When autoloading a Maxima script, avoid clobbering global state.
[maxima.git] / tests / rtest_elliptic.mac
blob9051b1fc0e0cf27713922299d274ebe64737b0e3
1 /* Tests of Jacobi elliptic functions and elliptic integrals */
3 kill(all);
4 done$
6 /* derivatives */
7 diff(jacobi_sn(u,m),u);
8 jacobi_cn(u,m)*jacobi_dn(u,m);
10 diff(jacobi_sn(u,m),m);
11 jacobi_cn(u,m)*jacobi_dn(u,m)*(u-elliptic_e(asin(jacobi_sn(u,m)),m)/(1-m))
12   /(2*m)
13  +jacobi_cn(u,m)^2*jacobi_sn(u,m)/(2*(1-m));
15 diff(jacobi_cn(u,m),u);
16 -jacobi_dn(u,m)*jacobi_sn(u,m);
18 diff(jacobi_cn(u,m),m);
19 -(jacobi_dn(u,m)*jacobi_sn(u,m)*(u-elliptic_e(asin(jacobi_sn(u,m)),m)/(1-m))/(2*m))
20   -(jacobi_cn(u,m)*jacobi_sn(u,m)^2/(2*(1-m)));
22 diff(jacobi_dn(u,m),u);
23 -m*jacobi_cn(u,m)*jacobi_sn(u,m);
25 diff(jacobi_dn(u,m),m);
26 -(jacobi_cn(u,m)*jacobi_sn(u,m)*(u-elliptic_e(asin(jacobi_sn(u,m)),m)/(1-m))/2)
27   -(jacobi_dn(u,m)*'jacobi_sn(u,m)^2/(2*(1-m)));
29 diff(inverse_jacobi_sn(u,m),u);
30 1/(sqrt(1-u^2)*sqrt(1-m*u^2));
32 diff(inverse_jacobi_sn(u,m),m);
33 ((elliptic_e(asin(u),m)-(1-m)*elliptic_f(asin(u),m))/m-(u*sqrt(1-u^2)/sqrt(1-m*u^2)))/(1-m);
35 diff(inverse_jacobi_cn(u,m),u);
36 -(1/(sqrt(1-u^2)*sqrt(m*u^2-m+1)));
38 diff(inverse_jacobi_cn(u,m),m);
39 ((elliptic_e(asin(sqrt(1-u^2)),m)-(1-m)*elliptic_f(asin(sqrt(1-u^2)),m))/m
40    -(sqrt(1-u^2)*abs(u)/sqrt(1-m*(1-u^2))))/(1-m);
42 diff(inverse_jacobi_dn(u,m),u);
43 1/(sqrt(1-u^2)*sqrt(u^2+m-1));
45 diff(inverse_jacobi_dn(u,m),m);
46 ((elliptic_e(asin(sqrt(1-u^2)/sqrt(m)),m)-(1-m)*elliptic_f(asin(sqrt(1-u^2)/sqrt(m)),m))/m
47   -sqrt(1-(1-u^2)/m)*sqrt(1-u^2)/(sqrt(m)*abs(u)))/(1-m)
48 -(sqrt(1-u^2)/(2*m^(3/2)*sqrt(1-(1-u^2)/m)*abs(u)));
50 diff(elliptic_e(phi,m),phi);
51 sqrt(1-m*sin(phi)^2);
53 diff(elliptic_e(phi,m),m);
54 (elliptic_e(phi,m)-elliptic_f(phi,m))/(2*m);
56 diff(elliptic_f(phi,m),phi);
57 1/sqrt(1-m*sin(phi)^2);
59 diff(elliptic_f(phi,m),m);
60 ((elliptic_e(phi,m)-(1-m)*elliptic_f(phi,m))/m
61    -(cos(phi)*sin(phi)/sqrt(1-m*sin(phi)^2)))
62  /(2*(1-m));
64 diff(elliptic_pi(n,phi,m),n);
65 ((-(n*sqrt(1-m*sin(phi)^2)*sin(2*phi))/(2*(1-n*sin(phi)^2)))
66  +((m-n)*elliptic_f(phi,m))/n+elliptic_e(phi,m)
67  +((n^2-m)*elliptic_pi(n,phi,m))/n)
68  /(2*(m-n)*(n-1));
70 diff(elliptic_pi(n,phi,m),phi);
71 1/(sqrt(1-m*sin(phi)^2)*(1-n*sin(phi)^2));
73 diff(elliptic_pi(n,phi,m),m);
74 ((-(m*sin(2*phi))/(2*(m-1)*sqrt(1-m*sin(phi)^2)))
75  +elliptic_e(phi,m)/(m-1)+elliptic_pi(n,phi,m))
76  /(2*(n-m));
78 diff(elliptic_kc(m),m);
79 (elliptic_ec(m)-(1-m)*elliptic_kc(m))/(2*(1-m)*m);
81 diff(elliptic_ec(m),m);
82 (elliptic_ec(m)-elliptic_kc(m))/(2*m);
84 /* Integrals */
86 integrate(jacobi_sn(u,m),u); /* A&S 16.24.1 */
87 log(jacobi_dn(u,m)-sqrt(m)*jacobi_cn(u,m))/sqrt(m);
89 integrate(jacobi_cn(u,m),u); /* A&S 16.24.2 */
90 acos(jacobi_dn(u,m))/sqrt(m);
92 integrate(jacobi_dn(u,m),u); /* A&S 16.24.3 */
93 asin(jacobi_sn(u,m));
95 integrate(jacobi_cd(u,m),u); /* A&S 16.24.4 */
96 log(sqrt(m)*jacobi_sd(u,m)+jacobi_nd(u,m))/sqrt(m);
98 /* Use functions.wolfram.com 09.35.21.001.01, not A&S 16.24.5 */
99 integrate(jacobi_sd(u,m),u);
100 -(sqrt(1-m*jacobi_cd(u,m)^2)*jacobi_dn(u,m)*asin(sqrt(m)*jacobi_cd(u,m))
101   /((1-m)*sqrt(m)));
103 /* Use functions.wolfram.com 09.32.21.0001.01, not A&S 16.24.6 */
104 integrate(jacobi_nd(u,m),u); 
105 sqrt(1-jacobi_cd(u,m)^2)*acos(jacobi_cd(u,m))/((1-m)*jacobi_sd(u,m));
107 integrate(jacobi_dc(u,m),u); /* A&S 16.24.7 */
108 log(jacobi_sc(u,m)+jacobi_nc(u,m));
110 integrate(jacobi_nc(u,m),u); /* A&S 16.24.8 */
111 log(sqrt(1-m)*jacobi_sc(u,m)+jacobi_dc(u,m))/sqrt(1-m);
113 integrate(jacobi_sc(u,m),u); /* A&S 16.24.9 */
114 log(sqrt(1-m)*jacobi_nc(u,m)+jacobi_dc(u,m))/sqrt(1-m);
116 integrate(jacobi_ns(u,m),u); /* A&S 16.24.10 */
117 log(jacobi_ds(u,m)-jacobi_cs(u,m));
119 /* Use functions.wolfram.com 09.30.21.0001.01, not A&S 16.24.11 */
120 integrate(jacobi_ds(u,m),u);
121 log((1-jacobi_cn(u,m))/jacobi_sn(u,m));
123 integrate(jacobi_cs(u,m),u); /* A&S 16.24.12 */
124 log(jacobi_ns(u,m)-jacobi_ds(u,m));
126 /* Check the integrals and derivatives by confirming
128        f(x,m)-diff(integral(x,m),x),x) = constant
130   Look at Taylor expansion about zero, rather than messing about with 
131   elliptic function.  This was sufficient to find several errors  */
132 (te(f,n):=ratsimp(
133            taylor(f(x,m)
134            -diff(integrate(f(x,m),x),x),x,0,n)),
135 ti(f,n):=ratsimp(
136            taylor(integrate(f(x,m),x),x,0,n)
137             -integrate(taylor(f(x,m),x,0,n-1),x)),
138 td(f,n):=ratsimp(
139            taylor(diff(f(x,m),x),x,0,n)
140             -diff(taylor(f(x,m),x,0,n+1),x)),
141 /* Compare analytic and numerical integral */
142 ni(f,x1,x2,m):= block(
143   [x,n,I,In,Ia,key:1,eps:1.0e-14],
144   I:integrate(f(x,n),x),
145   In:quad_qag(f(x,m),x,x1,x2,key),
146   Ia:subst([x=float(x2),n=float(m)],I)-subst([x=float(x1),n=float(m)],I),
147   if ( abs(Ia-In[1]) < eps ) then
148     true
149   else
150     [Ia,In[1]]
152 done);
153 done;
155 te(jacobi_sn,4);
158 te(jacobi_cn,4);
161 te(jacobi_dn,4);
164 te(jacobi_cd,4);
167 assume(m>0,m<1); /* also ok for m<0 and m>0 */
168 [m > 0, m < 1];
169 te(jacobi_sd,4);
171 forget(m>0,m<1);
172 [m > 0, m < 1];
174 te(jacobi_nd,4);
177 te(jacobi_dc,4);
180 te(jacobi_nc,4);
183 te(jacobi_sc,4);
186 /* jacobi_ns, jacobi_ds and jacobi_cs are singular at x=0 
187    Compare numerical and analytic integrals for a single case.
190 ni(jacobi_ns,1,2,1/2);
191 true;
193 ni(jacobi_ds,1,2,1/2);
194 true;
196 ni(jacobi_cs,1,2,1/2);
197 true;
199 kill(te,ti,td,ni);
200 done;
202 /* Slightly modified version of test_table taken from rtest_expintegral.mac */
204 (test_table(func,table,eps) :=
205 block([badpoints : [],
206        abserr    : 0,
207        maxerr    : -1],
208   for entry in table do
209     block([z : entry[1],
210            result, answer],
211       z : expand(rectform(bfloat(entry[1]))),
212       result : rectform(apply(func, z)),
213       answer : expand(rectform(bfloat(entry[2]))),
214       abserr : abs(result-answer),
215       maxerr : max(maxerr,abserr),
216       if abserr > eps then
217         badpoints : cons ([z,result,answer,abserr],badpoints)
218     ),
219   if badpoints # [] then
220     cons(maxerr,reverse(badpoints))
221   else
222     badpoints
223 ),done);
224 done;
226 /* These test values come from http://getnet.net/~cherry/testrf.mac */
229  * rf(1,2,0) = (gamma(1/4)^2/4/sqrt(2*%pi)
230  * rf(%i,-%i,0) = 1/2*integrate(1/sqrt(t^2+1)/sqrt(t),t,0,inf) = beta(1/4,1/4)/4;
231  * rf(1/2,1,0) = 1/sqrt(1/2)*rf(1,2,0) (See https://dlmf.nist.gov/19.20.E1)
232  */
233 (mrf:[[[1,2,0], gamma(1/4)^2/4/sqrt(2*%pi)],
234      [[0.5,1,0], gamma(1/4)^2/(4*sqrt(%pi))]],
235 done);
236 done;
238 (mrf2:[[[%i,-%i,0], beta(1/4,1/4)/4],
239      [[%i-1,%i,0],0.79612586584234b0-%i*(1.2138566698365b0)],
240      [[%i,-%i,2],1.0441445654064b0],
241      [[2,3,4],0.58408284167715b0],
242      [[%i-1,%i,1-%i],0.93912050218619b0-%i*(0.53296252018635b0)]],
243 done);
244 done;
246 test_table('carlson_rf, mrf, 8.3267b-17);
249 test_table('carlson_rf, mrf2, 1.2102b-8);
253  * rc(0,1/4) = 1/2*integrate(1/sqrt(t)/(t+1/4), t, 0, inf)
254  *           = %pi
256  * rc(9/4,2) = 1/2*integrate(1/sqrt(t+9/4)/(t+2), t, 0, inf)
257  *           = log(2)
258  * After doing a logcontract.
260  * rc(0,%i) = 1/2*integrate(1/sqrt(t)/(t+%i), t, 0, inf);
261  *          = (1-%i)*%pi/2^(3/2)
263  * rc(2,1) = 1/2*integrate(1/sqrt(t+2)/(t+1), t, 0, inf)
264  *         = (log(sqrt(2)+1)-log(sqrt(2)-1))/2
265  *         = -log(sqrt(2)-1)
267  * After doing a logcontract, ratsimp/algebraic, and logcontract with
268  *  logconcoeffp, and a sqrtdenest
270  * rc(0,1) = 1/2*integrate(1/sqrt(t)/(t+1), t, 0, inf)
271  *         = %pi/2
272  * rc(%i, %i+1) = 1/2*integrate(1/sqrt(t+%i)/(t+%i+1), t, 0, inf)
273  *     = (%pi-2*atan((-1)^(1/4)))/2
274  *     = %pi/4+%i/2*log(sqrt(2)-1)
276  * After applying rectform, ratsimp, logcontract, then another
277  * logcontract with logconcoeffp set to featurep(m) or ratnump(m).
279  */
280 (mrc:[[[0b0,1/4],bfloat(%pi)],
281       [[9/4,2b0],log(2b0)],
282       [[0b0,%i],(1-%i)*%pi/2^(3/2)],
283       [[-%i,%i],1.2260849569072b0-%i*(0.34471136988768b0)],
284       [[1/4,-2],log(2b0)/3],
285       [[%i,-1],0.77778596920447b0+%i*(0.19832484993429b0)],
286       [[0,1/4],%pi],
287       [[9/4,2],log(2)],
288       [[2,1],-log(sqrt(2)-1)],
289       [[-%i,%i],-log(sqrt(2)-1)/2+ %pi/4-%i*(log(sqrt(2)-1)/2+%pi/4)],
290       [[1/4,-2],log(2)/3],
291       [[%i,-1],sqrt(sqrt(2)/4-1/4)*atan(sqrt(sqrt(2)-1))-
292                sqrt(sqrt(2)/16+1/16)*log(-sqrt(2*sqrt(2)+2)+sqrt(2)+1)+
293                %i*(sqrt(sqrt(2)/4+1/4)*atan(sqrt(sqrt(2)-1))+sqrt(sqrt(2)/16-
294                1/16)*log(-sqrt(2*sqrt(2)+2)+sqrt(2)+1))],
295       [[0,1],%pi/2],
296       [[%i,%i+1],%pi/4+%i*log(sqrt(2)-1)/2]],
297 done);
298 done;
300 test_table('carlson_rc, mrc, 2.5b-14);
303 (mrj:[[[0,1,2,3],0.77688623778582b0],
304       [[2,3,4,5],0.14297579667157b0],
305       [[2,3,4,-1+%i],0.13613945827771b0-%i*(0.38207561624427b0)],
306       [[-1+%i,-1-%i,1,2],0.9414835884122b0],
307       [[-1+%i,-1-%i,1,-3+%i],-0.61127970812028b0-%i*(1.0684038390007b0)],
308       [[-1+%i,-2-%i,-%i,-1+%i],1.8249027393704b0-%i*(1.2218475784827b0)],
309       [[2,3,4,-0.5],0.24723819703052b0],
310       [[2,3,4,-5],-0.12711230042964b0]],
311 done);
312 done;
314 (mrj2:[[[%i,-%i,0,2],1.6490011662711b0],
315       [[%i,-%i,0,1-%i],1.8260115229009b0+%i*(1.2290661908643b0)]],
316 done);
317 done;
319 test_table('carlson_rj, mrj, 1b-13);
322 test_table('carlson_rj, mrj2, 2.3452b-8);
326  * rd(0,2,1) = (3*gamma(3/4)^2)/(sqrt(2)*sqrt(%pi))
328 (mrd:[[[0,2,1],(3*gamma(3/4)^2)/(sqrt(2)*sqrt(%pi))],
329       [[2,3,4],0.16510527294261b0],
330       [[-2-%i,-%i,-1+%i],1.8249027393704b0-%i*(1.2218475784827b0)]],
331 done);
332 done;
334 (mrd2:[[[%i,-%i,2],0.6593385415422b0],
335       [[0,%i,-%i],1.270819627191b0+%i*(2.7811120159521b0)],
336       [[0,%i-1,%i],-1.8577235439239b0-%i*(0.96193450888839b0)]],
337 done);
338 done;
340 test_table('carlson_rd, mrd, 6e-14);
343 test_table('carlson_rd, mrd2, 4.7818b-8);
346 /* Some tests of the Jacobian elliptic functions.  
347  * Just some tests at random points, to verify that we are accurate.
348  * The reference values were obtained from Mathematica, but we could 
349  * also just compute the values using a much larger fpprec.
350  */
352 test_table('jacobi_sn,
353            [[[1b0+%i*1b0, .7b0], 1.134045971912365274954b0 + 0.352252346922494477621b0*%i],
354             [[1b0+%i*1b0, 2b0], 0.98613318109123804740b0 + 0.09521910819727230780b0*%i],
355             [[1b0+%i*1b0, 2b0+3b0*%i], 0.94467077879445294981b0 - 0.19586410083100945528b0* %i],
356             [[1.785063352082689082581887b0 *%i + 8.890858759528509578661528b-1,
357               9.434463451695984398149033b-1 - 1.476052973708684178844821b-1 * %i],
358              1.345233534539675700312281b0 - 7.599023926615176284214056b-2 * %i]],
359            1b-15);
362 test_table('jacobi_cn,
363            [[[1b0+%i*1b0, .7b0], 0.571496591371764254029b0 - 0.698989917271916772991b0*%i ],
364             [[1b0+%i*1b0, 2b0], 0.33759463268642431412b0 - 0.27814044708010806377b0*%i],
365             [[1b0+%i*1b0, 2b0+3b0*%i], -0.52142079485827170824b0 - 0.35485177134179601850b0*%i],
366             [[100b0, .7b0], 0.93004753815774770476196b0]], 
367            6b-14);
370 test_table('jacobi_dn,
371            [[[1b0+%i*1b0, .7b0], 0.62297154372331777630564880787568b0 - 0.448863598031509643267241389621738b0 *%i ],
372             [[1b0+%i*1b0, 2b0], 0.1913322443206041462495606602242b0 - 0.9815253294150083432282549919753b0 * %i],
373             [[1b0+%i*1b0, 2b0+3b0*%i], 0.6147387452173944656984656771134b0 - 1.4819401302071697495918834416787b0 * %i],
374             [[100b0, .7b0], 0.95157337933724324055428565654872978b0],
375             [[1.785063352082689082581887b0 *%i + 8.890858759528509578661528b-1,
376               9.434463451695984398149033b-1 - 1.476052973708684178844821b-1 * %i],
377              -8.617730683333292717095686b-1 *%i - 2.663978258141280808361839b-1]], 
378            2b-15);
381 /* These routines for cn and dn work well for small (<= 1?) values of
382  * u and m.  They have known issues for large real values of u.
383  */
384 (ascending_transform(u,m) :=
385   block([root_m : expand(rectform(sqrt(m))), mu, root_mu1, v],
386     mu : expand(rectform(4*root_m/(1+root_m)^2)),
387     root_mu1 : expand(rectform((1-root_m)/(1+root_m))),
388     v : expand(rectform(u/(1+root_mu1))),
389     [v, mu, root_mu1]),
390  elliptic_dn_ascending(u,m) :=
391   if is(abs(m-1) < 4*10^(-fpprec)) then sech(u)
392   else
393     block([v, mu, root_mu1, new],
394       [v, mu, root_mu1] : ascending_transform(u,m),
395       new : elliptic_dn_ascending(v, mu),
396       expand(rectform((1-root_mu1)/mu*(new^2 + root_mu1)/new))),
397  elliptic_cn_ascending(u,m) :=
398   if is(abs(m-1) < 4*10^(-fpprec)) then sech(u)
399   else
400     block([v, mu, root_mu1, new],
401       [v, mu, root_mu1] : ascending_transform(u,m),
402       new : elliptic_dn_ascending(v, mu),
403       expand(rectform((1+root_mu1)/mu*(new^2-root_mu1)/new))),
404  done);
405 done;
407 /* Test with random values for the argument and parameter. */
408 (test_random(n, eps, testf, truef) :=
409   block([badpoints : [], maxerr : -1],
410     for k : 1 thru n do
411       block([z : bfloat(1-2*random(1d0) + %i * (1-2*random(1d0))),
412              m : bfloat(1-2*random(1d0) + %i * (1-2*random(1d0))),
413              ans, expected, abserr],
414         ans : testf(z, m),
415         expected : truef(z, m),
416         abserr : abs(ans-expected),
417         maxerr : max(maxerr, abserr),
418         if abserr > eps then
419           badpoints : cons([[z,m], ans, expected, abserr], badpoints)),
420     if badpoints # [] then
421       cons(maxerr, badpoints)
422     else
423       badpoints),
424  done);
425 done;
427 test_random(50, 2b-15, 'jacobi_dn, 'elliptic_dn_ascending);
430 test_random(50, 2b-15, 'jacobi_cn, 'elliptic_cn_ascending);
433 /* Test elliptic_f by using the fact that
435 inverse_jacobi_sn(x,m) = elliptic_f(asin(x), m)
438 (test_ef(x,m) := jacobi_sn(elliptic_f(asin(x),m), m), id(x,m):=x, done);
439 done;
441 test_random(100, 6b-13, 'test_ef, 'id);
444 /* Test elliptic_kc These values are from 
446  * http://functions.wolfram.com/EllipticIntegrals/EllipticK/03/01/
447  */
449 block([oldfpprec : fpprec, fpprec:100],
450   test_table('elliptic_kc,
451              [[[1/2], 8*%pi^(3/2)/gamma(-1/4)^2],
452               [[17-12*sqrt(2)], 2*(2+sqrt(2))*%pi^(3/2)/gamma(-1/4)^2],
453               [[-1], gamma(1/4)^2/4/sqrt(2*%pi)]],
454              2b-100));
457 /* Some tests for specific values */
458 inverse_jacobi_sn(1,m);
459 elliptic_kc(m);
461 inverse_jacobi_sn(x,0);
462 asin(x);
464 inverse_jacobi_sn(x,1);
465 log(tan(asin(x)/2 + %pi/4));
467 inverse_jacobi_sd(1/sqrt(1-m),m);
468 elliptic_kc(m);
470 inverse_jacobi_ds(sqrt(1-m),m);
471 elliptic_kc(m);
473 /* elliptic_kc(1) is undefined */
474 errcatch(elliptic_kc(1));
476 errcatch(elliptic_kc(1.0));
478 errcatch(elliptic_kc(1.0b0));
481 /* Test noun/verb results from elliptic functions */
482 diff(inverse_jacobi_sn(x,0),x);
483 1/sqrt(1-x^2);
485 diff(elliptic_pi(4/3,phi,0),phi);
486 sec(phi)^2/(1-tan(phi)^2/3);
488 diff(elliptic_pi(3/4,phi,0),phi);
489 sec(phi)^2/(1+tan(phi)^2/4);
491 diff(elliptic_pi(1,phi,0),phi);
492 sec(phi)^2;
494 /* signbfloat:false provokes "Is 0 positive, negative, or zero?" */
495 (signbfloat:false, 
496  diff(elliptic_pi(1,phi,0),phi));
497 sec(phi)^2;
499 (reset (signbfloat), 0);
502 /* Special values */
503 elliptic_ec(1);
506 elliptic_ec(0);
507 %pi/2;
509 elliptic_ec(1/2);
510 gamma(3/4)^2/(2*sqrt(%pi))+%pi^(3/2)/(4*gamma(3/4)^2);
512 elliptic_ec(-1);
513 sqrt(2)*elliptic_ec(1/2);
515 /* Test periodicity of elliptic_e */
516 elliptic_e(x, 1);
517 2*round(x/%pi)+sin(x-%pi*round(x/%pi));
519 elliptic_e(3,1/3);
520 elliptic_e(3-%pi,1/3)+2*elliptic_ec(1/3);
522 /* Bug #2629: elliptic_kc(3.0) not accurate */
523 test_table('elliptic_kc, [[[3.0], elliptic_kc(3b0)]], 1b-15);
526 /* Bug #2630: inverse_jacobi_cn(-2.0, 3.0) generates an error */
527 test_table('inverse_jacobi_cn, [[[-2.0, 3.0], 2.002154760912212-3.202503914656527*%i]],
528   1b-15);
531 /* Bug #2615: Numeric evaluation of inverse Jacobi elliptic functions is wrong for some inputs 
533 is(abs(jacobi_dn(inverse_jacobi_dn(-2.0,3.0), 3.0) + 2) < 1d-14);
534 true;
536 /* elliptical functions handling of non-rectangular complex numbers
537  * mailing list 2016-07-14 "Jacobi elliptic functions, maxima 5.38.1"
538  * mailing list 2020-05-28 "Unexpected result from ev"
539  */
541 /* list of functions here is everything that has a SIMP-[$%](ELLIPTIC|JACOBI)_FOO in sr/ellipt.lisp */
543 elliptic_fcns :
544   [ elliptic_e, elliptic_ec, elliptic_eu, elliptic_f, elliptic_kc, elliptic_pi,
545     inverse_jacobi_cd, inverse_jacobi_cn, inverse_jacobi_cs, inverse_jacobi_dc,
546     inverse_jacobi_dn, inverse_jacobi_ds, inverse_jacobi_nc, inverse_jacobi_nd,
547     inverse_jacobi_ns, inverse_jacobi_sc, inverse_jacobi_sd, inverse_jacobi_sn,
548     jacobi_am,
549     jacobi_cd, jacobi_cn, jacobi_cs, jacobi_dc, jacobi_dn, jacobi_ds,
550     jacobi_nc, jacobi_nd, jacobi_ns, jacobi_sc, jacobi_sd, jacobi_sn ];
551  */
553 /* code to generate test cases below
555 elliptic_fcns_1 : sublist (elliptic_fcns, lambda ([f], errcatch(f('a)) # []));
556 elliptic_fcns_2 : sublist (elliptic_fcns, lambda ([f], errcatch(f('a, 'b)) # []));
557 elliptic_fcns_3 : sublist (elliptic_fcns, lambda ([f], errcatch(f('a, 'b, 'c)) # []));
559 kill (u1, k1, n1);
561 with_stdout ("/tmp/foo.mac",
563   for f in elliptic_fcns_1
564     do block ([a : f('rectform(u1))],
565         print ('complex_floatp(a), ";"),
566         print (true, ";"),
567         print ("")),
568   
569   for f in elliptic_fcns_2
570     do block ([a : f('rectform(u1), k1)],
571         print ('complex_floatp(a), ";"),
572         print (true, ";"),
573         print ("")),
574   
575   for f in elliptic_fcns_3
576     do block ([a : f(n1, 'rectform(u1), k1)],
577         print ('complex_floatp(a), ";"),
578         print (true, ";"),
579         print ("")));
580  */
582 (u1 : 0.5*(1.0 - 0.25*%i)^2,
583  k1 : 0.775,
584  n1 : -1.375,
585  complex_floatp(e) := floatnump(realpart(e)) and floatnump(imagpart(e)),
586  0);
589 complex_floatp(elliptic_ec(rectform(u1))) ; 
590 true ; 
592 complex_floatp(elliptic_kc(rectform(u1))) ; 
593 true ; 
595 complex_floatp(elliptic_e(rectform(u1), k1)) ; 
596 true ; 
598 complex_floatp(elliptic_eu(rectform(u1), k1)) ; 
599 true ; 
601 complex_floatp(elliptic_f(rectform(u1), k1)) ; 
602 true ; 
604 complex_floatp(inverse_jacobi_cd(rectform(u1), k1)) ; 
605 true ; 
607 complex_floatp(inverse_jacobi_cn(rectform(u1), k1)) ; 
608 true ; 
610 complex_floatp(inverse_jacobi_cs(rectform(u1), k1)) ; 
611 true ; 
613 complex_floatp(inverse_jacobi_dc(rectform(u1), k1)) ; 
614 true ; 
616 complex_floatp(inverse_jacobi_dn(rectform(u1), k1)) ; 
617 true ; 
619 complex_floatp(inverse_jacobi_ds(rectform(u1), k1)) ; 
620 true ; 
622 complex_floatp(inverse_jacobi_nc(rectform(u1), k1)) ; 
623 true ; 
625 complex_floatp(inverse_jacobi_nd(rectform(u1), k1)) ; 
626 true ; 
628 complex_floatp(inverse_jacobi_ns(rectform(u1), k1)) ; 
629 true ; 
631 complex_floatp(inverse_jacobi_sc(rectform(u1), k1)) ; 
632 true ; 
634 complex_floatp(inverse_jacobi_sd(rectform(u1), k1)) ; 
635 true ; 
637 complex_floatp(inverse_jacobi_sn(rectform(u1), k1)) ; 
638 true ; 
640 complex_floatp(jacobi_am(rectform(u1), k1)) ; 
641 true ; 
643 complex_floatp(jacobi_cd(rectform(u1), k1)) ; 
644 true ; 
646 complex_floatp(jacobi_cn(rectform(u1), k1)) ; 
647 true ; 
649 complex_floatp(jacobi_cs(rectform(u1), k1)) ; 
650 true ; 
652 complex_floatp(jacobi_dc(rectform(u1), k1)) ; 
653 true ; 
655 complex_floatp(jacobi_dn(rectform(u1), k1)) ; 
656 true ; 
658 complex_floatp(jacobi_ds(rectform(u1), k1)) ; 
659 true ; 
661 complex_floatp(jacobi_nc(rectform(u1), k1)) ; 
662 true ; 
664 complex_floatp(jacobi_nd(rectform(u1), k1)) ; 
665 true ; 
667 complex_floatp(jacobi_ns(rectform(u1), k1)) ; 
668 true ; 
670 complex_floatp(jacobi_sc(rectform(u1), k1)) ; 
671 true ; 
673 complex_floatp(jacobi_sd(rectform(u1), k1)) ; 
674 true ; 
676 complex_floatp(jacobi_sn(rectform(u1), k1)) ; 
677 true ; 
679 complex_floatp(elliptic_pi(n1, rectform(u1), k1)) ; 
680 true ; 
683 with_stdout ("/tmp/bar.mac",
685   for f in elliptic_fcns_1
686     do block ([a : f('rectform(u1))],
687         print ('complex_bfloatp(a), ";"),
688         print (true, ";"),
689         print ("")),
691   for f in elliptic_fcns_2
692     do block ([a : f('rectform(u1), k1)],
693         print ('complex_bfloatp(a), ";"),
694         print (true, ";"),
695         print ("")),
697   for f in elliptic_fcns_3
698     do block ([a : f(n1, 'rectform(u1), k1)],
699         print ('complex_bfloatp(a), ";"),
700         print (true, ";"),
701         print ("")));
702  */
704 (u1 : 0.5b0*(1.0b0 - 0.25b0*%i)^2,
705  k1 : 0.775b0,
706  n1 : -1.375b0,
707  complex_bfloatp(e) := bfloatp(realpart(e)) and bfloatp(imagpart(e)),
708  0);
711 complex_bfloatp(elliptic_ec(rectform(u1))) ; 
712 true ; 
714 complex_bfloatp(elliptic_kc(rectform(u1))) ; 
715 true ; 
717 complex_bfloatp(elliptic_e(rectform(u1), k1)) ; 
718 true ; 
720 complex_bfloatp(elliptic_eu(rectform(u1), k1)) ; 
721 true ; 
723 complex_bfloatp(elliptic_f(rectform(u1), k1)) ; 
724 true ; 
726 complex_bfloatp(inverse_jacobi_cd(rectform(u1), k1)) ; 
727 true ; 
729 complex_bfloatp(inverse_jacobi_cn(rectform(u1), k1)) ; 
730 true ; 
732 complex_bfloatp(inverse_jacobi_cs(rectform(u1), k1)) ; 
733 true ; 
735 complex_bfloatp(inverse_jacobi_dc(rectform(u1), k1)) ; 
736 true ; 
738 complex_bfloatp(inverse_jacobi_dn(rectform(u1), k1)) ; 
739 true ; 
741 complex_bfloatp(inverse_jacobi_ds(rectform(u1), k1)) ; 
742 true ; 
744 complex_bfloatp(inverse_jacobi_nc(rectform(u1), k1)) ; 
745 true ; 
747 complex_bfloatp(inverse_jacobi_nd(rectform(u1), k1)) ; 
748 true ; 
750 complex_bfloatp(inverse_jacobi_ns(rectform(u1), k1)) ; 
751 true ; 
753 complex_bfloatp(inverse_jacobi_sc(rectform(u1), k1)) ; 
754 true ; 
756 complex_bfloatp(inverse_jacobi_sd(rectform(u1), k1)) ; 
757 true ; 
759 complex_bfloatp(inverse_jacobi_sn(rectform(u1), k1)) ; 
760 true ; 
762 complex_bfloatp(jacobi_am(rectform(u1), k1)) ; 
763 true ; 
765 complex_bfloatp(jacobi_cd(rectform(u1), k1)) ; 
766 true ; 
768 complex_bfloatp(jacobi_cn(rectform(u1), k1)) ; 
769 true ; 
771 complex_bfloatp(jacobi_cs(rectform(u1), k1)) ; 
772 true ; 
774 complex_bfloatp(jacobi_dc(rectform(u1), k1)) ; 
775 true ; 
777 complex_bfloatp(jacobi_dn(rectform(u1), k1)) ; 
778 true ; 
780 complex_bfloatp(jacobi_ds(rectform(u1), k1)) ; 
781 true ; 
783 complex_bfloatp(jacobi_nc(rectform(u1), k1)) ; 
784 true ; 
786 complex_bfloatp(jacobi_nd(rectform(u1), k1)) ; 
787 true ; 
789 complex_bfloatp(jacobi_ns(rectform(u1), k1)) ; 
790 true ; 
792 complex_bfloatp(jacobi_sc(rectform(u1), k1)) ; 
793 true ; 
795 complex_bfloatp(jacobi_sd(rectform(u1), k1)) ; 
796 true ; 
798 complex_bfloatp(jacobi_sn(rectform(u1), k1)) ; 
799 true ; 
801 complex_bfloatp(elliptic_pi(n1, rectform(u1), k1)) ; 
802 true ; 
804 /* Test derivatives at a few random points using numerical differentiation
805    
806    func  - a function
807    df - deriviative of func wrt p-th argument
808    p - derivative wrt p-th argument 
809    vars - variables in expression deriv
810    table - table of points to evaluate
811    eps - report errors > eps
812    delta - offset for numerical derivative
814 (test_deriv(func,df,vars,p,table,eps,delta) :=
815 block([badpoints : [],
816        abserr    : 0,
817        maxerr    : -1],
818   for %z in table do
819     block([%z0, %z1, deriv, nderiv],
820       %z1:makelist(if i=p then %z[i]+delta else %z[i],i,1,length(%z)),
821       %z0:makelist(if i=p then %z[i]-delta else %z[i],i,1,length(%z)),
822       nderiv: float((apply(func,%z1)-apply(func,%z0))/(2*delta)),
823       deriv : float(subst(maplist("=",vars,%z),df)),
824       abserr : abs(nderiv-deriv),
825       maxerr : max(maxerr,abserr),
826       if abserr > eps then
827         badpoints : cons ([%z,deriv,nderiv,abserr],badpoints)
828     ),
829   if badpoints # [] then
830     cons(maxerr,badpoints)
831   else
832     badpoints
833     ),done);
834 done;
836 /* test points for two-arg functions */
837 (l2: [[0.2, 0.3], [0.2, 0.5], [1.5, 0.7], [0.5, 0.99], [1.5, 0.8],
838     [1.57, 0.8], [1.5707, 0.8], [-1.0, 0.8], [2.0, 0.8]],done);
839 done;
841 /* test points for three-arg functions */
842 /* bug #3221 elliptic_pi() wrong for 2nd arg > float(%pi/2) */
843 (l3: [[0.1,0.2,0.3],[0.1,0.2,0.5],[0.2,1.5,0.7],[0.001,0.5,0.99],
844     [0.2,1.5,0.8],[0.2,1.57,0.8],[0.2,1.5707,0.8]],done);
845 done;
847 test_deriv('elliptic_f,diff(elliptic_f(z,m),z),[z,m],1,l2,2.0e-7,1.0e-8);
850 test_deriv('elliptic_f,diff(elliptic_f(z,m),m),[z,m],2,l2,2.0e-7,1.0e-8);
853 test_deriv('elliptic_e,diff(elliptic_e(z,m),z),[z,m],1,l2,2.0e-7,1.0e-8);
856 test_deriv('elliptic_e,diff(elliptic_e(z,m),m),[z,m],2,l2,2.0e-7,1.0e-8);
859 test_deriv('elliptic_pi,diff(elliptic_pi(n,z,m),n),[n,z,m],1,l3,2.0e-7,1.0e-8);
862 test_deriv('elliptic_pi,diff(elliptic_pi(n,z,m),z),[n,z,m],2,l3,2.0e-7,1.0e-8);
865 test_deriv('elliptic_pi,diff(elliptic_pi(n,z,m),m),[n,z,m],3,l3,2.0e-7,1.0e-8);
868 closeto(e,tol):=block([numer:true,abse],abse:abs(e),if(abse<tol) then true else abse);
869 closeto(e,tol):=block([numer:true,abse],abse:abs(e),if(abse<tol) then true else abse);
872  * Some additional tests for complex values. Use the fact that
873  * elliptic_pi(0, phi, m) = elliptic_f(phi,m);
874  */
875 closeto(elliptic_pi(0, 0.5, 0.25)- elliptic_f(0.5, 0.25), 1e-15);
876 true;
877 closeto(elliptic_pi(0, 0.5*%i, 0.25)- elliptic_f(0.5*%i, 0.25), 1e-15);
878 true;
880 closeto(elliptic_pi(0, 0.5b0, 0.25b0)- elliptic_f(0.5b0, 0.25b0), 1e-15);
881 true;
882 closeto(elliptic_pi(0, 0.5b0*%i, 0.25b0)- elliptic_f(0.5b0*%i, 0.25b0), 1e-15);
883 true;
885 /* Test for Bug 3221 */
887 closeto(elliptic_pi(0.2,1.57,0.8) - 2.5770799919605668, 1e-14);
888 true;
889 closeto(elliptic_pi(0.2,1.58,0.8) - 2.60502920656026151, 1e-14);
890 true;
891 closeto(elliptic_pi(0.20,2.00,0.80) - 3.6543835090829025, 7.47e-11);
892 true;
894 (oldfpprec:fpprec,fpprec:32);
896 closeto(elliptic_pi(0.2b0,1.57b0,0.8b0) - 2.5770799919605668058849721013196b0, 3.09b-32);
897 true;
898 closeto(elliptic_pi(0.2b0,1.58b0,0.8b0) - 2.6050292065602615145481924917132b0, 1.24b-32);
899 true;
900 closeto(elliptic_pi(0.2b0,2.0b0,0.8b0) - 3.6543835090082902501670624830938b0, 1b-32);
901 true;
903 /* Test for #3733 $gamma vs %gamma confusion */
904 elliptic_ec(1/2) - gamma(3/4)^2/(2*sqrt(%pi))- %pi^(3/2)/(4*gamma(3/4)^2);
907 /* #3745 Quoting either elliptic_f [or elliptic_e] inhibits simplification */
908 'elliptic_f(5,0);
911 'elliptic_e(5,0);
914 /* #3967 elliptic_e(5*%pi/4,1) inconsistent with numerical evaluation */
915 elliptic_e(5*%pi/4, 1);
916 1/sqrt(2) + 2;
918 closeto(elliptic_e(5*%pi/4,1) - elliptic_e(float(5*%pi/4),1), 1e-16);
919 true;
921 /* #3746 derivative of inverse_jacobi_sn is noun/verb confused. 
922    The same is true for inverse_jacobi_cn and inverse_jacobi_dn, but the 
923    remaining nine inverse Jacobi functions ns, nc, nd, sc, cs, sd, ds, cd, 
924    and dc don't have defined m derivatives.*/
925 subst(u=0, diff(inverse_jacobi_sn(u,m),m));
928 subst(u=1, diff(inverse_jacobi_cn(u,m),m));
931 subst(u=1, diff(inverse_jacobi_dn(u,m),m));
934 /* Carlson's integrals */
936 (assume(xp > 0), done);
937 done$
939 carlson_rc(xp,xp);
940 1/sqrt(xp)$
942 carlson_rc(0,1);
943 %pi/2$
945 carlson_rc(0, 1/4);
946 %pi$
948 carlson_rc(2, 1);
949 log(sqrt(2)+1)$
951 carlson_rc(%i, %i+1);
952 %pi/4 + %i*log(sqrt(2)-1)/2$
954 carlson_rc(0, %i);
955 ((1-%i)*%pi)/2^(3/2)$
957 carlson_rc(((1+xp)/2)^2, xp);
958 log(xp)/(xp-1)$
960 /* log(x) = (x-1)*carlson_rc(((1+x)/2)^2,x) */
961 closeto((2-1)*carlson_rc(((1+2)/2)^2,2.0) - log(2.0), 1e-16);
962 true$
964 /* asin(x) = x*carlson_rc(1-x^2,1) */
965 closeto(1/2*carlson_rc(1-1/4,1.0) - asin(0.5), 2.2205e-16);
966 true$
968 /* acos(x) = sqrt(1-x^2)*carlson_rc(x^2,1) */
969 closeto(sqrt(1.0-1/4)*carlson_rc(1/4,1.0) - acos(0.5), 2.2205e-16);
970 true$
972 /* atan(x) = x*carlson_rc(1,1+x^2) */
973 closeto(1*carlson_rc(1.0,1+1^2) - atan(1.0), 1.1103e-16);
974 true$
976 /* asinh(x) = x*carlson_rc(1+x^2,1) */
977 closeto(1*carlson_rc(1+1^2,1.0) - asinh(1.0), 2.2205e-16);
978 true$
980 /* acosh(x) = sqrt(x^2-1)*carlson_rc(x^2,1) */
981 closeto(sqrt(2^2-1.0)*carlson_rc(2^2,1.0) - acosh(2.0), 2.2205e-16);
982 true$
984 /* atanh(x) = x*carlson_rc(1,1-x^2) */
985 closeto(1/2*carlson_rc(1.0,1-(1/2)^2) - atanh(0.5), 1.1103e-16);
986 true$
988 carlson_rd(x, x, x);
989 1/x^(3/2)$
991 carlson_rd(0,2,1);
992 3*sqrt(%pi)*gamma(3/4)/gamma(1/4)$
994 carlson_rd(2,0,1);
995 3*sqrt(%pi)*gamma(3/4)/gamma(1/4)$
997 closeto(carlson_rd(0, 2, 1.0) - float(3*sqrt(%pi)*gamma(3/4)/gamma(1/4)), 6.66134e-16);
998 true$
1000 carlson_rf(x,x,x);
1001 1/sqrt(x)$
1004  * Rf(1,2,0) has an analytic solution.  Check some of the permutations too.
1005  */
1006 carlson_rf(1,2,0);
1007 gamma(1/4)^2/(4*sqrt(2*%pi))$
1009 carlson_rf(2,1,0);
1010 gamma(1/4)^2/(4*sqrt(2*%pi))$
1012 carlson_rf(0, 1,2);
1013 gamma(1/4)^2/(4*sqrt(2*%pi))$
1015 /* Check permutations of Rf(%i, -%i, 0) too */
1016 carlson_rf(%i,-%i,0);
1017 gamma(1/4)^2/(4*sqrt(%pi))$
1019 carlson_rf(-%i,%i,0);
1020 gamma(1/4)^2/(4*sqrt(%pi))$
1022 carlson_rf(0, %i,-%i);
1023 gamma(1/4)^2/(4*sqrt(%pi))$
1026 carlson_rj(x,x,x,x);
1027 1/x^(3/2)$
1029 carlson_rj(x,y,z,z);
1030 carlson_rd(x,y,z)$
1032 (assume(pp>0,yp>0), done);
1033 done$
1035 carlson_rj(xp,xp,xp,pp);
1036 3*(1/sqrt(xp)-carlson_rc(xp,pp))/(pp-xp)$
1038 carlson_rj(0,yp,yp,pp);
1039 3*%pi/(2*(yp*sqrt(pp)+pp*sqrt(yp)))$
1041 carlson_rj(x,yp,yp,pp);
1042 3/(pp-yp)*(carlson_rc(x,yp) - carlson_rc(x,pp))$
1044 carlson_rj(x,yp,yp,yp);
1045 (3*(carlson_rc(x,yp)-sqrt(x)/yp))/(2*(yp-x))$
1047 (fpprec:oldfpprec,kill(l2,l3,test_deriv,oldfpprec),done);
1048 done;