]> www.ginac.de Git - ginac.git/blobdiff - ginac/inifcns_trans.cpp
* Added realsymbol class.
[ginac.git] / ginac / inifcns_trans.cpp
index fae653182f4f7c9076526c5cd1dba36ac2e4abef..08758f563461f209892e654acb967e3ef1e57901 100644 (file)
@@ -1,9 +1,10 @@
 /** @file inifcns_trans.cpp
  *
  *  Implementation of transcendental (and trigonometric and hyperbolic)
- *  functions.
- *
- *  GiNaC Copyright (C) 1999 Johannes Gutenberg University Mainz, Germany
+ *  functions. */
+
+/*
+ *  GiNaC Copyright (C) 1999-2004 Johannes Gutenberg University Mainz, Germany
  *
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
 #include "constant.h"
 #include "numeric.h"
 #include "power.h"
+#include "operators.h"
+#include "relational.h"
+#include "symbol.h"
+#include "pseries.h"
+#include "utils.h"
+
+namespace GiNaC {
 
 //////////
 // exponential function
 //////////
 
-ex exp_evalf(ex const & x)
-{
-    BEGIN_TYPECHECK
-        TYPECHECK(x,numeric)
-    END_TYPECHECK(exp(x))
-    
-    return exp(ex_to_numeric(x)); // -> numeric exp(numeric)
-}
-
-ex exp_eval(ex const & x)
-{
-    // exp(0) -> 1
-    if (x.is_zero()) {
-        return exONE();
-    }
-    // exp(n*Pi*I/2) -> {+1|+I|-1|-I}
-    ex TwoExOverPiI=(2*x)/(Pi*I);
-    if (TwoExOverPiI.info(info_flags::integer)) {
-        numeric z=mod(ex_to_numeric(TwoExOverPiI),numeric(4));
-        if (z.is_equal(numZERO()))
-            return exONE();
-        if (z.is_equal(numONE()))
-            return ex(I);
-        if (z.is_equal(numTWO()))
-            return exMINUSONE();
-        if (z.is_equal(numTHREE()))
-            return ex(-I);
-    }
-    // exp(log(x)) -> x
-    if (is_ex_the_function(x, log))
-        return x.op(0);
-    
-    // exp(float)
-    if (x.info(info_flags::numeric) && !x.info(info_flags::rational))
-        return exp_evalf(x);
-    
-    return exp(x).hold();
-}    
+static ex exp_evalf(const ex & x)
+{
+       if (is_exactly_a<numeric>(x))
+               return exp(ex_to<numeric>(x));
+       
+       return exp(x).hold();
+}
+
+static ex exp_eval(const ex & x)
+{
+       // exp(0) -> 1
+       if (x.is_zero()) {
+               return _ex1;
+       }
+       // exp(n*Pi*I/2) -> {+1|+I|-1|-I}
+       const ex TwoExOverPiI=(_ex2*x)/(Pi*I);
+       if (TwoExOverPiI.info(info_flags::integer)) {
+               const numeric z = mod(ex_to<numeric>(TwoExOverPiI),_num4);
+               if (z.is_equal(_num0))
+                       return _ex1;
+               if (z.is_equal(_num1))
+                       return ex(I);
+               if (z.is_equal(_num2))
+                       return _ex_1;
+               if (z.is_equal(_num3))
+                       return ex(-I);
+       }
+       // exp(log(x)) -> x
+       if (is_ex_the_function(x, log))
+               return x.op(0);
+       
+       // exp(float)
+       if (x.info(info_flags::numeric) && !x.info(info_flags::crational))
+               return exp(ex_to<numeric>(x));
+       
+       return exp(x).hold();
+}
 
-ex exp_diff(ex const & x, unsigned diff_param)
+static ex exp_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
+       GINAC_ASSERT(deriv_param==0);
 
-    return exp(x);
+       // d/dx exp(x) -> exp(x)
+       return exp(x);
 }
 
-REGISTER_FUNCTION(exp, exp_eval, exp_evalf, exp_diff, NULL);
+REGISTER_FUNCTION(exp, eval_func(exp_eval).
+                       evalf_func(exp_evalf).
+                       derivative_func(exp_deriv).
+                       latex_name("\\exp"));
 
 //////////
 // natural logarithm
 //////////
 
-ex log_evalf(ex const & x)
-{
-    BEGIN_TYPECHECK
-        TYPECHECK(x,numeric)
-    END_TYPECHECK(log(x))
-    
-    return log(ex_to_numeric(x)); // -> numeric log(numeric)
-}
-
-ex log_eval(ex const & x)
-{
-    if (x.info(info_flags::numeric)) {
-        // log(1) -> 0
-        if (x.is_equal(exONE()))
-            return exZERO();
-        // log(-1) -> I*Pi
-        if (x.is_equal(exMINUSONE()))
-            return (I*Pi);
-        // log(I) -> Pi*I/2
-        if (x.is_equal(I))
-            return (I*Pi*numeric(1,2));
-        // log(-I) -> -Pi*I/2
-        if (x.is_equal(-I))
-            return (I*Pi*numeric(-1,2));
-        // log(0) -> throw singularity
-        if (x.is_equal(exZERO()))
-            throw(std::domain_error("log_eval(): log(0)"));
-        // log(float)
-        if (!x.info(info_flags::rational))
-            return log_evalf(x);
-    }
-    
-    return log(x).hold();
-}    
+static ex log_evalf(const ex & x)
+{
+       if (is_exactly_a<numeric>(x))
+               return log(ex_to<numeric>(x));
+       
+       return log(x).hold();
+}
 
-ex log_diff(ex const & x, unsigned diff_param)
+static ex log_eval(const ex & x)
 {
-    ASSERT(diff_param==0);
+       if (x.info(info_flags::numeric)) {
+               if (x.is_zero())         // log(0) -> infinity
+                       throw(pole_error("log_eval(): log(0)",0));
+               if (x.info(info_flags::real) && x.info(info_flags::negative))
+                       return (log(-x)+I*Pi);
+               if (x.is_equal(_ex1))  // log(1) -> 0
+                       return _ex0;
+               if (x.is_equal(I))       // log(I) -> Pi*I/2
+                       return (Pi*I*_num1_2);
+               if (x.is_equal(-I))      // log(-I) -> -Pi*I/2
+                       return (Pi*I*_num_1_2);
+               // log(float)
+               if (!x.info(info_flags::crational))
+                       return log(ex_to<numeric>(x));
+       }
+       // log(exp(t)) -> t (if -Pi < t.imag() <= Pi):
+       if (is_ex_the_function(x, exp)) {
+               const ex &t = x.op(0);
+               if (is_a<symbol>(t) && t.info(info_flags::real)) {
+                       return t;
+               }
+               if (t.info(info_flags::numeric)) {
+                       const numeric &nt = ex_to<numeric>(t);
+                       if (nt.is_real())
+                               return t;
+               }
+       }
+       
+       return log(x).hold();
+}
 
-    return power(x, -1);
+static ex log_deriv(const ex & x, unsigned deriv_param)
+{
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx log(x) -> 1/x
+       return power(x, _ex_1);
 }
 
-REGISTER_FUNCTION(log, log_eval, log_evalf, log_diff, NULL);
+static ex log_series(const ex &arg,
+                     const relational &rel,
+                     int order,
+                     unsigned options)
+{
+       GINAC_ASSERT(is_a<symbol>(rel.lhs()));
+       ex arg_pt;
+       bool must_expand_arg = false;
+       // maybe substitution of rel into arg fails because of a pole
+       try {
+               arg_pt = arg.subs(rel, subs_options::no_pattern);
+       } catch (pole_error) {
+               must_expand_arg = true;
+       }
+       // or we are at the branch point anyways
+       if (arg_pt.is_zero())
+               must_expand_arg = true;
+       
+       if (must_expand_arg) {
+               // method:
+               // This is the branch point: Series expand the argument first, then
+               // trivially factorize it to isolate that part which has constant
+               // leading coefficient in this fashion:
+               //   x^n + x^(n+1) +...+ Order(x^(n+m))  ->  x^n * (1 + x +...+ Order(x^m)).
+               // Return a plain n*log(x) for the x^n part and series expand the
+               // other part.  Add them together and reexpand again in order to have
+               // one unnested pseries object.  All this also works for negative n.
+               pseries argser;          // series expansion of log's argument
+               unsigned extra_ord = 0;  // extra expansion order
+               do {
+                       // oops, the argument expanded to a pure Order(x^something)...
+                       argser = ex_to<pseries>(arg.series(rel, order+extra_ord, options));
+                       ++extra_ord;
+               } while (!argser.is_terminating() && argser.nops()==1);
+
+               const symbol &s = ex_to<symbol>(rel.lhs());
+               const ex &point = rel.rhs();
+               const int n = argser.ldegree(s);
+               epvector seq;
+               // construct what we carelessly called the n*log(x) term above
+               const ex coeff = argser.coeff(s, n);
+               // expand the log, but only if coeff is real and > 0, since otherwise
+               // it would make the branch cut run into the wrong direction
+               if (coeff.info(info_flags::positive))
+                       seq.push_back(expair(n*log(s-point)+log(coeff), _ex0));
+               else
+                       seq.push_back(expair(log(coeff*pow(s-point, n)), _ex0));
+
+               if (!argser.is_terminating() || argser.nops()!=1) {
+                       // in this case n more (or less) terms are needed
+                       // (sadly, to generate them, we have to start from the beginning)
+                       const ex newarg = ex_to<pseries>((arg/coeff).series(rel, order+n, options)).shift_exponents(-n).convert_to_poly(true);
+                       return pseries(rel, seq).add_series(ex_to<pseries>(log(newarg).series(rel, order, options)));
+               } else  // it was a monomial
+                       return pseries(rel, seq);
+       }
+       if (!(options & series_options::suppress_branchcut) &&
+            arg_pt.info(info_flags::negative)) {
+               // method:
+               // This is the branch cut: assemble the primitive series manually and
+               // then add the corresponding complex step function.
+               const symbol &s = ex_to<symbol>(rel.lhs());
+               const ex &point = rel.rhs();
+               const symbol foo;
+               const ex replarg = series(log(arg), s==foo, order).subs(foo==point, subs_options::no_pattern);
+               epvector seq;
+               seq.push_back(expair(-I*csgn(arg*I)*Pi, _ex0));
+               seq.push_back(expair(Order(_ex1), order));
+               return series(replarg - I*Pi + pseries(rel, seq), rel, order);
+       }
+       throw do_taylor();  // caught by function::series()
+}
+
+REGISTER_FUNCTION(log, eval_func(log_eval).
+                       evalf_func(log_evalf).
+                       derivative_func(log_deriv).
+                       series_func(log_series).
+                       latex_name("\\ln"));
 
 //////////
 // sine (trigonometric function)
 //////////
 
-ex sin_evalf(ex const & x)
-{
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(sin(x))
-    
-    return sin(ex_to_numeric(x)); // -> numeric sin(numeric)
-}
-
-ex sin_eval(ex const & x)
-{
-    // sin(n*Pi) -> 0
-    ex xOverPi=x/Pi;
-    if (xOverPi.info(info_flags::integer))
-        return exZERO();
-    
-    // sin((2n+1)*Pi/2) -> {+|-}1
-    ex xOverPiMinusHalf=xOverPi-exHALF();
-    if (xOverPiMinusHalf.info(info_flags::even))
-        return exONE();
-    else if (xOverPiMinusHalf.info(info_flags::odd))
-        return exMINUSONE();
-    
-    if (is_ex_exactly_of_type(x, function)) {
-        ex t=x.op(0);
-        // sin(asin(x)) -> x
-        if (is_ex_the_function(x, asin))
-            return t;
-        // sin(acos(x)) -> (1-x^2)^(1/2)
-        if (is_ex_the_function(x, acos))
-            return power(exONE()-power(t,exTWO()),exHALF());
-        // sin(atan(x)) -> x*(1+x^2)^(-1/2)
-        if (is_ex_the_function(x, atan))
-            return t*power(exONE()+power(t,exTWO()),exMINUSHALF());
-    }
-    
-    // sin(float) -> float
-    if (x.info(info_flags::numeric) && !x.info(info_flags::rational))
-        return sin_evalf(x);
-    
-    return sin(x).hold();
-}
-
-ex sin_diff(ex const & x, unsigned diff_param)
-{
-    ASSERT(diff_param==0);
-    
-    return cos(x);
-}
-
-REGISTER_FUNCTION(sin, sin_eval, sin_evalf, sin_diff, NULL);
+static ex sin_evalf(const ex & x)
+{
+       if (is_exactly_a<numeric>(x))
+               return sin(ex_to<numeric>(x));
+       
+       return sin(x).hold();
+}
+
+static ex sin_eval(const ex & x)
+{
+       // sin(n/d*Pi) -> { all known non-nested radicals }
+       const ex SixtyExOverPi = _ex60*x/Pi;
+       ex sign = _ex1;
+       if (SixtyExOverPi.info(info_flags::integer)) {
+               numeric z = mod(ex_to<numeric>(SixtyExOverPi),_num120);
+               if (z>=_num60) {
+                       // wrap to interval [0, Pi)
+                       z -= _num60;
+                       sign = _ex_1;
+               }
+               if (z>_num30) {
+                       // wrap to interval [0, Pi/2)
+                       z = _num60-z;
+               }
+               if (z.is_equal(_num0))  // sin(0)       -> 0
+                       return _ex0;
+               if (z.is_equal(_num5))  // sin(Pi/12)   -> sqrt(6)/4*(1-sqrt(3)/3)
+                       return sign*_ex1_4*sqrt(_ex6)*(_ex1+_ex_1_3*sqrt(_ex3));
+               if (z.is_equal(_num6))  // sin(Pi/10)   -> sqrt(5)/4-1/4
+                       return sign*(_ex1_4*sqrt(_ex5)+_ex_1_4);
+               if (z.is_equal(_num10)) // sin(Pi/6)    -> 1/2
+                       return sign*_ex1_2;
+               if (z.is_equal(_num15)) // sin(Pi/4)    -> sqrt(2)/2
+                       return sign*_ex1_2*sqrt(_ex2);
+               if (z.is_equal(_num18)) // sin(3/10*Pi) -> sqrt(5)/4+1/4
+                       return sign*(_ex1_4*sqrt(_ex5)+_ex1_4);
+               if (z.is_equal(_num20)) // sin(Pi/3)    -> sqrt(3)/2
+                       return sign*_ex1_2*sqrt(_ex3);
+               if (z.is_equal(_num25)) // sin(5/12*Pi) -> sqrt(6)/4*(1+sqrt(3)/3)
+                       return sign*_ex1_4*sqrt(_ex6)*(_ex1+_ex1_3*sqrt(_ex3));
+               if (z.is_equal(_num30)) // sin(Pi/2)    -> 1
+                       return sign;
+       }
+       
+       if (is_exactly_a<function>(x)) {
+               const ex &t = x.op(0);
+               // sin(asin(x)) -> x
+               if (is_ex_the_function(x, asin))
+                       return t;
+               // sin(acos(x)) -> sqrt(1-x^2)
+               if (is_ex_the_function(x, acos))
+                       return sqrt(_ex1-power(t,_ex2));
+               // sin(atan(x)) -> x/sqrt(1+x^2)
+               if (is_ex_the_function(x, atan))
+                       return t*power(_ex1+power(t,_ex2),_ex_1_2);
+       }
+       
+       // sin(float) -> float
+       if (x.info(info_flags::numeric) && !x.info(info_flags::crational))
+               return sin(ex_to<numeric>(x));
+       
+       return sin(x).hold();
+}
+
+static ex sin_deriv(const ex & x, unsigned deriv_param)
+{
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx sin(x) -> cos(x)
+       return cos(x);
+}
+
+REGISTER_FUNCTION(sin, eval_func(sin_eval).
+                       evalf_func(sin_evalf).
+                       derivative_func(sin_deriv).
+                       latex_name("\\sin"));
 
 //////////
 // cosine (trigonometric function)
 //////////
 
-ex cos_evalf(ex const & x)
-{
-    BEGIN_TYPECHECK
-        TYPECHECK(x,numeric)
-    END_TYPECHECK(cos(x))
-    
-    return cos(ex_to_numeric(x)); // -> numeric cos(numeric)
-}
-
-ex cos_eval(ex const & x)
-{
-    // cos(n*Pi) -> {+|-}1
-    ex xOverPi=x/Pi;
-    if (xOverPi.info(info_flags::even))
-        return exONE();
-    else if (xOverPi.info(info_flags::odd))
-        return exMINUSONE();
-    
-    // cos((2n+1)*Pi/2) -> 0
-    ex xOverPiMinusHalf=xOverPi-exHALF();
-    if (xOverPiMinusHalf.info(info_flags::integer))
-        return exZERO();
-    
-    if (is_ex_exactly_of_type(x, function)) {
-        ex t=x.op(0);
-        // cos(acos(x)) -> x
-        if (is_ex_the_function(x, acos))
-            return t;
-        // cos(asin(x)) -> (1-x^2)^(1/2)
-        if (is_ex_the_function(x, asin))
-            return power(exONE()-power(t,exTWO()),exHALF());
-        // cos(atan(x)) -> (1+x^2)^(-1/2)
-        if (is_ex_the_function(x, atan))
-            return power(exONE()+power(t,exTWO()),exMINUSHALF());
-    }
-    
-    // cos(float) -> float
-    if (x.info(info_flags::numeric) && !x.info(info_flags::rational))
-        return cos_evalf(x);
-    
-    return cos(x).hold();
-}
-
-ex cos_diff(ex const & x, unsigned diff_param)
-{
-    ASSERT(diff_param==0);
-
-    return numMINUSONE()*sin(x);
-}
-
-REGISTER_FUNCTION(cos, cos_eval, cos_evalf, cos_diff, NULL);
+static ex cos_evalf(const ex & x)
+{
+       if (is_exactly_a<numeric>(x))
+               return cos(ex_to<numeric>(x));
+       
+       return cos(x).hold();
+}
+
+static ex cos_eval(const ex & x)
+{
+       // cos(n/d*Pi) -> { all known non-nested radicals }
+       const ex SixtyExOverPi = _ex60*x/Pi;
+       ex sign = _ex1;
+       if (SixtyExOverPi.info(info_flags::integer)) {
+               numeric z = mod(ex_to<numeric>(SixtyExOverPi),_num120);
+               if (z>=_num60) {
+                       // wrap to interval [0, Pi)
+                       z = _num120-z;
+               }
+               if (z>=_num30) {
+                       // wrap to interval [0, Pi/2)
+                       z = _num60-z;
+                       sign = _ex_1;
+               }
+               if (z.is_equal(_num0))  // cos(0)       -> 1
+                       return sign;
+               if (z.is_equal(_num5))  // cos(Pi/12)   -> sqrt(6)/4*(1+sqrt(3)/3)
+                       return sign*_ex1_4*sqrt(_ex6)*(_ex1+_ex1_3*sqrt(_ex3));
+               if (z.is_equal(_num10)) // cos(Pi/6)    -> sqrt(3)/2
+                       return sign*_ex1_2*sqrt(_ex3);
+               if (z.is_equal(_num12)) // cos(Pi/5)    -> sqrt(5)/4+1/4
+                       return sign*(_ex1_4*sqrt(_ex5)+_ex1_4);
+               if (z.is_equal(_num15)) // cos(Pi/4)    -> sqrt(2)/2
+                       return sign*_ex1_2*sqrt(_ex2);
+               if (z.is_equal(_num20)) // cos(Pi/3)    -> 1/2
+                       return sign*_ex1_2;
+               if (z.is_equal(_num24)) // cos(2/5*Pi)  -> sqrt(5)/4-1/4x
+                       return sign*(_ex1_4*sqrt(_ex5)+_ex_1_4);
+               if (z.is_equal(_num25)) // cos(5/12*Pi) -> sqrt(6)/4*(1-sqrt(3)/3)
+                       return sign*_ex1_4*sqrt(_ex6)*(_ex1+_ex_1_3*sqrt(_ex3));
+               if (z.is_equal(_num30)) // cos(Pi/2)    -> 0
+                       return _ex0;
+       }
+       
+       if (is_exactly_a<function>(x)) {
+               const ex &t = x.op(0);
+               // cos(acos(x)) -> x
+               if (is_ex_the_function(x, acos))
+                       return t;
+               // cos(asin(x)) -> sqrt(1-x^2)
+               if (is_ex_the_function(x, asin))
+                       return sqrt(_ex1-power(t,_ex2));
+               // cos(atan(x)) -> 1/sqrt(1+x^2)
+               if (is_ex_the_function(x, atan))
+                       return power(_ex1+power(t,_ex2),_ex_1_2);
+       }
+       
+       // cos(float) -> float
+       if (x.info(info_flags::numeric) && !x.info(info_flags::crational))
+               return cos(ex_to<numeric>(x));
+       
+       return cos(x).hold();
+}
+
+static ex cos_deriv(const ex & x, unsigned deriv_param)
+{
+       GINAC_ASSERT(deriv_param==0);
+
+       // d/dx cos(x) -> -sin(x)
+       return -sin(x);
+}
+
+REGISTER_FUNCTION(cos, eval_func(cos_eval).
+                       evalf_func(cos_evalf).
+                       derivative_func(cos_deriv).
+                       latex_name("\\cos"));
 
 //////////
 // tangent (trigonometric function)
 //////////
 
-ex tan_evalf(ex const & x)
-{
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(tan(x)) // -> numeric tan(numeric)
-    
-    return tan(ex_to_numeric(x));
-}
-
-ex tan_eval(ex const & x)
-{
-    // tan(n*Pi/3) -> {0|3^(1/2)|-(3^(1/2))}
-    ex ThreeExOverPi=numTHREE()*x/Pi;
-    if (ThreeExOverPi.info(info_flags::integer)) {
-        numeric z=mod(ex_to_numeric(ThreeExOverPi),numeric(3));
-        if (z.is_equal(numZERO()))
-            return exZERO();
-        if (z.is_equal(numONE()))
-            return power(exTHREE(),exHALF());
-        if (z.is_equal(numTWO()))
-            return -power(exTHREE(),exHALF());
-    }
-    
-    // tan((2n+1)*Pi/2) -> throw
-    ex ExOverPiMinusHalf=x/Pi-exHALF();
-    if (ExOverPiMinusHalf.info(info_flags::integer))
-        throw (std::domain_error("tan_eval(): infinity"));
-    
-    if (is_ex_exactly_of_type(x, function)) {
-        ex t=x.op(0);
-        // tan(atan(x)) -> x
-        if (is_ex_the_function(x, atan))
-            return t;
-        // tan(asin(x)) -> x*(1+x^2)^(-1/2)
-        if (is_ex_the_function(x, asin))
-            return t*power(exONE()-power(t,exTWO()),exMINUSHALF());
-        // tan(acos(x)) -> (1-x^2)^(1/2)/x
-        if (is_ex_the_function(x, acos))
-            return power(t,exMINUSONE())*power(exONE()-power(t,exTWO()),exHALF());
-    }
-    
-    // tan(float) -> float
-    if (x.info(info_flags::numeric) && !x.info(info_flags::rational)) {
-        return tan_evalf(x);
-    }
-    
-    return tan(x).hold();
-}
-
-ex tan_diff(ex const & x, unsigned diff_param)
-{
-    ASSERT(diff_param==0);
-    
-    return (1+power(tan(x),exTWO()));
-}
-
-REGISTER_FUNCTION(tan, tan_eval, tan_evalf, tan_diff, NULL);
+static ex tan_evalf(const ex & x)
+{
+       if (is_exactly_a<numeric>(x))
+               return tan(ex_to<numeric>(x));
+       
+       return tan(x).hold();
+}
+
+static ex tan_eval(const ex & x)
+{
+       // tan(n/d*Pi) -> { all known non-nested radicals }
+       const ex SixtyExOverPi = _ex60*x/Pi;
+       ex sign = _ex1;
+       if (SixtyExOverPi.info(info_flags::integer)) {
+               numeric z = mod(ex_to<numeric>(SixtyExOverPi),_num60);
+               if (z>=_num60) {
+                       // wrap to interval [0, Pi)
+                       z -= _num60;
+               }
+               if (z>=_num30) {
+                       // wrap to interval [0, Pi/2)
+                       z = _num60-z;
+                       sign = _ex_1;
+               }
+               if (z.is_equal(_num0))  // tan(0)       -> 0
+                       return _ex0;
+               if (z.is_equal(_num5))  // tan(Pi/12)   -> 2-sqrt(3)
+                       return sign*(_ex2-sqrt(_ex3));
+               if (z.is_equal(_num10)) // tan(Pi/6)    -> sqrt(3)/3
+                       return sign*_ex1_3*sqrt(_ex3);
+               if (z.is_equal(_num15)) // tan(Pi/4)    -> 1
+                       return sign;
+               if (z.is_equal(_num20)) // tan(Pi/3)    -> sqrt(3)
+                       return sign*sqrt(_ex3);
+               if (z.is_equal(_num25)) // tan(5/12*Pi) -> 2+sqrt(3)
+                       return sign*(sqrt(_ex3)+_ex2);
+               if (z.is_equal(_num30)) // tan(Pi/2)    -> infinity
+                       throw (pole_error("tan_eval(): simple pole",1));
+       }
+       
+       if (is_exactly_a<function>(x)) {
+               const ex &t = x.op(0);
+               // tan(atan(x)) -> x
+               if (is_ex_the_function(x, atan))
+                       return t;
+               // tan(asin(x)) -> x/sqrt(1+x^2)
+               if (is_ex_the_function(x, asin))
+                       return t*power(_ex1-power(t,_ex2),_ex_1_2);
+               // tan(acos(x)) -> sqrt(1-x^2)/x
+               if (is_ex_the_function(x, acos))
+                       return power(t,_ex_1)*sqrt(_ex1-power(t,_ex2));
+       }
+       
+       // tan(float) -> float
+       if (x.info(info_flags::numeric) && !x.info(info_flags::crational)) {
+               return tan(ex_to<numeric>(x));
+       }
+       
+       return tan(x).hold();
+}
+
+static ex tan_deriv(const ex & x, unsigned deriv_param)
+{
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx tan(x) -> 1+tan(x)^2;
+       return (_ex1+power(tan(x),_ex2));
+}
+
+static ex tan_series(const ex &x,
+                     const relational &rel,
+                     int order,
+                     unsigned options)
+{
+       GINAC_ASSERT(is_a<symbol>(rel.lhs()));
+       // method:
+       // Taylor series where there is no pole falls back to tan_deriv.
+       // On a pole simply expand sin(x)/cos(x).
+       const ex x_pt = x.subs(rel, subs_options::no_pattern);
+       if (!(2*x_pt/Pi).info(info_flags::odd))
+               throw do_taylor();  // caught by function::series()
+       // if we got here we have to care for a simple pole
+       return (sin(x)/cos(x)).series(rel, order+2, options);
+}
+
+REGISTER_FUNCTION(tan, eval_func(tan_eval).
+                       evalf_func(tan_evalf).
+                       derivative_func(tan_deriv).
+                       series_func(tan_series).
+                       latex_name("\\tan"));
 
 //////////
 // inverse sine (arc sine)
 //////////
 
-ex asin_evalf(ex const & x)
+static ex asin_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(asin(x))
-    
-    return asin(ex_to_numeric(x)); // -> numeric asin(numeric)
+       if (is_exactly_a<numeric>(x))
+               return asin(ex_to<numeric>(x));
+       
+       return asin(x).hold();
 }
 
-ex asin_eval(ex const & x)
+static ex asin_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // asin(0) -> 0
-        if (x.is_zero())
-            return x;
-        // asin(1/2) -> Pi/6
-        if (x.is_equal(exHALF()))
-            return numeric(1,6)*Pi;
-        // asin(1) -> Pi/2
-        if (x.is_equal(exONE()))
-            return numeric(1,2)*Pi;
-        // asin(-1/2) -> -Pi/6
-        if (x.is_equal(exMINUSHALF()))
-            return numeric(-1,6)*Pi;
-        // asin(-1) -> -Pi/2
-        if (x.is_equal(exMINUSONE()))
-            return numeric(-1,2)*Pi;
-        // asin(float) -> float
-        if (!x.info(info_flags::rational))
-            return asin_evalf(x);
-    }
-    
-    return asin(x).hold();
+       if (x.info(info_flags::numeric)) {
+               // asin(0) -> 0
+               if (x.is_zero())
+                       return x;
+               // asin(1/2) -> Pi/6
+               if (x.is_equal(_ex1_2))
+                       return numeric(1,6)*Pi;
+               // asin(1) -> Pi/2
+               if (x.is_equal(_ex1))
+                       return _num1_2*Pi;
+               // asin(-1/2) -> -Pi/6
+               if (x.is_equal(_ex_1_2))
+                       return numeric(-1,6)*Pi;
+               // asin(-1) -> -Pi/2
+               if (x.is_equal(_ex_1))
+                       return _num_1_2*Pi;
+               // asin(float) -> float
+               if (!x.info(info_flags::crational))
+                       return asin(ex_to<numeric>(x));
+       }
+       
+       return asin(x).hold();
 }
 
-ex asin_diff(ex const & x, unsigned diff_param)
+static ex asin_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return power(1-power(x,exTWO()),exMINUSHALF());
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx asin(x) -> 1/sqrt(1-x^2)
+       return power(1-power(x,_ex2),_ex_1_2);
 }
 
-REGISTER_FUNCTION(asin, asin_eval, asin_evalf, asin_diff, NULL);
+REGISTER_FUNCTION(asin, eval_func(asin_eval).
+                        evalf_func(asin_evalf).
+                        derivative_func(asin_deriv).
+                        latex_name("\\arcsin"));
 
 //////////
 // inverse cosine (arc cosine)
 //////////
 
-ex acos_evalf(ex const & x)
+static ex acos_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(acos(x))
-    
-    return acos(ex_to_numeric(x)); // -> numeric acos(numeric)
+       if (is_exactly_a<numeric>(x))
+               return acos(ex_to<numeric>(x));
+       
+       return acos(x).hold();
 }
 
-ex acos_eval(ex const & x)
+static ex acos_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // acos(1) -> 0
-        if (x.is_equal(exONE()))
-            return exZERO();
-        // acos(1/2) -> Pi/3
-        if (x.is_equal(exHALF()))
-            return numeric(1,3)*Pi;
-        // acos(0) -> Pi/2
-        if (x.is_zero())
-            return numeric(1,2)*Pi;
-        // acos(-1/2) -> 2/3*Pi
-        if (x.is_equal(exMINUSHALF()))
-            return numeric(2,3)*Pi;
-        // acos(-1) -> Pi
-        if (x.is_equal(exMINUSONE()))
-            return Pi;
-        // acos(float) -> float
-        if (!x.info(info_flags::rational))
-            return acos_evalf(x);
-    }
-    
-    return acos(x).hold();
+       if (x.info(info_flags::numeric)) {
+               // acos(1) -> 0
+               if (x.is_equal(_ex1))
+                       return _ex0;
+               // acos(1/2) -> Pi/3
+               if (x.is_equal(_ex1_2))
+                       return _ex1_3*Pi;
+               // acos(0) -> Pi/2
+               if (x.is_zero())
+                       return _ex1_2*Pi;
+               // acos(-1/2) -> 2/3*Pi
+               if (x.is_equal(_ex_1_2))
+                       return numeric(2,3)*Pi;
+               // acos(-1) -> Pi
+               if (x.is_equal(_ex_1))
+                       return Pi;
+               // acos(float) -> float
+               if (!x.info(info_flags::crational))
+                       return acos(ex_to<numeric>(x));
+       }
+       
+       return acos(x).hold();
 }
 
-ex acos_diff(ex const & x, unsigned diff_param)
+static ex acos_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return numMINUSONE()*power(1-power(x,exTWO()),exMINUSHALF());
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx acos(x) -> -1/sqrt(1-x^2)
+       return -power(1-power(x,_ex2),_ex_1_2);
 }
 
-REGISTER_FUNCTION(acos, acos_eval, acos_evalf, acos_diff, NULL);
+REGISTER_FUNCTION(acos, eval_func(acos_eval).
+                        evalf_func(acos_evalf).
+                        derivative_func(acos_deriv).
+                        latex_name("\\arccos"));
 
 //////////
 // inverse tangent (arc tangent)
 //////////
 
-ex atan_evalf(ex const & x)
+static ex atan_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-        TYPECHECK(x,numeric)
-    END_TYPECHECK(atan(x))
-    
-    return atan(ex_to_numeric(x)); // -> numeric atan(numeric)
+       if (is_exactly_a<numeric>(x))
+               return atan(ex_to<numeric>(x));
+       
+       return atan(x).hold();
 }
 
-ex atan_eval(ex const & x)
+static ex atan_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // atan(0) -> 0
-        if (x.is_equal(exZERO()))
-            return exZERO();
-        // atan(float) -> float
-        if (!x.info(info_flags::rational))
-            return atan_evalf(x);
-    }
-    
-    return atan(x).hold();
-}    
+       if (x.info(info_flags::numeric)) {
+               // atan(0) -> 0
+               if (x.is_zero())
+                       return _ex0;
+               // atan(1) -> Pi/4
+               if (x.is_equal(_ex1))
+                       return _ex1_4*Pi;
+               // atan(-1) -> -Pi/4
+               if (x.is_equal(_ex_1))
+                       return _ex_1_4*Pi;
+               if (x.is_equal(I) || x.is_equal(-I))
+                       throw (pole_error("atan_eval(): logarithmic pole",0));
+               // atan(float) -> float
+               if (!x.info(info_flags::crational))
+                       return atan(ex_to<numeric>(x));
+       }
+       
+       return atan(x).hold();
+}
 
-ex atan_diff(ex const & x, unsigned diff_param)
+static ex atan_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
+       GINAC_ASSERT(deriv_param==0);
+
+       // d/dx atan(x) -> 1/(1+x^2)
+       return power(_ex1+power(x,_ex2), _ex_1);
+}
 
-    return power(1+x*x, -1);
+static ex atan_series(const ex &arg,
+                      const relational &rel,
+                      int order,
+                      unsigned options)
+{
+       GINAC_ASSERT(is_a<symbol>(rel.lhs()));
+       // method:
+       // Taylor series where there is no pole or cut falls back to atan_deriv.
+       // There are two branch cuts, one runnig from I up the imaginary axis and
+       // one running from -I down the imaginary axis.  The points I and -I are
+       // poles.
+       // On the branch cuts and the poles series expand
+       //     (log(1+I*x)-log(1-I*x))/(2*I)
+       // instead.
+       const ex arg_pt = arg.subs(rel, subs_options::no_pattern);
+       if (!(I*arg_pt).info(info_flags::real))
+               throw do_taylor();     // Re(x) != 0
+       if ((I*arg_pt).info(info_flags::real) && abs(I*arg_pt)<_ex1)
+               throw do_taylor();     // Re(x) == 0, but abs(x)<1
+       // care for the poles, using the defining formula for atan()...
+       if (arg_pt.is_equal(I) || arg_pt.is_equal(-I))
+               return ((log(1+I*arg)-log(1-I*arg))/(2*I)).series(rel, order, options);
+       if (!(options & series_options::suppress_branchcut)) {
+               // method:
+               // This is the branch cut: assemble the primitive series manually and
+               // then add the corresponding complex step function.
+               const symbol &s = ex_to<symbol>(rel.lhs());
+               const ex &point = rel.rhs();
+               const symbol foo;
+               const ex replarg = series(atan(arg), s==foo, order).subs(foo==point, subs_options::no_pattern);
+               ex Order0correction = replarg.op(0)+csgn(arg)*Pi*_ex_1_2;
+               if ((I*arg_pt)<_ex0)
+                       Order0correction += log((I*arg_pt+_ex_1)/(I*arg_pt+_ex1))*I*_ex_1_2;
+               else
+                       Order0correction += log((I*arg_pt+_ex1)/(I*arg_pt+_ex_1))*I*_ex1_2;
+               epvector seq;
+               seq.push_back(expair(Order0correction, _ex0));
+               seq.push_back(expair(Order(_ex1), order));
+               return series(replarg - pseries(rel, seq), rel, order);
+       }
+       throw do_taylor();
 }
 
-REGISTER_FUNCTION(atan, atan_eval, atan_evalf, atan_diff, NULL);
+REGISTER_FUNCTION(atan, eval_func(atan_eval).
+                        evalf_func(atan_evalf).
+                        derivative_func(atan_deriv).
+                        series_func(atan_series).
+                        latex_name("\\arctan"));
 
 //////////
 // inverse tangent (atan2(y,x))
 //////////
 
-ex atan2_evalf(ex const & y, ex const & x)
+static ex atan2_evalf(const ex &y, const ex &x)
 {
-    BEGIN_TYPECHECK
-        TYPECHECK(y,numeric)
-        TYPECHECK(x,numeric)
-    END_TYPECHECK(atan2(y,x))
-    
-    return atan(ex_to_numeric(y),ex_to_numeric(x)); // -> numeric atan(numeric)
+       if (is_exactly_a<numeric>(y) && is_exactly_a<numeric>(x))
+               return atan2(ex_to<numeric>(y), ex_to<numeric>(x));
+       
+       return atan2(y, x).hold();
 }
 
-ex atan2_eval(ex const & y, ex const & x)
+static ex atan2_eval(const ex & y, const ex & x)
 {
-    if (y.info(info_flags::numeric) && !y.info(info_flags::rational) &&
-        x.info(info_flags::numeric) && !x.info(info_flags::rational)) {
-        return atan2_evalf(y,x);
-    }
-    
-    return atan2(y,x).hold();
+       if (y.info(info_flags::numeric) && !y.info(info_flags::crational) &&
+               x.info(info_flags::numeric) && !x.info(info_flags::crational)) {
+               return atan2_evalf(y,x);
+       }
+       
+       return atan2(y,x).hold();
 }    
 
-ex atan2_diff(ex const & y, ex const & x, unsigned diff_param)
+static ex atan2_deriv(const ex & y, const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param<2);
-
-    if (diff_param==0) {
-        // d/dy atan(y,x)
-        return power(x*(1+y*y/(x*x)),-1);
-    }
-    // d/dx atan(y,x)
-    return -y*power(x*x+y*y,-1);
+       GINAC_ASSERT(deriv_param<2);
+       
+       if (deriv_param==0) {
+               // d/dy atan(y,x)
+               return x*power(power(x,_ex2)+power(y,_ex2),_ex_1);
+       }
+       // d/dx atan(y,x)
+       return -y*power(power(x,_ex2)+power(y,_ex2),_ex_1);
 }
 
-REGISTER_FUNCTION(atan2, atan2_eval, atan2_evalf, atan2_diff, NULL);
+REGISTER_FUNCTION(atan2, eval_func(atan2_eval).
+                         evalf_func(atan2_evalf).
+                         derivative_func(atan2_deriv));
 
 //////////
 // hyperbolic sine (trigonometric function)
 //////////
 
-ex sinh_evalf(ex const & x)
+static ex sinh_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(sinh(x))
-    
-    return sinh(ex_to_numeric(x)); // -> numeric sinh(numeric)
+       if (is_exactly_a<numeric>(x))
+               return sinh(ex_to<numeric>(x));
+       
+       return sinh(x).hold();
 }
 
-ex sinh_eval(ex const & x)
+static ex sinh_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // sinh(0) -> 0
-        if (x.is_zero())
-            return exZERO();
-        // sinh(float) -> float
-        if (!x.info(info_flags::rational))
-            return sinh_evalf(x);
-    }
-    
-    if (is_ex_exactly_of_type(x, function)) {
-        ex t=x.op(0);
-        // sinh(asinh(x)) -> x
-        if (is_ex_the_function(x, asinh))
-            return t;
-        // sinh(acosh(x)) -> (x-1)^(1/2) * (x+1)^(1/2)
-        if (is_ex_the_function(x, acosh))
-            return power(t-exONE(),exHALF())*power(t+exONE(),exHALF());
-        // sinh(atanh(x)) -> x*(1-x^2)^(-1/2)
-        if (is_ex_the_function(x, atanh))
-            return t*power(exONE()-power(t,exTWO()),exMINUSHALF());
-    }
-    
-    return sinh(x).hold();
+       if (x.info(info_flags::numeric)) {
+               if (x.is_zero())  // sinh(0) -> 0
+                       return _ex0;        
+               if (!x.info(info_flags::crational))  // sinh(float) -> float
+                       return sinh(ex_to<numeric>(x));
+       }
+       
+       if ((x/Pi).info(info_flags::numeric) &&
+               ex_to<numeric>(x/Pi).real().is_zero())  // sinh(I*x) -> I*sin(x)
+               return I*sin(x/I);
+       
+       if (is_exactly_a<function>(x)) {
+               const ex &t = x.op(0);
+               // sinh(asinh(x)) -> x
+               if (is_ex_the_function(x, asinh))
+                       return t;
+               // sinh(acosh(x)) -> sqrt(x-1) * sqrt(x+1)
+               if (is_ex_the_function(x, acosh))
+                       return sqrt(t-_ex1)*sqrt(t+_ex1);
+               // sinh(atanh(x)) -> x/sqrt(1-x^2)
+               if (is_ex_the_function(x, atanh))
+                       return t*power(_ex1-power(t,_ex2),_ex_1_2);
+       }
+       
+       return sinh(x).hold();
 }
 
-ex sinh_diff(ex const & x, unsigned diff_param)
+static ex sinh_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return cosh(x);
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx sinh(x) -> cosh(x)
+       return cosh(x);
 }
 
-REGISTER_FUNCTION(sinh, sinh_eval, sinh_evalf, sinh_diff, NULL);
+REGISTER_FUNCTION(sinh, eval_func(sinh_eval).
+                        evalf_func(sinh_evalf).
+                        derivative_func(sinh_deriv).
+                        latex_name("\\sinh"));
 
 //////////
 // hyperbolic cosine (trigonometric function)
 //////////
 
-ex cosh_evalf(ex const & x)
+static ex cosh_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(cosh(x))
-    
-    return cosh(ex_to_numeric(x)); // -> numeric cosh(numeric)
+       if (is_exactly_a<numeric>(x))
+               return cosh(ex_to<numeric>(x));
+       
+       return cosh(x).hold();
 }
 
-ex cosh_eval(ex const & x)
+static ex cosh_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // cosh(0) -> 1
-        if (x.is_zero())
-            return exONE();
-        // cosh(float) -> float
-        if (!x.info(info_flags::rational))
-            return cosh_evalf(x);
-    }
-    
-    if (is_ex_exactly_of_type(x, function)) {
-        ex t=x.op(0);
-        // cosh(acosh(x)) -> x
-        if (is_ex_the_function(x, acosh))
-            return t;
-        // cosh(asinh(x)) -> (1+x^2)^(1/2)
-        if (is_ex_the_function(x, asinh))
-            return power(exONE()+power(t,exTWO()),exHALF());
-        // cosh(atanh(x)) -> (1-x^2)^(-1/2)
-        if (is_ex_the_function(x, atanh))
-            return power(exONE()-power(t,exTWO()),exMINUSHALF());
-    }
-    
-    return cosh(x).hold();
+       if (x.info(info_flags::numeric)) {
+               if (x.is_zero())  // cosh(0) -> 1
+                       return _ex1;
+               if (!x.info(info_flags::crational))  // cosh(float) -> float
+                       return cosh(ex_to<numeric>(x));
+       }
+       
+       if ((x/Pi).info(info_flags::numeric) &&
+               ex_to<numeric>(x/Pi).real().is_zero())  // cosh(I*x) -> cos(x)
+               return cos(x/I);
+       
+       if (is_exactly_a<function>(x)) {
+               const ex &t = x.op(0);
+               // cosh(acosh(x)) -> x
+               if (is_ex_the_function(x, acosh))
+                       return t;
+               // cosh(asinh(x)) -> sqrt(1+x^2)
+               if (is_ex_the_function(x, asinh))
+                       return sqrt(_ex1+power(t,_ex2));
+               // cosh(atanh(x)) -> 1/sqrt(1-x^2)
+               if (is_ex_the_function(x, atanh))
+                       return power(_ex1-power(t,_ex2),_ex_1_2);
+       }
+       
+       return cosh(x).hold();
 }
 
-ex cosh_diff(ex const & x, unsigned diff_param)
+static ex cosh_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return sinh(x);
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx cosh(x) -> sinh(x)
+       return sinh(x);
 }
 
-REGISTER_FUNCTION(cosh, cosh_eval, cosh_evalf, cosh_diff, NULL);
+REGISTER_FUNCTION(cosh, eval_func(cosh_eval).
+                        evalf_func(cosh_evalf).
+                        derivative_func(cosh_deriv).
+                        latex_name("\\cosh"));
 
 //////////
 // hyperbolic tangent (trigonometric function)
 //////////
 
-ex tanh_evalf(ex const & x)
+static ex tanh_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(tanh(x))
-    
-    return tanh(ex_to_numeric(x)); // -> numeric tanh(numeric)
+       if (is_exactly_a<numeric>(x))
+               return tanh(ex_to<numeric>(x));
+       
+       return tanh(x).hold();
 }
 
-ex tanh_eval(ex const & x)
+static ex tanh_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // tanh(0) -> 0
-        if (x.is_zero())
-            return exZERO();
-        // tanh(float) -> float
-        if (!x.info(info_flags::rational))
-            return tanh_evalf(x);
-    }
-    
-    if (is_ex_exactly_of_type(x, function)) {
-        ex t=x.op(0);
-        // tanh(atanh(x)) -> x
-        if (is_ex_the_function(x, atanh))
-            return t;
-        // tanh(asinh(x)) -> x*(1+x^2)^(-1/2)
-        if (is_ex_the_function(x, asinh))
-            return t*power(exONE()+power(t,exTWO()),exMINUSHALF());
-        // tanh(acosh(x)) -> (x-1)^(1/2)*(x+1)^(1/2)/x
-        if (is_ex_the_function(x, acosh))
-            return power(t-exONE(),exHALF())*power(t+exONE(),exHALF())*power(t,exMINUSONE());
-    }
-    
-    return tanh(x).hold();
+       if (x.info(info_flags::numeric)) {
+               if (x.is_zero())  // tanh(0) -> 0
+                       return _ex0;
+               if (!x.info(info_flags::crational))  // tanh(float) -> float
+                       return tanh(ex_to<numeric>(x));
+       }
+       
+       if ((x/Pi).info(info_flags::numeric) &&
+               ex_to<numeric>(x/Pi).real().is_zero())  // tanh(I*x) -> I*tan(x);
+               return I*tan(x/I);
+       
+       if (is_exactly_a<function>(x)) {
+               const ex &t = x.op(0);
+               // tanh(atanh(x)) -> x
+               if (is_ex_the_function(x, atanh))
+                       return t;
+               // tanh(asinh(x)) -> x/sqrt(1+x^2)
+               if (is_ex_the_function(x, asinh))
+                       return t*power(_ex1+power(t,_ex2),_ex_1_2);
+               // tanh(acosh(x)) -> sqrt(x-1)*sqrt(x+1)/x
+               if (is_ex_the_function(x, acosh))
+                       return sqrt(t-_ex1)*sqrt(t+_ex1)*power(t,_ex_1);
+       }
+       
+       return tanh(x).hold();
 }
 
-ex tanh_diff(ex const & x, unsigned diff_param)
+static ex tanh_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return exONE()-power(tanh(x),exTWO());
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx tanh(x) -> 1-tanh(x)^2
+       return _ex1-power(tanh(x),_ex2);
 }
 
-REGISTER_FUNCTION(tanh, tanh_eval, tanh_evalf, tanh_diff, NULL);
+static ex tanh_series(const ex &x,
+                      const relational &rel,
+                      int order,
+                      unsigned options)
+{
+       GINAC_ASSERT(is_a<symbol>(rel.lhs()));
+       // method:
+       // Taylor series where there is no pole falls back to tanh_deriv.
+       // On a pole simply expand sinh(x)/cosh(x).
+       const ex x_pt = x.subs(rel, subs_options::no_pattern);
+       if (!(2*I*x_pt/Pi).info(info_flags::odd))
+               throw do_taylor();  // caught by function::series()
+       // if we got here we have to care for a simple pole
+       return (sinh(x)/cosh(x)).series(rel, order+2, options);
+}
+
+REGISTER_FUNCTION(tanh, eval_func(tanh_eval).
+                        evalf_func(tanh_evalf).
+                        derivative_func(tanh_deriv).
+                        series_func(tanh_series).
+                        latex_name("\\tanh"));
 
 //////////
 // inverse hyperbolic sine (trigonometric function)
 //////////
 
-ex asinh_evalf(ex const & x)
+static ex asinh_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(asinh(x))
-    
-    return asinh(ex_to_numeric(x)); // -> numeric asinh(numeric)
+       if (is_exactly_a<numeric>(x))
+               return asinh(ex_to<numeric>(x));
+       
+       return asinh(x).hold();
 }
 
-ex asinh_eval(ex const & x)
+static ex asinh_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // asinh(0) -> 0
-        if (x.is_zero())
-            return exZERO();
-        // asinh(float) -> float
-        if (!x.info(info_flags::rational))
-            return asinh_evalf(x);
-    }
-    
-    return asinh(x).hold();
+       if (x.info(info_flags::numeric)) {
+               // asinh(0) -> 0
+               if (x.is_zero())
+                       return _ex0;
+               // asinh(float) -> float
+               if (!x.info(info_flags::crational))
+                       return asinh(ex_to<numeric>(x));
+       }
+       
+       return asinh(x).hold();
 }
 
-ex asinh_diff(ex const & x, unsigned diff_param)
+static ex asinh_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return power(1+power(x,exTWO()),exMINUSHALF());
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx asinh(x) -> 1/sqrt(1+x^2)
+       return power(_ex1+power(x,_ex2),_ex_1_2);
 }
 
-REGISTER_FUNCTION(asinh, asinh_eval, asinh_evalf, asinh_diff, NULL);
+REGISTER_FUNCTION(asinh, eval_func(asinh_eval).
+                         evalf_func(asinh_evalf).
+                         derivative_func(asinh_deriv));
 
 //////////
 // inverse hyperbolic cosine (trigonometric function)
 //////////
 
-ex acosh_evalf(ex const & x)
+static ex acosh_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(acosh(x))
-    
-    return acosh(ex_to_numeric(x)); // -> numeric acosh(numeric)
+       if (is_exactly_a<numeric>(x))
+               return acosh(ex_to<numeric>(x));
+       
+       return acosh(x).hold();
 }
 
-ex acosh_eval(ex const & x)
+static ex acosh_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // acosh(0) -> Pi*I/2
-        if (x.is_zero())
-            return Pi*I*numeric(1,2);
-        // acosh(1) -> 0
-        if (x.is_equal(exONE()))
-            return exZERO();
-        // acosh(-1) -> Pi*I
-        if (x.is_equal(exMINUSONE()))
-            return Pi*I;
-        // acosh(float) -> float
-        if (!x.info(info_flags::rational))
-            return acosh_evalf(x);
-    }
-    
-    return acosh(x).hold();
+       if (x.info(info_flags::numeric)) {
+               // acosh(0) -> Pi*I/2
+               if (x.is_zero())
+                       return Pi*I*numeric(1,2);
+               // acosh(1) -> 0
+               if (x.is_equal(_ex1))
+                       return _ex0;
+               // acosh(-1) -> Pi*I
+               if (x.is_equal(_ex_1))
+                       return Pi*I;
+               // acosh(float) -> float
+               if (!x.info(info_flags::crational))
+                       return acosh(ex_to<numeric>(x));
+       }
+       
+       return acosh(x).hold();
 }
 
-ex acosh_diff(ex const & x, unsigned diff_param)
+static ex acosh_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return power(x-1,exMINUSHALF())*power(x+1,exMINUSHALF());
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx acosh(x) -> 1/(sqrt(x-1)*sqrt(x+1))
+       return power(x+_ex_1,_ex_1_2)*power(x+_ex1,_ex_1_2);
 }
 
-REGISTER_FUNCTION(acosh, acosh_eval, acosh_evalf, acosh_diff, NULL);
+REGISTER_FUNCTION(acosh, eval_func(acosh_eval).
+                         evalf_func(acosh_evalf).
+                         derivative_func(acosh_deriv));
 
 //////////
 // inverse hyperbolic tangent (trigonometric function)
 //////////
 
-ex atanh_evalf(ex const & x)
+static ex atanh_evalf(const ex & x)
 {
-    BEGIN_TYPECHECK
-       TYPECHECK(x,numeric)
-    END_TYPECHECK(atanh(x))
-    
-    return atanh(ex_to_numeric(x)); // -> numeric atanh(numeric)
+       if (is_exactly_a<numeric>(x))
+               return atanh(ex_to<numeric>(x));
+       
+       return atanh(x).hold();
 }
 
-ex atanh_eval(ex const & x)
+static ex atanh_eval(const ex & x)
 {
-    if (x.info(info_flags::numeric)) {
-        // atanh(0) -> 0
-        if (x.is_zero())
-            return exZERO();
-        // atanh({+|-}1) -> throw
-        if (x.is_equal(exONE()) || x.is_equal(exONE()))
-            throw (std::domain_error("atanh_eval(): infinity"));
-        // atanh(float) -> float
-        if (!x.info(info_flags::rational))
-            return atanh_evalf(x);
-    }
-    
-    return atanh(x).hold();
+       if (x.info(info_flags::numeric)) {
+               // atanh(0) -> 0
+               if (x.is_zero())
+                       return _ex0;
+               // atanh({+|-}1) -> throw
+               if (x.is_equal(_ex1) || x.is_equal(_ex_1))
+                       throw (pole_error("atanh_eval(): logarithmic pole",0));
+               // atanh(float) -> float
+               if (!x.info(info_flags::crational))
+                       return atanh(ex_to<numeric>(x));
+       }
+       
+       return atanh(x).hold();
 }
 
-ex atanh_diff(ex const & x, unsigned diff_param)
+static ex atanh_deriv(const ex & x, unsigned deriv_param)
 {
-    ASSERT(diff_param==0);
-    
-    return power(exONE()-power(x,exTWO()),exMINUSONE());
+       GINAC_ASSERT(deriv_param==0);
+       
+       // d/dx atanh(x) -> 1/(1-x^2)
+       return power(_ex1-power(x,_ex2),_ex_1);
 }
 
-REGISTER_FUNCTION(atanh, atanh_eval, atanh_evalf, atanh_diff, NULL);
+static ex atanh_series(const ex &arg,
+                       const relational &rel,
+                       int order,
+                       unsigned options)
+{
+       GINAC_ASSERT(is_a<symbol>(rel.lhs()));
+       // method:
+       // Taylor series where there is no pole or cut falls back to atanh_deriv.
+       // There are two branch cuts, one runnig from 1 up the real axis and one
+       // one running from -1 down the real axis.  The points 1 and -1 are poles
+       // On the branch cuts and the poles series expand
+       //     (log(1+x)-log(1-x))/2
+       // instead.
+       const ex arg_pt = arg.subs(rel, subs_options::no_pattern);
+       if (!(arg_pt).info(info_flags::real))
+               throw do_taylor();     // Im(x) != 0
+       if ((arg_pt).info(info_flags::real) && abs(arg_pt)<_ex1)
+               throw do_taylor();     // Im(x) == 0, but abs(x)<1
+       // care for the poles, using the defining formula for atanh()...
+       if (arg_pt.is_equal(_ex1) || arg_pt.is_equal(_ex_1))
+               return ((log(_ex1+arg)-log(_ex1-arg))*_ex1_2).series(rel, order, options);
+       // ...and the branch cuts (the discontinuity at the cut being just I*Pi)
+       if (!(options & series_options::suppress_branchcut)) {
+               // method:
+               // This is the branch cut: assemble the primitive series manually and
+               // then add the corresponding complex step function.
+               const symbol &s = ex_to<symbol>(rel.lhs());
+               const ex &point = rel.rhs();
+               const symbol foo;
+               const ex replarg = series(atanh(arg), s==foo, order).subs(foo==point, subs_options::no_pattern);
+               ex Order0correction = replarg.op(0)+csgn(I*arg)*Pi*I*_ex1_2;
+               if (arg_pt<_ex0)
+                       Order0correction += log((arg_pt+_ex_1)/(arg_pt+_ex1))*_ex1_2;
+               else
+                       Order0correction += log((arg_pt+_ex1)/(arg_pt+_ex_1))*_ex_1_2;
+               epvector seq;
+               seq.push_back(expair(Order0correction, _ex0));
+               seq.push_back(expair(Order(_ex1), order));
+               return series(replarg - pseries(rel, seq), rel, order);
+       }
+       throw do_taylor();
+}
+
+REGISTER_FUNCTION(atanh, eval_func(atanh_eval).
+                         evalf_func(atanh_evalf).
+                         derivative_func(atanh_deriv).
+                         series_func(atanh_series));
+
+
+} // namespace GiNaC