[Date Index]
[Thread Index]
[Author Index]
Re: Simple recursive assignment
*To*: mathgroup at smc.vnet.net
*Subject*: [mg16028] Re: [mg15996] Simple recursive assignment
*From*: "Wolf, Hartmut" <hwolf at debis.com>
*Date*: Sat, 20 Feb 1999 02:52:12 -0500
*Organization*: debis Systemhaus
*References*: <199902190827.DAA18900@smc.vnet.net.>
*Sender*: owner-wri-mathgroup at wolfram.com
Ciao Peltio,
Peltio schrieb:
>
> Can someone spare a little time for a trivial question?
>
> The assignement
> a[0]=1.;
> a[1]=2.;
> a[n_]:=a[n]=a[n-1]-a[n-2]/n
> works fine, all right...
> a[5]
> 0.291667
>
equally fine will work that
In[4]:=b[0]=1.;
b[1]=2.;
In[5]:=expr[n_]:= b[n-1]-b[n-2]/n
In[6]:=b[n_]:=b[n]=expr[n]
In[7]:=b[5]
Out[7]=0.291667
but not
> expr=b[n-1]-b[n-2]/n;
>
> b[0]=1;
> b[1]=2;
> b[n_]:=b[n]=expr
>
> b[3]
> $RecursionLimit::reclim: etc.
>
Why? From where shall expr get it's value for n? Case 1: n is just a
Symbol with no value. Then if you call b[3] then you call expr, and that
calls b[n-1] and b[n-2] and n hasn't got a value! which means calling
expr, which means calling b[n-1] and b[n-2] with the *same* n! (if n had
a value before, say 5, then expr is defined as b[4]-b[3]/5 and b[4] is
the same as expr as is also b[3]) ad infinitum!
But if you do the following, you will succeed:
In[9]:= Remove[b, expr, n]
In[10]:= b[0]=1.;
b[1]=2.;
In[11]:= expr= b[n-1]-b[n-2]/n;
In[12]:= b[nn_]:= b[nn]=Block[{n=nn}, expr]
In[13]:= b[5]
Out[13]= 0.291667
(Just a remark: if you use (n=nn; expr) instead of Block[...] the you
will also terminate, but get a wrong result. Can you see why?)
ciao, Hartmut
H. Wolf, debis Systemhaus, Darmstadt, Germania
=======[mailto:hwolf at debis.com]=======
Prev by Date:
**Re: Simple recursive assignment**
Next by Date:
**Re: Must Play[] generate a graphic?**
Previous by thread:
**Simple recursive assignment**
Next by thread:
**Re: Simple recursive assignment**
| |