[Numbers] How far from a first release?

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

[Numbers] How far from a first release?

Gilles Sadowski
Hello.

There are 15 open issues in
   https://issues.apache.org/jira/browse/NUMBERS

See also:
   https://issues.apache.org/jira/browse/NUMBERS-25

Can we consider a first release?
Should it be a "beta"?[1]
Please indicate what you think the status is of each
module.
Please indicate which issues could be postponed.
Please assign yourself some issues...
What is the status of "commons-numbers-complex"?

Is there some more code we can move from "Commons Math"?
Please remove from "Math" the code which you have ported
to "Numbers"; see
   https://issues.apache.org/jira/browse/MATH-1416


Thanks,
Gilles

[1] Unless I'm mistaken, the previous experiments did not
     bring any insights as how to improve the subsequent
     release.


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: [Numbers] How far from a first release?

Eric Barnhill
As far as commons-complex goes. To conform the library to C standards, the
entire approach to NaN had to be rethought. In fact it is not clear from
the code what the justification was for the NaN approach in the present
code. To handle corner cases of nearly every operation in accordance with
the standard, NaNs and infinities in real or imaginary parts need to be
handled in flexible ways; and nearly every operation in the previous
Complex library simply checked for NaN (in either Re or Im) at the
beginning, then returned NaN immediately if this was found, and sometimes
also for Inf. I stripped all these out, then set up a testing regime to see
how close the trig methods came to matching the C standards for the corner
cases without additional code to check those cases. What remains is to code
in handling of the corner cases when the "default" response of the complex
object does not match the C standards.

Other progress has been made. All the C++ syntactic sugar is in. Some new
methods in C++ 11 (the proj method comes to mind) were added.

I had to let it sit because I needed to learn a bunch of R and Julia for
work, but if there is a release coming, I will use it as motivation to get
complex finished.

Eric


On Sat, Jun 3, 2017 at 10:26 AM, Gilles <[hidden email]>
wrote:

> Hello.
>
> There are 15 open issues in
>   https://issues.apache.org/jira/browse/NUMBERS
>
> See also:
>   https://issues.apache.org/jira/browse/NUMBERS-25
>
> Can we consider a first release?
> Should it be a "beta"?[1]
> Please indicate what you think the status is of each
> module.
> Please indicate which issues could be postponed.
> Please assign yourself some issues...
> What is the status of "commons-numbers-complex"?
>
> Is there some more code we can move from "Commons Math"?
> Please remove from "Math" the code which you have ported
> to "Numbers"; see
>   https://issues.apache.org/jira/browse/MATH-1416
>
>
> Thanks,
> Gilles
>
> [1] Unless I'm mistaken, the previous experiments did not
>     bring any insights as how to improve the subsequent
>     release.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>
>
Reply | Threaded
Open this post in threaded view
|

Re: [Numbers] How far from a first release?

Gilles Sadowski
Hello Eric.

On Sun, 4 Jun 2017 22:03:44 +0200, Eric Barnhill wrote:

> As far as commons-complex goes. To conform the library to C
> standards, the
> entire approach to NaN had to be rethought. In fact it is not clear
> from
> the code what the justification was for the NaN approach in the
> present
> code. To handle corner cases of nearly every operation in accordance
> with
> the standard, NaNs and infinities in real or imaginary parts need to
> be
> handled in flexible ways; and nearly every operation in the previous
> Complex library simply checked for NaN (in either Re or Im) at the
> beginning, then returned NaN immediately if this was found, and
> sometimes
> also for Inf. I stripped all these out, then set up a testing regime
> to see
> how close the trig methods came to matching the C standards for the
> corner
> cases without additional code to check those cases. What remains is
> to code
> in handling of the corner cases when the "default" response of the
> complex
> object does not match the C standards.
>
> Other progress has been made. All the C++ syntactic sugar is in. Some
> new
> methods in C++ 11 (the proj method comes to mind) were added.
>
> I had to let it sit because I needed to learn a bunch of R and Julia
> for
> work, but if there is a release coming, I will use it as motivation
> to get
> complex finished.

Thanks for the update (and the work!).
[I can't comment on the NaN handling...]

Do you think that you could commit the current state of that work
to "master" (possibly tagging with "@Ignore" the unit tests that
still fail)?
That would enable someone else to upgrade the "Commons Math" code
to use the class from "Commons Numbers" (cf. MATH-1416).
There are also issues with "ComplexUtils" which someone else could
take on if you don't have the time.[1]

As for the release, I'm inquiring about what would hold it.

Best regards,
Gilles

[1] But please do comment on them.


> Eric
>
>
> On Sat, Jun 3, 2017 at 10:26 AM, Gilles
> <[hidden email]>
> wrote:
>
>> Hello.
>>
>> There are 15 open issues in
>>   https://issues.apache.org/jira/browse/NUMBERS
>>
>> See also:
>>   https://issues.apache.org/jira/browse/NUMBERS-25
>>
>> Can we consider a first release?
>> Should it be a "beta"?[1]
>> Please indicate what you think the status is of each
>> module.
>> Please indicate which issues could be postponed.
>> Please assign yourself some issues...
>> What is the status of "commons-numbers-complex"?
>>
>> Is there some more code we can move from "Commons Math"?
>> Please remove from "Math" the code which you have ported
>> to "Numbers"; see
>>   https://issues.apache.org/jira/browse/MATH-1416
>>
>>
>> Thanks,
>> Gilles
>>
>> [1] Unless I'm mistaken, the previous experiments did not
>>     bring any insights as how to improve the subsequent
>>     release.
>>


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]