MathGroup Archive 2006

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

Search the Archive

Re: Re: HoldFirst confusion

  • To: mathgroup at smc.vnet.net
  • Subject: [mg65545] Re: [mg65537] Re: HoldFirst confusion
  • From: "Carl K. Woll" <carlw at wolfram.com>
  • Date: Fri, 7 Apr 2006 06:14:21 -0400 (EDT)
  • References: <e1095n$lkf$1@smc.vnet.net> <200604061053.GAA19559@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

Jon Harrop wrote:
> Yaroslav Bulatov wrote:
> 
>>The following gives an error
>>f = 5; f[a_] = a
>>
>>Set is HoldFirst, so why does it evaluate it's first argument f[a_]?
> 
> 
> Conversely, why do the attributes of Set and SetDelayed indicate that they
> hold their first argument when the kernel evaluates it.
> 

As David Bailey wrote, Set and SetDelayed do have HoldFirst attributes 
so that they can see the left hand side before evaluation. That doesn't 
mean that they are not able to evaluate all or part of the left hand 
side in doing their function. In this case, Set and SetDelayed need to 
know the head  of the expression to attach a DownValue to the 
appropriate head. Here is an example where you can see how things are 
working. First create two HoldFirst functions, and set one equal to the 
other:

SetAttributes[{f,g},HoldFirst]

f=g

Now, give f a down value:

f[1+2]:=foo

Check Information for f:

In[5]:=
??f

Global`f

Attributes[f] = {HoldFirst}

f = g

Notice that there is no DownValue associated with f[1+2]=f. This is 
because SetDelayed first evaluated the head of f[1+2] and discovered 
that it was really g. Now check the Information for g:

In[6]:=
??g

Global`g

Attributes[g] = {HoldFirst}

g[1 + 2] := foo

Notice that g has gotten a DownValue, and the argument of g is 1+2 and 
not 3. To summarize, SetDelayed evaluated the head but not the argument 
when given a first argument of f[1+2].

> On a related note, how can you define and use a downvalue that matches
> g[1+2] without evaluating the 1+2?
> 

One possibility:

SetAttributes[g,HoldFirst]

g[Plus[a__]]=Hold[a]

In[10]:=
g[1+2]

Out[10]=
Hold[1,2]

Carl Woll
Wolfram Research


  • Prev by Date: Re: Re: HoldFirst confusion
  • Next by Date: Re: Sphere
  • Previous by thread: Re: Re: HoldFirst confusion
  • Next by thread: Re: HoldFirst confusion