Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: Function with optional default argument cannot cache results
Replies: 3   Last Post: May 1, 2013 9:40 PM

 Messages: [ Previous | Next ]
 Sseziwa Mukasa Posts: 108 Registered: 8/26/07
Re: Function with optional default argument cannot cache results
Posted: May 1, 2013 9:40 PM

On May 1, 2013, at 9:30 AM, Dan O'Brien <danobrie@gmail.com> wrote:

> Thanks for your response. What you say was clear to start with, I should have asked a better question. The question is why wouldn't mathematica assign downvalues to function evaluations when the default arguments (y_:2) are used?

I can't answer why, but one reasone why may be that it would conflict with memoized functions. For example in your case you want f[x_,y_:2] to create a downvalue f[x_,2]:>{Pause[1],1,2} but your function definition creates a downvalue f[1,2]:>{Null,1,2} so now it's not clear what value you should return for f[1,2].

>
> The reason I ask this is because I was playing with a complicated function I defined where I would cache the values. It occurred to me that one of the parameters typically didn't change and so I gave it a default value so that I didn't need to specify it every time. Then I quickly realized the function evalutations where not being cached (downvalues were not being assigned). That is when I specify f[1], mathematica should know that that actually means f[1,2] and give it a downvalue upon evaluation. So my question is, why wouldn't mathematica do this? Would it break some other intended functionality, or, to implement that sort of recognition is just not worth it?

My guess as to why defaults don't create downvalues is above. But you can work around this by defining values for the single argument version of your function f[x_]:=f[x,default]=...

(Debug) In[18]:= f[x_, y_] := f[x, y] = {Pause[1], 1, 2}
f[x_] := f[x, 2]
(Debug) In[20]:= f[1, 2] // AbsoluteTiming
f[1, 2] // AbsoluteTiming
f[1] // AbsoluteTiming
f[1] // AbsoluteTiming
(Debug) Out[20]= {1.000464, {Null, 1, 2}}
(Debug) Out[21]= {3.*10^-6, {Null, 1, 2}}
(Debug) Out[22]= {6.*10^-6, {Null, 1, 2}}
(Debug) Out[23]= {4.*10^-6, {Null, 1, 2}}

>
> -Dan
>
>
> On 5/1/2013 7:46 AM, Sseziwa Mukasa wrote:

>> On May 1, 2013, at 3:37 AM, Dan O'Brien <danobrie@gmail.com> wrote:
>>

>>> Is there a discussion somewhere on why this is?
>>>
>>> In[1]:= \$Version
>>> f[x_, y_: 2] := f[x, y] = {Pause[1], x, y}
>>> f[1, 2] // AbsoluteTiming
>>> f[1, 2] // AbsoluteTiming
>>> f[1] // AbsoluteTiming
>>> f[1] // AbsoluteTiming
>>>
>>> Out[1]= "9.0 for Microsoft Windows (64-bit) (January 25, 2013)"
>>>
>>> Out[3]= {1.014002, {Null, 1, 2}}
>>>
>>> Out[4]= {0., {Null, 1, 2}}
>>>
>>> Out[5]= {1.029602, {Null, 1, 2}}
>>>
>>> Out[6]= {1.014002, {Null, 1, 2}}
>>>

>> The value that's cached for f[1,2] is {Null,1,2} not {Pause[1],1,2}. Out[4] returns the cached value, f[1] doesn't match the cached pattern so f[x_,y_:2] is evaluated again in the case of Out[5] and Out[6]. You can see this by looking at the DownValues of f:
>>
>> (Debug) In[7]:= DownValues[f]
>> (Debug) Out[7]= {HoldPattern[f[1, 2]] :> {Null, 1, 2},
>> HoldPattern[f[x_, y_ : 2]] :> (f[x, y] = {Pause[1], x, y})}
>>
>> It's not clear what you are trying to accomplish, perhaps you want:
>>
>> (Debug) In[39]:= f[x_, y_: 2] := f[x, y] = {Hold[Pause[1]], x, y}
>> ReleaseHold[f[1, 2]] // AbsoluteTiming
>> ReleaseHold[f[1, 2]] // AbsoluteTiming
>> ReleaseHold[f[1]] // AbsoluteTiming
>> ReleaseHold[f[1]] // AbsoluteTiming
>> (Debug) Out[40]= {1.000602, {Null, 1, 2}}
>> (Debug) Out[41]= {1.000028, {Null, 1, 2}}
>> (Debug) Out[42]= {1.000572, {Null, 1, 2}}
>> (Debug) Out[43]= {1.000699, {Null, 1, 2}}

>

Date Subject Author
5/1/13 Sseziwa Mukasa
5/1/13 Dan O'Brien