module tpseval; % Evaluator for truncated power series.
% Authors: Julian Padget & Alan Barnes
% The evaluator interprets the results of the compilation phase and
% is also rule driven until I get round to getting the compilation
% phase to produce directly executable code
% The evaluation functions live on the erule property of the name.
fluid '(ps ps!:order!-limit ps!:max!-order);
% Printing functions now in module tpsconv
% symbolic procedure ps!:prin!: p;
% if constantpsp p then
% maprint(prepsqxx ps!:get!-term(p,0), 0)
% else
% (lambda (first,u,delta,symbolic!-exp!-pt,about,atinf);
% << if !*nat and posn!*<20 then orig!*:=posn!*;
% atinf:=(about='ps!:inf);
% ps!:find!-order p;
% delta:=prepf((ps!:depvar p) .** 1 .*1 .+
% (negf if atinf then nil
% % expansion about infinity
% else if idp about then !*k2f about
% else if ps!:numberp about then !*n2f about
% else if (u:=!*pre2dp about) then !*n2f u
% else !*k2f(symbolic!-exp!-pt:= compress
% append(explode ps!:depvar p, explode '0))));
% if symbolic!-exp!-pt then prin2!* "[";
% prin2!* "{";
% for i:=(ps!:order p): ps!:exp!-lim do
% << u:=ps!:get!-term(p,i);
% if null u then u := ps!:evaluate!-next(p,i);
% if not null numr u then
% <<if minusf numr u then <<u:=negsq u; prin2!* " - ">>
% else if not first then prin2!* " + ";
% first := nil;
% if posn!*>55 then <<terpri!* nil;prin2!* " ">>;
% if denr u neq 1 then prin2!* "(";
% if u neq '(1 . 1) then
% maprint(prepsqxx u,get('times,'infix))
% else if i=0 then prin2!* 1;
% if denr u neq 1 then prin2!* ")";
% if i neq 0 and u neq '(1 . 1) then prin2!* "*";
% if i neq 0 then
% xprinf(!*p2f mksp(delta,
% if atinf then -i else i),nil,nil)
% >>
% >>;
% if first then prin2!* "0";
% if posn!*>55 then terpri!* nil;
% u:=ps!:exp!-lim +1;
% if (u=1) and not atinf and (about neq 0) then
% prin2!* " + O"
% else prin2!* " + O(";
% xprinf(!*p2f mksp(delta,if atinf then -u else u),nil,nil);
% if (u=1) and not atinf and (about neq 0) then
% prin2!* "}"
% else prin2!* ")}";
% if symbolic!-exp!-pt then
% << if posn!*>45 then terpri!* nil;
% prin2!* " where ";
% prin2!* symbolic!-exp!-pt;
% prin2!* " = ";
% maprin about;
% prin2!* "]"
% >>;
% terpri!* nil;
% >>)
% (t,nil,nil,nil,ps!:expansion!-point p,nil);
%
symbolic procedure ps!:unknown!-order ps;
(lambda (u, v);
if v >= u then u
else rerror(tps,20,
list("Can't find the order of ",ps!:value ps)))
(ps!:order ps, ps!:last!-term ps);
symbolic procedure ps!:find!-order ps;
if null ps then 0
else if idp ps then ps % second arg of DF etc are identifiers
else if ps!:numberp ps then 0
else if eqcar(ps,'!:ps!:) then <<
if idp cdr ps then ps!:unknown!-order ps
else if atom ps!:expression ps then ps!:order ps
else ps!:find!-order1(ps)>>
else rerror(tps,21,"Unexpected form in ps!:find!-order");
symbolic procedure ps!:find!-order1(ps);
begin scalar psoperator,psord,pslast;
psord:=ps!:order ps;
pslast:=ps!:last!-term ps;
if psord leq pslast then
return psord;
psoperator:=ps!:operator ps;
psord:=apply(get(psoperator,'ps!:order!-fn), list ps);
ps!:set!-order(ps,psord);
ps!:set!-last!-term(ps,psord-1);
if ps!:value ps =0 then
% prevents infinite loop if we have exact cancellation
<<psord:=0; ps!:set!-last!-term(ps, ps!:max!-order)>>
else while ps!:evaluate!-next(ps,psord)=(nil ./ 1 ) do
% in case we have finite # of cancellations in a sum or difference
<<psord:=psord+1;
if psord > ps!:order!-limit then
rerror(tps,22,list("Expression ", ps!:value ps,
" has zero expansion to order ",
psord))
% We may not always be able to recognise zero,
% so give up after specified number of iterations.
>>;
return psord
end;
symbolic (ps!:order!-limit:=100);
% symbolic here seems to be essential in Cambridge Lisp systems
put('psordlim, 'simpfn, 'simppsordlim);
symbolic procedure simppsordlim u;
begin integer n;
n:=ps!:order!-limit;
if u then ps!:order!-limit := ieval carx(u,'psordlim);
return (if n=0 then nil ./ 1 else n ./ 1);
end;
put('plus,'ps!:order!-fn, 'ps!:plus!-orderfn);
put('int,'ps!:order!-fn,'ps!:int!-orderfn);
put('df,'ps!:order!-fn,'ps!:df!-orderfn);
put('quotient,'ps!:order!-fn, 'ps!:quotient!-orderfn);
put('times,'ps!:order!-fn, 'ps!:times!-orderfn);
put('minus,'ps!:order!-fn, 'ps!:minus!-orderfn);
put('difference,'ps!:order!-fn, 'ps!:difference!-orderfn);
symbolic procedure ps!:int!-orderfn ps;
begin scalar u,v;
v := ps!:depvar ps;
u := ps!:find!-order(rand1 ps!:expression ps);
return
if v=rand2 ps!:expression ps then
if ps!:expansion!-point ps neq 'ps!:inf then
if u=-1 then rerror(tps,23,"Logarithmic Singularity")
else u+1
else % expansion about infinity
if u=1 then rerror(tps,24,"Logarithmic Singularity")
else u-1
else u;
end;
symbolic procedure ps!:df!-orderfn ps;
begin scalar u, v, pt, dfvar;
v:= ps!:expression ps;
u := ps!:find!-order(rand1 v);
dfvar := rand2 v;
pt := ps!:expansion!-point ps;
return
if ps!:depvar ps = dfvar then
if pt neq 'ps!:inf then
if u=0 then 0 else u-1
else if u=0 then 2 else u+1 % expansion about infinity
else if depends(pt, dfvar) then if u=0 then 0 else u-1
else u;
end;
symbolic procedure ps!:quotient!-orderfn ps;
begin scalar u,v;
v := ps!:expression ps;
u := ps!:find!-order(rand1 v);
v := ps!:find!-order(rand2 v);
return difference(u,v);
end;
symbolic procedure ps!:times!-orderfn ps;
begin scalar u,v;
v := ps!:expression ps;
u := ps!:find!-order(rand1 v);
v := ps!:find!-order(rand2 v);
return plus2(u,v);
end;
%symbolic procedure ps!:plus!-orderfn ps;
% eval cons('min , mapcar(rands ps!:expression ps, 'ps!:find!-order));
symbolic procedure ps!:plus!-orderfn ps;
% Re-worked by ACN to avoid excessive numbers of args.
begin scalar w, m;
if null ps then return 0;
w := mapcar(rands ps!:expression ps, 'ps!:find!-order);
m := car w;
for each z in cdr w do if z < m then m := z;
return m
end;
symbolic procedure ps!:minus!-orderfn ps;
ps!:find!-order(rand1 ps!:expression ps);
symbolic procedure ps!:difference!-orderfn ps;
begin scalar u,v;
v := ps!:expression ps;
u := ps!:find!-order(rand1 v);
v := ps!:find!-order(rand2 v);
return min2(u,v);
end;
put('sqrt,'ps!:order!-fn,'ps!:sqrt!-orderfn);
put('sqrt,'ps!:erule,'ps!:sqrt!-erule);
symbolic procedure ps!:sqrt!-orderfn ps;
begin scalar u;
u:=ps!:find!-order rand1 ps!:expression ps;
return
(if v*2=u then v else rerror(tps,25,"Branch Point in Sqrt"))
where v=u/2
end;
symbolic procedure ps!:sqrt!-erule(a,n);
begin scalar aa,x,y,z;
aa:=rand1 a; z:= nil ./ 1;
y:=ps!:order aa;
x:=ps!:order(ps); %order of sqrt ps
if n=x then return simpexpt(list(prepsqxx ps!:evaluate(aa,y),
'(quotient 1 2)));
for k:=1:n-x do
z:=addsq(z,
multsq(((lambda y; if y=0 then nil else y)
(k*3-2*n+y)) ./ 1,
multsq(ps!:evaluate(aa,k+y),
ps!:evaluate(ps,n-k))));
return quotsq(z,multsq(2*(n-x) ./ 1,ps!:evaluate(aa,y)))
end;
% alternative algorithm (for order 0 only)
% for i:=1:n-1 do
% z:=addsq(z,multsq(multsq( i ./ 1,ps!:evaluate(ps,i)),
% ps!:evaluate(ps,n-i)));
% z:=multsq(z, 1 ./ (n+1));
% return quotsq(addsq(ps!:evaluate(aa,n),negsq z),
% multsq(2 ./ 1,ps!:evaluate(b,x)))
put('cbrt,'ps!:order!-fn,'ps!:cbrt!-orderfn);
put('cbrt,'ps!:erule,'ps!:cbrt!-erule);
symbolic procedure ps!:cbrt!-orderfn ps;
begin scalar u;
u:=ps!:find!-order rand1 ps!:expression ps;
return
(if v*3=u then v else rerror(tps,26,"Branch Point in Cbrt"))
where v=u/3
end;
symbolic procedure ps!:cbrt!-erule(a,n);
begin scalar aa,x,y,z;
aa:=rand1 a; z:= nil ./ 1;
y:=ps!:order aa;
x:=ps!:order(ps); %order of cbrt ps
if n=x then return simpexpt(list(prepsqxx ps!:evaluate(aa,y),
'(quotient 1 3)));
for k:=1:n-x do
z:=addsq(z,
multsq(((lambda y; if y=0 then nil else y)
(k*4-3*n+y)) ./ 1,
multsq(ps!:evaluate(aa,k+y),
ps!:evaluate(ps,n-k))));
return quotsq(z,multsq(3*(n-x) ./ 1,ps!:evaluate(aa,y)))
end;
symbolic procedure ps!:evaluate(ps,i);
begin scalar term;
term:=ps!:get!-term (ps,i);
if term then return term;
for j:=ps!:last!-term(ps)+1:i do
term:= ps!:evaluate!-next(ps,j);
return term;
end;
symbolic procedure ps!:evaluate!-next(ps,n);
% The appropriate evaluation rule for the operator
% in the ps is selected and invoked
begin scalar next;
next := apply(get(ps!:operator ps,'ps!:erule),
list(ps!:expression ps,n));
ps!:set!-term(ps,n,next:=simp!* prepsqxx next);
return next;
end;
symbolic procedure ps!:plus!-erule(a,n);
begin scalar z;
z := nil ./ 1;
foreach term in rands a do
z:=addsq(z, ps!:evaluate(term, n));
return z
end;
put('plus,'ps!:erule,'ps!:plus!-erule);
symbolic procedure ps!:minus!-erule(a,n);
negsq ps!:evaluate(rand1 a,n);
put('minus,'ps!:erule,'ps!:minus!-erule);
symbolic procedure ps!:difference!-erule(a,n);
addsq(ps!:evaluate(rand1 a,n),
negsq ps!:evaluate(rand2 a,n));
put('difference,'ps!:erule,'ps!:difference!-erule);
symbolic procedure ps!:times!-erule(a,n);
begin scalar aa,b,x,y,y1,z;
aa:=rand1 a; b:= rand2 a; z:= nil ./ 1;
x:=ps!:order(aa);
y:=ps!:order(ps); % order of product ps
y1 := ps!:order b;
for i := 0:n-y do if n-x-i>=y1 then
z:= addsq(z,multsq(ps!:evaluate(aa,i+x),
ps!:evaluate(b,n-x-i)));
return z
end;
put('times,'ps!:erule,'ps!:times!-erule);
symbolic procedure ps!:quotient!-erule(a,n);
begin scalar aa,b,x,y,z;
aa:=rand1 a; b:=rand2 a; z:= nil ./ 1;
y:=ps!:order(b);
x:=ps!:order(ps); %order of quotient ps
for i:=1:n-x do
z:=addsq(z,multsq(ps!:evaluate(b,i+y),
ps!:evaluate(ps,n-i)));
return quotsq(addsq(ps!:evaluate(aa,n+y),negsq z),
ps!:evaluate(b,y))
end;
put('quotient,'ps!:erule,'ps!:quotient!-erule);
% the next two functions deal more efficiently with common special
% cases of multiplication or division by a constant
% the constmult operator is produced by
% ps!:times!-crule and ps!:quotient!-crule
%
put('psmult,'ps!:order!-fn, 'ps!:constmult!-orderfn);
put('psmult,'ps!:erule,'ps!:constmult!-erule);
symbolic procedure ps!:constmult!-orderfn ps;
ps!:find!-order rand2 ps!:expression ps;
symbolic procedure ps!:constmult!-erule(a,n);
multsq(rand1 a, ps!:evaluate(rand2 a,n));
symbolic procedure ps!:df!-erule(a,n);
begin scalar dfvar, series, about;
dfvar := rand2 a;
series := rand1 a;
about := ps!:expansion!-point series;
return
if dfvar = ps!:depvar series then
if about neq 'ps!:inf then
multsq((n+1) ./ 1,ps!:evaluate(series, n+1))
else multsq((1-n) ./ 1,ps!:evaluate(series, n-1))
else if depends(about, dfvar) then
addsq(diffsq(ps!:evaluate(series,n),dfvar),
multsq((-n-1) ./ 1, multsq(ps!:evaluate(series,n+1),
diffsq(simp!* about,dfvar))))
else diffsq(ps!:evaluate(series,n),dfvar);
end;
put('df,'ps!:erule,'ps!:df!-erule);
symbolic procedure ps!:int!-erule(a,n);
if rand2 a=ps!:depvar rand1 a then
if ps!:expansion!-point rand1 a neq 'ps!:inf then
quotsq(ps!:evaluate(rand1 a,n-1), n ./ 1)
else quotsq(ps!:evaluate(rand1 a,n+1),-n ./ 1)
else simpint list(prepsqxx ps!:evaluate(rand1 a,n),rand2 a);
put('int,'ps!:erule,'ps!:int!-erule);
symbolic procedure ps!:expt!-orderfn ps;
begin scalar u, v, w, expres;
expres := ps!:expression ps;
u:= ps!:find!-order rand1 expres;
v:= rand2 expres;
w := cadddr expres;
if cdr(v:=divide(u * v,w))=0 then return car v
else rerror(tps,27,"Branch Point in EXPT")
end;
symbolic procedure ps!:expt!-erule(a,n);
begin scalar base,x,y,z,p,q;
base:= rand1 a;
p:=rand2 a; q:=cadddr a;
y:=ps!:order(base);
z:= ps!:order ps; % order of exponential
if n=z then <<
if q =1 then x := p else x := list('quotient, p, q);
return simpexpt(list(prepsqxx ps!:evaluate(base,y),x))>>
else << x:= nil ./ 1;
for k:=1:n-z do
x:=addsq(x,
multsq(((lambda num; if num=0 then nil else num)
(k*p+q*(k-n+z))) ./ q,
multsq(ps!:evaluate(base,k+y),
ps!:evaluate(ps,n-k))));
return quotsq(x,multsq((n-z) ./ 1, ps!:evaluate(base,y)))
>>;
end;
put('expt,'ps!:erule, 'ps!:expt!-erule);
put('expt,'ps!:order!-fn,'ps!:expt!-orderfn);
symbolic procedure ps!:exp!-orderfn ps;
if ps!:find!-order rand1 ps!:expression ps<0 then
rerror(tps, 28, "Essential Singularity in EXP")
else 0;
symbolic procedure ps!:exp!-erule(a,n);
begin scalar exp1, x;
exp1:= rand1 a;
if n=0 then
return simpexpt(list('e, prepsqxx ps!:evaluate(exp1,0)));
x:= nil ./ 1;
for k:=0:n-1 do
x:=addsq(x, multsq((n-k) ./ 1,
multsq(ps!:evaluate(exp1,n-k),
ps!:evaluate(ps,k))));
return quotsq(x, n ./ 1);
end;
put('exp,'ps!:erule, 'ps!:exp!-erule);
put('exp,'ps!:order!-fn,'ps!:exp!-orderfn);
endmodule;
end;