From: SourceForge.net <noreply@so...>  20100504 12:38:17

Bugs item #2996542, was opened at 20100504 16:38 Message generated for change (Tracker Item Submitted) made by aleckalinin You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100504 13:08:25

Bugs item #2996542, was opened at 20100504 14:38 Message generated for change (Comment added) made by crategus You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Dieter Kaiser (crategus) Date: 20100504 15:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100517 20:01:56

Bugs item #2996542, was opened at 20100504 08:38 Message generated for change (Comment added) made by rtoy You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Raymond Toy (rtoy) Date: 20100517 16:01 Message: The new result comes from the new routine defintlogexp, which is called relatively early in defint. Perhaps it should be called later? I have not investigated this aspect yet.  Comment By: Dieter Kaiser (crategus) Date: 20100504 09:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100517 22:20:24

Bugs item #2996542, was opened at 20100504 14:38 Message generated for change (Comment added) made by crategus You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Dieter Kaiser (crategus) Date: 20100518 00:20 Message: We have the new algorithm of defintlogexp since Maxima 5.19. But the behavior for the log function has changed between 5.20 and 5.21. Therefore, I think it is not the algorithm of defintlogexp which has changed the integral of the log function, but some code which has been introduced later. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100517 22:01 Message: The new result comes from the new routine defintlogexp, which is called relatively early in defint. Perhaps it should be called later? I have not investigated this aspect yet.  Comment By: Dieter Kaiser (crategus) Date: 20100504 15:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100518 03:31:45

Bugs item #2996542, was opened at 20100504 08:38 Message generated for change (Comment added) made by rtoy You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Raymond Toy (rtoy) Date: 20100517 23:31 Message: You are correct. Looks like the issue comes from dintlog. In 5.19.2, the antideriv was tried first, then logx1. In 5.21, logx1 is tried first, then antideriv. I do not know which is better. The commit message says the change helps remove the limit from some integrals. But it makes this particular integral not as nice.  Comment By: Dieter Kaiser (crategus) Date: 20100517 18:20 Message: We have the new algorithm of defintlogexp since Maxima 5.19. But the behavior for the log function has changed between 5.20 and 5.21. Therefore, I think it is not the algorithm of defintlogexp which has changed the integral of the log function, but some code which has been introduced later. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100517 16:01 Message: The new result comes from the new routine defintlogexp, which is called relatively early in defint. Perhaps it should be called later? I have not investigated this aspect yet.  Comment By: Dieter Kaiser (crategus) Date: 20100504 09:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100519 23:35:35

Bugs item #2996542, was opened at 20100504 08:38 Message generated for change (Comment added) made by rtoy You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Raymond Toy (rtoy) Date: 20100519 19:35 Message: One possible solution. In dintlog, bind $gamma_expand to T when calling logx1. Then maxima returns a*log(a)a. However, this change causes 90 in rtestint to fail. The result is k1*gamma(k1) instead of gamma(1+k1).  Comment By: Raymond Toy (rtoy) Date: 20100517 23:31 Message: You are correct. Looks like the issue comes from dintlog. In 5.19.2, the antideriv was tried first, then logx1. In 5.21, logx1 is tried first, then antideriv. I do not know which is better. The commit message says the change helps remove the limit from some integrals. But it makes this particular integral not as nice.  Comment By: Dieter Kaiser (crategus) Date: 20100517 18:20 Message: We have the new algorithm of defintlogexp since Maxima 5.19. But the behavior for the log function has changed between 5.20 and 5.21. Therefore, I think it is not the algorithm of defintlogexp which has changed the integral of the log function, but some code which has been introduced later. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100517 16:01 Message: The new result comes from the new routine defintlogexp, which is called relatively early in defint. Perhaps it should be called later? I have not investigated this aspect yet.  Comment By: Dieter Kaiser (crategus) Date: 20100504 09:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100523 20:15:58

Bugs item #2996542, was opened at 20100504 14:38 Message generated for change (Comment added) made by crategus You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None Status: Open Resolution: None Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Dieter Kaiser (crategus) Date: 20100523 22:15 Message: Some more examples of integrals which have changed. In general the following type of integral: integrate(log(x)^n, x, 0, a); where n is a positive integer has changed. Some examples: (%i1) assume(a>0)$ (%i2) integrate(log(x)^2,x,0,a); (%o2) gamma_incomplete(3,log(a)) (%i3) integrate(log(x)^2,x,0,a),gamma_expand:true; (%o3) a*log(a)^22*a*log(a)+2*a (%i4) integrate(log(x)^3,x,0,a); (%o4) gamma_incomplete(4,log(a)) (%i5) integrate(log(x)^3,x,0,a),gamma_expand:true; (%o5) a*log(a)^33*a*log(a)^2+6*a*log(a)6*a (%i6) integrate(log(x)^4,x,0,a); (%o6) gamma_incomplete(5,log(a)) (%i7) integrate(log(x)^4,x,0,a),gamma_expand:true; (%o7) a*log(a)^44*a*log(a)^3+12*a*log(a)^224*a*log(a)+24*a But we have also new integrals which do not work in Maxima 5.20. These are of the type integrate(log(x)^(n/2), x, 0, a); where n is an integer. Some examples are: (%i8) integrate(log(x)^(1/2),x,0,a); (%o8) %i*gamma_incomplete(3/2,log(a)) (%i9) integrate(log(x)^(1/2),x,0,a),gamma_expand:true; (%o9) (sqrt(%pi)*%i*erfc(sqrt(log(a)))+2*%i*a*sqrt(log(a)))/2 (%i10) integrate(log(x)^(3/2),x,0,a); (%o10) %i*(3*sqrt(%pi)*erf(sqrt(log(a)))/4 a*sqrt(log(a))*log(a)+3*a*sqrt(log(a))/2+3*sqrt(%pi)/4) (%i11) integrate(log(x)^(5/2),x,0,a); (%o11) %i*gamma_incomplete(7/2,log(a)) (%i12) integrate(log(x)^(5/2),x,0,a),gamma_expand:true; (%o12) (15*sqrt(%pi)*%i*erfc(sqrt(log(a))) +sqrt(log(a))*(8*%i*a*log(a)^220*%i*a*log(a)+30*%i*a)) /8 So the best seems to be as suggested to bind the flag $gamma_expand to T. We get the change for gamma(k1+1) because the Gamma function uses the flag $gamma_expand too. Perhaps, we should introduce a second flag $gamma_incomplete_expand to have the possibility to get a more specific expansion of the Incomplete Gamma function. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100520 01:35 Message: One possible solution. In dintlog, bind $gamma_expand to T when calling logx1. Then maxima returns a*log(a)a. However, this change causes 90 in rtestint to fail. The result is k1*gamma(k1) instead of gamma(1+k1).  Comment By: Raymond Toy (rtoy) Date: 20100518 05:31 Message: You are correct. Looks like the issue comes from dintlog. In 5.19.2, the antideriv was tried first, then logx1. In 5.21, logx1 is tried first, then antideriv. I do not know which is better. The commit message says the change helps remove the limit from some integrals. But it makes this particular integral not as nice.  Comment By: Dieter Kaiser (crategus) Date: 20100518 00:20 Message: We have the new algorithm of defintlogexp since Maxima 5.19. But the behavior for the log function has changed between 5.20 and 5.21. Therefore, I think it is not the algorithm of defintlogexp which has changed the integral of the log function, but some code which has been introduced later. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100517 22:01 Message: The new result comes from the new routine defintlogexp, which is called relatively early in defint. Perhaps it should be called later? I have not investigated this aspect yet.  Comment By: Dieter Kaiser (crategus) Date: 20100504 15:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 
From: SourceForge.net <noreply@so...>  20100710 14:31:42

Bugs item #2996542, was opened at 20100504 14:38 Message generated for change (Comment added) made by crategus You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Lisp Core  Integration Group: None >Status: Closed >Resolution: Fixed Priority: 5 Private: No Submitted By: akalinin (aleckalinin) Assigned to: Nobody/Anonymous (nobody) Summary: log(x) integration is incorrect Initial Comment: Correct behaviour in previous verison: wxMaxima 0.8.4, Maxima 5.20.1: integrate(log(x), x, 0, a) > a log(a) a Incorrect behaviour in current version: wxMaxima 0.8.5, Maxima 5.21.0: integrate(log(x), x, 0, a) > gamma_incomplete(2,log(a))  >Comment By: Dieter Kaiser (crategus) Date: 20100710 16:31 Message: Fixed in defint.lisp revision 1.78. As suggested the option variable $gamma_expand has been bound to TRUE when calling logx1 in the routine dintlog. Closing this bug report as fixed. Dieter Kaiser  Comment By: Dieter Kaiser (crategus) Date: 20100523 22:15 Message: Some more examples of integrals which have changed. In general the following type of integral: integrate(log(x)^n, x, 0, a); where n is a positive integer has changed. Some examples: (%i1) assume(a>0)$ (%i2) integrate(log(x)^2,x,0,a); (%o2) gamma_incomplete(3,log(a)) (%i3) integrate(log(x)^2,x,0,a),gamma_expand:true; (%o3) a*log(a)^22*a*log(a)+2*a (%i4) integrate(log(x)^3,x,0,a); (%o4) gamma_incomplete(4,log(a)) (%i5) integrate(log(x)^3,x,0,a),gamma_expand:true; (%o5) a*log(a)^33*a*log(a)^2+6*a*log(a)6*a (%i6) integrate(log(x)^4,x,0,a); (%o6) gamma_incomplete(5,log(a)) (%i7) integrate(log(x)^4,x,0,a),gamma_expand:true; (%o7) a*log(a)^44*a*log(a)^3+12*a*log(a)^224*a*log(a)+24*a But we have also new integrals which do not work in Maxima 5.20. These are of the type integrate(log(x)^(n/2), x, 0, a); where n is an integer. Some examples are: (%i8) integrate(log(x)^(1/2),x,0,a); (%o8) %i*gamma_incomplete(3/2,log(a)) (%i9) integrate(log(x)^(1/2),x,0,a),gamma_expand:true; (%o9) (sqrt(%pi)*%i*erfc(sqrt(log(a)))+2*%i*a*sqrt(log(a)))/2 (%i10) integrate(log(x)^(3/2),x,0,a); (%o10) %i*(3*sqrt(%pi)*erf(sqrt(log(a)))/4 a*sqrt(log(a))*log(a)+3*a*sqrt(log(a))/2+3*sqrt(%pi)/4) (%i11) integrate(log(x)^(5/2),x,0,a); (%o11) %i*gamma_incomplete(7/2,log(a)) (%i12) integrate(log(x)^(5/2),x,0,a),gamma_expand:true; (%o12) (15*sqrt(%pi)*%i*erfc(sqrt(log(a))) +sqrt(log(a))*(8*%i*a*log(a)^220*%i*a*log(a)+30*%i*a)) /8 So the best seems to be as suggested to bind the flag $gamma_expand to T. We get the change for gamma(k1+1) because the Gamma function uses the flag $gamma_expand too. Perhaps, we should introduce a second flag $gamma_incomplete_expand to have the possibility to get a more specific expansion of the Incomplete Gamma function. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100520 01:35 Message: One possible solution. In dintlog, bind $gamma_expand to T when calling logx1. Then maxima returns a*log(a)a. However, this change causes 90 in rtestint to fail. The result is k1*gamma(k1) instead of gamma(1+k1).  Comment By: Raymond Toy (rtoy) Date: 20100518 05:31 Message: You are correct. Looks like the issue comes from dintlog. In 5.19.2, the antideriv was tried first, then logx1. In 5.21, logx1 is tried first, then antideriv. I do not know which is better. The commit message says the change helps remove the limit from some integrals. But it makes this particular integral not as nice.  Comment By: Dieter Kaiser (crategus) Date: 20100518 00:20 Message: We have the new algorithm of defintlogexp since Maxima 5.19. But the behavior for the log function has changed between 5.20 and 5.21. Therefore, I think it is not the algorithm of defintlogexp which has changed the integral of the log function, but some code which has been introduced later. Dieter Kaiser  Comment By: Raymond Toy (rtoy) Date: 20100517 22:01 Message: The new result comes from the new routine defintlogexp, which is called relatively early in defint. Perhaps it should be called later? I have not investigated this aspect yet.  Comment By: Dieter Kaiser (crategus) Date: 20100504 15:08 Message: Yes, the result has changed between Maxima 5.21 and 5.20. I do not know the reason at this time, but the new result is not really wrong. It simplifies to the old result with the flag gamma_expand: (%i1) assume(a>0)$ (%i2) integrate(log(x),x,0,a),gamma_expand:true; (%o2) a*log(a)a Dieter Kaiser  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=2996542&group_id=4933 