Re: Should Pure Functions Require &

*To*: mathgroup at smc.vnet.net*Subject*: [mg29781] Re: Should Pure Functions Require &*From*: "Orestis Vantzos" <atelesforos at hotmail.com>*Date*: Sun, 8 Jul 2001 20:35:56 -0400 (EDT)*Organization*: National Technical University of Athens, Greece*References*: <9i8qaq$r05$1@smc.vnet.net>*Sender*: owner-wri-mathgroup at wolfram.com

My objection is based on the fact that there is no true "declaring mechanism" for functions, in the sense that you never really declare what arguments a function is supposed to handle. True, there is the pattern you provide with each SetDelayed evaluation, but that is not definite, since a future SetDelayed could relate to a completely different pattern. To sum up, Mathematica can never be sure that a certain function expects a true Function as its n-th argument, and therefore it can not automatically decide that Select[....,3+#] really means Select[....,3+#&]. Orestis "Ersek, Ted R" <ErsekTR at navair.navy.mil> wrote in message news:9i8qaq$r05$1 at smc.vnet.net... > Earlier I wrote: > -------------------- > > I stated wondering if all would work well if pure functions didn't require > > & at the end. I am thinking it would be great if a future version of > > Mathematica would make the use of & optional. > > > > So for example we could use > > Select[data, #!=0] > > instead of > > Select[data, #!=0&] > > > > > > and we could use > > #^2 /@expr > > instead of > > #^2& /@expr > > > > I would want to have pure functions ending with & optional rather than > > prohibited for backward compatibility. Wouldn't life be better if we > > didn't have to use &. Is there a reason why my suggestion would not work? > > > ----------------- > Orestis Vantzos, > asked whether Select[data, #!=0] > should do what Select[data, #!=0&] does now, > or what Select[data, #!0]& does now. > > In that case one clearly wants Select[data, #!=0&] > since the other case is a pure function that always returns an empty > list. > > ---------------- > The way I would like to see it the kernel would put an & at a > suitable place in the following situations. > 1 A head has one or more #, #n, ##, or ##n but no &. > 2 Use of expr/;test, _?test, __?test, ___?test where test has > " ". > 3 The right side of Set, or SetDelayed has " ". > 4 The second argument of Select, MatrixQ, VectorQ has " ". > 5 An argument of a functional programming construct includes #, #n, > ##, or ##n but no & and a function is expected in this argument. > > Examples of 5 > > In[1]:= g = {##+1, ##+2}; > Through[ g[{x,y,z}] ] > > This would return the same thing as if we had g = {##+1&, ##+2& } > since Through expects an argument of the form p[func1, func2][x] > > > In[2]:= g= {##+1, ##+2}; > Apply[ g, {x,y,z} ] > > This would return the same thing as if we had g= {##+1, ##+2}& > since Apply expects the first argument to be a function. > > > In[3]:= Clear[g]; Apply[g, {x,y,z}] > > Out[3]= g[x,y,z] > > In this case (g) has no #, #n, ##, ##n so an (&) would not be > assumed. > > > I haven't found a case where the "missing" (&) could go "here" or > "there" and both decisions would be useful. Also I am not aware of a use for > #, #n, ##, ##n without an (&). > > ------------------ > > Regards, > > Ted Ersek > > Download Mathematica tips, tricks from > > http://www.verbeia.com/mathematica/tips/Tricks.html > > >