Yes, I could well use the quadratics as the representation, since there 
will be only one root.  I thought about that.  I was trying to see if 
the continued-fraction form had any value in the application I was 
working on, hence the quest for working code.  You are right that the 
periods will vary.  The more I think about this the less I like it; 
still if the code existed it would provide some interesting moments of 
exploration.

Henry Rich

Cliff Reiter wrote:
> I think you will need terms of the form a + c*%:b in your products.
> This requires using different periodic portions
> 
>     qp_to_pcf  x:_15 _4 1 NB. 2+%:19
> +-+-----------+
> |6|2 1 3 1 2 8|
> +-+-----------+
>     qp_to_pcf  x:_37 _2 1 NB. 1+2*%:19
> +-+----+
> |7|6 12|
> +-+----+
> 
> It is entirely tractable to create an inverse to qp_to_pcf and then
> multiply the quadratics given the form you suggest. However,
> I have not seen that done, only described by example.
> I still think that it might be more practical to use rational
> approximations...
> Is it possible you could use quadratics as your representation?
> Just brainstorming...
> 
> Henry Rich wrote:
>> Yes, of interest, please.  (and thanks Ambrus, but I would like to stick 
>> with continued-fraction form).
>>
>> What I meant was, something that operated on the continued fraction 
>> without converting to rational, which it seemed that mcf did.
>>
>> You are right that in general products of these are not repeating, but
>> (a + %:b) * (c + %:b) would be, and I am trying to see if I can make use 
>> of that in something I am working on.
>>
>> Henry Rich
>>
>> Cliff Reiter wrote:
>>> Henry,
>>> I am confused.
>>>
>>>     1 2 3 4 mcf 1 1 2 1 1
>>> 2 2 5 3
>>>
>>> takes two finite continued fractions and gives the continued
>>> fraction for the product. What do you mean by finite cf-product
>>> if not that?
>>>
>>> In the book for my number theory workshop, I also have
>>> utilities for converting between (preperiod;period) and
>>> quadratic polynomial representations (section G.2). Of interest?
>>>
>>> Maybe you should share an illustration of what you want.
>>>
>>> Best,
>>> Cliff
>>>
>>> Henry Rich wrote:
>>>> The continued fractions I want to work with are infinite repeating, so 
>>>> converting to rational doesn't help.  The products will also be infinite 
>>>> repeating, though.  If we can come up with a finite cf-product, I could 
>>>> adapt it to my uses.
>>>>
>>>> Henry Rich
>>>>
>>>> Dan Bron wrote:
>>>>> Henry wrote:
>>>>>> Does anyone have J code for multiplying two numbers expressed as 
>>>>>> simple continued fractions, producing a continued-fraction result?
>>>>> Roger responded:
>>>>>>  As a first approximation, convert to two single (rational) numbers; 
>>>>>>  mutliply; convert back.  *&.conv
>>>>> Here's one way to do that:
>>>>>
>>>>>      NB.  Continued fraction to decimal
>>>>>      cf2d =:  (+%)/                              
>>>>>      
>>>>>      NB.  Decimal to continued fraction
>>>>>      d2cf =:  }:@:<.@:(%@:(-<.)^:(_&>)^:a:&.x:)  
>>>>>      
>>>>>      NB.  continued fraction <-> decimal   
>>>>>      cf   =:  cf2d :. d2cf
>>>>>      
>>>>>      NB.  Multiply continued fractions
>>>>>      cfM  =:  *&.cf
>>>>>
>>>>> I'm sure there are superior algorithms, in particular for  d2cf  .  I'm
>>>>> thinking along the line of Euler's GCD algorithm.  There's a lot of depth
>>>>> hidden in J's rational numbers and the relevant primitives (e.g.  +.  | 
>>>>> #:  ).
>>>>>
>>>>> There might also be a direct way (without the intermediate conversion to
>>>>> decimal), but no ideas occur to me, personally.
>>>>>
>>>>> -Dan
>>>>> ----------------------------------------------------------------------
>>>>> For information about J forums see http://www.jsoftware.com/forums.htm
>>>>>
>>>> ----------------------------------------------------------------------
>>>> For information about J forums see http://www.jsoftware.com/forums.htm
>>>>
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
> 
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to