Services & ResourcesWolfram ForumsMathGroup Archive

[Date Index] [Thread Index] [Author Index]

Re: Confusing behaviour with caching of evaluated expressions ?

• To: mathgroup at smc.vnet.net
• Subject: [mg63745] Re: [mg63727] Confusing behaviour with caching of evaluated expressions ?
• From: "Carl K. Woll" <carlw at wolfram.com>
• Date: Fri, 13 Jan 2006 04:48:23 -0500 (EST)
• References: <200601120822.DAA10481@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

Chris Rodgers wrote:
> Hi,
> 
> Can anyone shed any light upon the following unexpected output from 
> Mathematica?
> 
> I have come across some rather strange behaviour in the way Mathematica 
> handles expressions which depend on a "flag" variable e.g.:
> 
> F[x_] := G /; flag==1
> 
> It appears that Mathematica forgets that flag is a variable that might 
> change in the future. Thus, if one defines another expression e.g.:
> 
> flag=0
> expr1= F[z1]
> expr2:=F[z2]
> 
> expr1 and expr2 will both currently evaluate to F[z1] and F[z2] as 
> expected. BUT, if we now change the flag and try to prod them into 
> evaluating to give G, things become tricky:
> 
> flag=1
> expr1
> expr2
> Evaluate[expr1]
> Evaluate[expr2]
> FullSimplify[expr1]
> FullSimplify[expr2]
> 
> Even doing something like
> 
> F[z1]
> expr1
> 
> doesn't work. (The F[z1] returns G correctly, but expr1 doesn't realise 
> that it should change to give G.)
> 
> The only way to make it work again seems to be to "change" z1 or z2. For 
> example,
> 
> z1=z1+0
> expr1
> 
> RandomFunctionIMadeUp[z2]^:=33
> expr2
> 
> These two assignments are enough to make Mathematica update its cache(?).
> 
> Can anyone explain to me why this might be "by design"?
> 
> Can anyone tell me a way to make Mathematica fully evaluate an 
> expression e.g. expr1 without getting stuck in this way? Is there a 
> SuperFullActuallyEvaluate[expr1] commmand?? :-)
> 
> Many thanks in advance,
> 
> Chris.

The command you are looking for is Update. After changing flag, do 
Update[F]. Then, F will use the new value of flag.

Carl Woll
Wolfram Research

• References:
  • Confusing behaviour with caching of evaluated expressions ?
    • From: Chris Rodgers <rodgers@physchem.NOSPAMox.aREMOVEc.uk>