Eclipse Community Forums: TMF (Xtext) » Serialization of EFloat

Help

Home

Home » Modeling » TMF (Xtext) » Serialization of EFloat(seems to fail for small float values)

Show: Today's Messages :: Show Polls :: Message Navigator

Serialization of EFloat [message #652920]

Mon, 07 February 2011 15:35

Martin Jung

Messages: 4
Registered: July 2009

Junior Member

Hi!

I was playing around with xtext reading files, transforming the Object trees values and serializing the modified tree when I stumbled over a XtextSerializationException.

I boiled it down to this:

I am using a Terminal rule for representing Floating point numbers like this (being lazy, i copied it somwhere):

terminal REALTYPE returns ecore::EFloat : (('-'|'+')? ( INT '.' INT | '.' INT | INT '.') ('e' ('-'|'+') INT)?) | ('-' INT);

It turns out that Serialization fails when small values of floats are used. Turned out further that exactly when float values "serialized" by standard java float->String conversion (as in System.err.println) switch over to scientific notation (which at my installation is at 0.00042f or 4.2E-4). As long as standard java does not switch the values rendering to scientific notation I do not get the XtextSerializationException.

The Question now is twofold: Is this behavior known or am I doing something wrong?
If I do nothing wrong, how do I get rid of the XtextSerializationException?

Any Help appreciated!

Report message to a moderator

Re: Serialization of EFloat [message #652922 is a reply to message #652920]

Mon, 07 February 2011 15:58

Christian Dietrich

Messages: 14665
Registered: July 2009

Senior Member

Hi,

what about writing a ValueConverter as docs say.

~ Christian

Twitter : @chrdietrich
Blog : https://www.dietrich-it.de

Report message to a moderator

Re: Serialization of EFloat [message #652929 is a reply to message #652920]

Mon, 07 February 2011 15:53

Sebastian Zarnekow

Messages: 3118
Registered: July 2009

Senior Member

Hi Martin,

you'll have to define your own value converter that creates a string
which matches your terminal rule. Please refer to the docs for details.

Regards,
Sebastian
--
Need professional support for Eclipse Modeling?
Go visit: http://xtext.itemis.com

Am 07.02.11 16:35, schrieb Martin Jung:
> Hi!
>
> I was playing around with xtext reading files, transforming the Object
> trees values and serializing the modified tree when I stumbled over a
> XtextSerializationException.
>
> I boiled it down to this:
>
> I am using a Terminal rule for representing Floating point numbers like
> this (being lazy, i copied it somwhere):
> terminal REALTYPE returns ecore::EFloat : (('-'|'+')? ( INT '.' INT |
> '.' INT | INT '.') ('e' ('-'|'+') INT)?) | ('-' INT);
>
> It turns out that Serialization fails when small values of floats are
> used. Turned out further that exactly when float values "serialized" by
> standard java float->String conversion (as in System.err.println) switch
> over to scientific notation (which at my installation is at 0.00042f or
> 4.2E-4). As long as standard java does not switch the values rendering
> to scientific notation I do not get the XtextSerializationException.
>
> The Question now is twofold: Is this behavior known or am I doing
> something wrong?
> If I do nothing wrong, how do I get rid of the XtextSerializationException?
>
> Any Help appreciated!

Report message to a moderator

Re: Serialization of EFloat [message #653041 is a reply to message #652929]

Tue, 08 February 2011 08:41

Martin Jung

Messages: 4
Registered: July 2009

Junior Member

Hi Christian & Sebastian.

Thanks for the fast RTFM, I deserved that (thats what you get from decaf).

As expected, the ValueConverter does exactly what I was trying to do.

Cheers,
Martin

Report message to a moderator

Previous Topic:	Resolving references time
Next Topic:	Inferred superclasses

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

]

Current Time: Fri Apr 26 09:15:03 GMT 2024

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter