Richard Suchenwirth 2002-12-01 - Here are some little pieces that don't deserve a page of their own yet.
A sum to get e edit
Reading "What is J?" (the ASCIIfied successor to APL, see
Playing APL), one nice example was how to approximate the Euler number e:
+/ % ! i. 9
which in more conventional math notation is
_8_
\ 1
> ---
/__ i!
i=0
Too lazy to start and reimplement J in Tcl, I decided to express this in Tcl as natural as possible:
proc Sum {itName from to body} {
upvar 1 $itName it
set res 0
for {set it $from} {$it<=$to} {incr it} {
set res [uplevel 1 expr $res+$body]
}
set res
}
proc fact n {expr {$n<2? 1: wide($n)*[fact [incr n -1]]}}
It worked alright (the wide() cast was necessary to enter the 64 bit realm), but in the 18th iteration got saturated into a slightly wrong result, while
expr itself knows better:
% Sum i 0 17 {1./[fact $i]}
2.7182818284590455
% expr exp(1)
2.7182818284590451
Continued fraction edit
Also in the J paper, it is shown how to approximate the
golden ratio x-1 = 1/x with the continued fraction
(+ %) / 15 $ 1
which in more conventional spelling is
1
1 + ------------------
1
1 + --------------
1
1 + ----------
1 + ...
ad infinitum (or for a number of iterations at least, 15 in the example). To wrap this nicely in Tcl, I invented an extra notation where @ stands for "the whole thing again". Execution has to start from bottom right of course, with a default value of 1, and work its way up for the specified number of iterations:
proc continuedFraction {body iterations} {
regsub -all @ $body {$res} body
set res 1
for {set i 0} {$i<$iterations} {incr i} {
set res [expr $body]
}
set res
}
% continuedFraction {1 + 1./@} 15
1.6180344478216819
With increasing number of iterations, the results oscillate ever closer around one value, until from 38 up they don't change any more:
% continuedFraction {1 + 1./@} 38
1.6180339887498949