumberParser provides an implementation that will correctly round the number to nearest Float (or ScaledDecimal eventually)

� I've seen too many Smalltalk implementations departing from universal standards with this respect (IOW, wrong)

- NumerParser tries to be efficient (as far as Smalltalk can be), especially in the cases with many digits

The first point is mandatory.

The second is just an optimization.

Now that you have more insight about rationale which guided those decisions, it's up to you :)

Cheers

Nicolas

2013/10/28 Marcus Denker <marcus.denker@inria.fr>

On 28 Oct 2013, at 17:55, Yuriy Tymchuk <yuriy.tymchuk@me.com> wrote:

On 28 Oct 2013, at 17:49, Marcus Denker <marcus.denker@inria.fr> wrote:

On 28 Oct 2013, at 10:41, Yuriy Tymchuk <yuriy.tymchuk@me.com> wrote:

So what we decide to do with this?

I �can work on that case, but I have 2 questions:

- do we care to provide parsing of numbers in other notations as Pharo one?

That�s a good question.. I completely misunderstood the code there because that sounds a bit

strange that a language supports other languages syntax, but not �inside� the language but somehow as a tool�

I fear that even if someone needs Fotran number parsing, the last place to look would be the Pharo base kernel

classes�

There are 2 extremes here. If you want to parse a fortran number, you probably also want to parse other fortran parts, and so you have some dedicated parser.

But if you want to just parse some general number and it�s +2, then you can�t easily do that.

- what is the number syntax that we want to support?

Yes, we we need to document it. Right now it is �some extended Squeak format� which is just defined by the implementation.

I can levee the implementation that is used now (a sq one). Or use extended. But it should be really nice to think about what we want to have instead of what we can use.

If we have to decide, we should take the simple path. The fact that I am completely confused about this is not a good sign (maybe more related to my intelligence� :-)

Marcus