Use Advanced Search to search the entire archive.
[jsr363-experts] Re: RI for Java ME
- From: Werner Keil <
>
- To: "
" <
>
- Subject: [jsr363-experts] Re: RI for Java ME
- Date: Fri, 30 Oct 2015 10:06:51 -0700
Hi,
Thanks to SFO's decent Wifi and desks, I was able to use a ~2h delay of my
flight (and rerouted via Chicago because the original NY transfer would
just have been 25 min!;-) to address
https://java.net/jira/browse/UNITSOFMEASUREMENT-3. The RI contains no more
trace of ME incompatible stuff and slimmed down to approx. 155kb JAR size
(the exploded class folder is pretty much 300k total)
I ran the ME demo
https://github.com/unitsofmeasurement/uom-demos/tree/master/javame/medemo
changed to use the RI in the emulator and it worked :-)
Please check it out Leo or others if you can against both emulator and real
devices. If it works with my old Raspberry Pi I have in Switzerland, I may
give that a try in the ME SDK for DevoXX.
This was a really great ultimate outcome of the Hackergarten sessions and
also shows, we didn't get the JCP Award for nothing.
There are almost no major issues left for the Public Draft sprint. The
license issue with Eclipse and others are solved from our end, but we'll
keep that open till the Orbit ticket can be opened again and hopefully
resolved.
Happy Halloween,
Werner
On Thu, Oct 29, 2015 at 8:26 PM, Werner Keil
<
>
wrote:
>
Dear Experts,
>
>
Just in case anybody missed it, congratulations once more for JSR 363
>
winning the most significant JSR JCP Award this week at JavaOne.
>
>
Those who were here are either already on their way back home or soon
>
leaving (I also do first thing tomorrow morning)
>
>
During the last few sessions I took a closer look at RI issues related to
>
Java ME 8 Embedded.
>
A branch for that broke some number of unit tests while eliminating all
>
dependencies on methods or types not available in ME 8. Gradually merging
>
or re-applying them to the master one by one, I think there's a
>
breakthrough to get rid of pretty much all that's left.
>
>
Calls to Math are substituted with a ME compliant helper class. At least
>
one method in Character is also substituted in a ME compliant way. A major
>
crux was the Field object, but it turns out, SimpleUnitFormat does have
>
quite some parse() functionality of its own, at least in a Locale-neutral
>
way. So enabling that for all relevant units (the key is also the label()
>
method here especially for new or derived units) looks like we can get rid
>
of EBNFUnitFormat at least in the ME compliant RI.
>
It has advantages for SE, so for a Java SE variant we should keep it.
>
>
This looks like some helper classes and parser or token elements used by
>
SymbolMap / EBNFUnitFormat can probably be removed. And the RI size should
>
decrease, at least to around 150kb I would estimate. Thus the RI and API
>
together should be under 200k. If that is a problem for certain devices, we
>
may have to look into modularization of the RI, but first let`s see how to
>
get it running in the same way as the minimalistic enum implementation
>
already did.
>
>
This is not a show-stopper for Public Review, but I am confident it can be
>
done in the course of next week or soon after. And we won't propose the
>
draft till another week or so giving the PMO and everybody else enough time
>
to recover from JavaOne.
>
>
Regards,
>
>
>
Werner
>