From: SourceForge.net <noreply@so...>  20030723 18:25:15

Bugs item #776441, was opened at 20030723 14:25 Message generated for change (Tracker Item Submitted) made by Item Submitter You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 Category: None Group: None Status: Open Resolution: None Priority: 5 Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 
From: SourceForge.net <noreply@so...>  20030725 14:26:55

Bugs item #776441, was opened at 20030723 14:25 Message generated for change (Comment added) made by macrakis You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 Category: None Group: None Status: Open Resolution: None Priority: 5 Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  >Comment By: Stavros Macrakis (macrakis) Date: 20030725 10:26 Message: Logged In: YES user_id=588346 This not only screws up SORT etc., but even basic simplification, since simplus, simptimes, etc. depend on great: q+r+s => (x+1)^2+x^2+x*(x+2) q+s+r => x^2+x*(x+2)+(x+1)^2 (q+s+r)(q+s+r) => x^2x^2 (q+s+r)(s+q+r) => x^2x^2 (q+r+s)(q+s+r) => x (x + 2)  x (x + 2)  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 
From: SourceForge.net <noreply@so...>  20030805 04:35:45

Bugs item #776441, was opened at 20030723 14:25 Message generated for change (Comment added) made by macrakis You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 Category: None Group: None Status: Open Resolution: None Priority: 5 Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  >Comment By: Stavros Macrakis (macrakis) Date: 20030805 00:35 Message: Logged In: YES user_id=588346 More amusing consequences: q+r+s => (x+1)^2+x^2+x*(x+2) expand(%,0,0) => x^2+x*(x+2)+(x+1)^2 expand(%,0,0) => x*(x+2)+(x+1)^2+x^2 expand(%,0,0) => (x+1)^2+x^2+x*(x+2) q+r+srqs => (x+1)^2+x^2+x*(x+2)(x+1)^2x^2x* (x+2) expand(%,0,0) => x^2x^2 expand(%,0,0) => 0 I haven't found an example where simptimes fails, though. Fateman reports that this bug is also found in commercial Macsyma 2.4, and calls it a Methuselah bug because it has persisted for so long  presumably it has been around for 30+ years.  Comment By: Stavros Macrakis (macrakis) Date: 20030725 10:26 Message: Logged In: YES user_id=588346 This not only screws up SORT etc., but even basic simplification, since simplus, simptimes, etc. depend on great: q+r+s => (x+1)^2+x^2+x*(x+2) q+s+r => x^2+x*(x+2)+(x+1)^2 (q+s+r)(q+s+r) => x^2x^2 (q+s+r)(s+q+r) => x^2x^2 (q+r+s)(q+s+r) => x (x + 2)  x (x + 2)  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 
From: SourceForge.net <noreply@so...>  20030808 23:50:22

Bugs item #776441, was opened at 20030723 20:25 Message generated for change (Comment added) made by wjenkner You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 Category: None Group: None Status: Open Resolution: None Priority: 5 Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  >Comment By: Wolfgang Jenkner (wjenkner) Date: 20030809 01:50 Message: Logged In: YES user_id=581700 This one doesn't even involve MEXPT (I found it while checking one of the cases needed for proving that ORDLIST implements a consistent way of extending a given total order on a set of simplified expressions to their simplified sums and products. So it doesn't...) (C1) orderlessp(t/2,t); (D1) TRUE (C2) orderlessp(t,t+1/4); (D2) TRUE (C3) orderlessp(t/2,t+1/4); (D3) FALSE The point is that t/2 is ((MTIMES SIMP) ((RAT SIMP) 1 2) $t) and, lexicographically, we have (t, 1/2) < (t, 1), (t, 0) < (t, 1/4) and (t, 1/2, *) > (t, 1/4, +). So t corresponds to (t, 1) in the first comparison and to (t, 0) in the second comparison. Trouble. Floats instead of rational numbers give the same results, by the way. This one is more like Stavros's examples. (C1) orderlessp((x+1)^2,x^21); (D1) TRUE (C2) orderlessp(x^21,x^2); (D2) TRUE (C3) orderlessp((x+1)^2,x^2); (D3) FALSE Maybe powers whose exponents are positive integers should be treated like products. Actually, ORDMEXPT does this already but it isn't always called by ORDFN, for whatever reason. Anyway, here is the patch I'm currently experimenting with (it solves all the issues reported by Stavros and also the last example above, but in light of the other example it is certainly far from being a complete solution. It might even be totally wrong since I have no reason to believe that it is more than a simple palliative and that it would make things more consistent). ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Index: simp.lisp =================================================================== RCS file: /cvsroot/maxima/maxima/src/simp.lisp,v retrieving revision 1.5 diff C2 r1.5 simp.lisp *** simp.lisp 5 Mar 2003 01:36:26 0000 1.5  simp.lisp 8 Aug 2003 19:10:57 0000 *************** *** 1848,1854 **** ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT) (MPLUSP (CADR Y))) (NOT (ORDMEXPT Y X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X)))  1848,1854  ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT)) (NOT (ORDMEXPT Y X))) + ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X))) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~  Comment By: Stavros Macrakis (macrakis) Date: 20030805 06:35 Message: Logged In: YES user_id=588346 More amusing consequences: q+r+s => (x+1)^2+x^2+x*(x+2) expand(%,0,0) => x^2+x*(x+2)+(x+1)^2 expand(%,0,0) => x*(x+2)+(x+1)^2+x^2 expand(%,0,0) => (x+1)^2+x^2+x*(x+2) q+r+srqs => (x+1)^2+x^2+x*(x+2)(x+1)^2x^2x* (x+2) expand(%,0,0) => x^2x^2 expand(%,0,0) => 0 I haven't found an example where simptimes fails, though. Fateman reports that this bug is also found in commercial Macsyma 2.4, and calls it a Methuselah bug because it has persisted for so long  presumably it has been around for 30+ years.  Comment By: Stavros Macrakis (macrakis) Date: 20030725 16:26 Message: Logged In: YES user_id=588346 This not only screws up SORT etc., but even basic simplification, since simplus, simptimes, etc. depend on great: q+r+s => (x+1)^2+x^2+x*(x+2) q+s+r => x^2+x*(x+2)+(x+1)^2 (q+s+r)(q+s+r) => x^2x^2 (q+s+r)(s+q+r) => x^2x^2 (q+r+s)(q+s+r) => x (x + 2)  x (x + 2)  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 
From: SourceForge.net <noreply@so...>  20060708 17:13:17

Bugs item #776441, was opened at 20030723 12:25 Message generated for change (Settings changed) made by robert_dodier You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&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 Group: None Status: Open Resolution: None >Priority: 7 Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  >Comment By: Robert Dodier (robert_dodier) Date: 20060708 11:13 Message: Logged In: YES user_id=501686 Increasing the priority on this one  potential for subtle breakage in various contexts.  Comment By: Wolfgang Jenkner (wjenkner) Date: 20030808 17:50 Message: Logged In: YES user_id=581700 This one doesn't even involve MEXPT (I found it while checking one of the cases needed for proving that ORDLIST implements a consistent way of extending a given total order on a set of simplified expressions to their simplified sums and products. So it doesn't...) (C1) orderlessp(t/2,t); (D1) TRUE (C2) orderlessp(t,t+1/4); (D2) TRUE (C3) orderlessp(t/2,t+1/4); (D3) FALSE The point is that t/2 is ((MTIMES SIMP) ((RAT SIMP) 1 2) $t) and, lexicographically, we have (t, 1/2) < (t, 1), (t, 0) < (t, 1/4) and (t, 1/2, *) > (t, 1/4, +). So t corresponds to (t, 1) in the first comparison and to (t, 0) in the second comparison. Trouble. Floats instead of rational numbers give the same results, by the way. This one is more like Stavros's examples. (C1) orderlessp((x+1)^2,x^21); (D1) TRUE (C2) orderlessp(x^21,x^2); (D2) TRUE (C3) orderlessp((x+1)^2,x^2); (D3) FALSE Maybe powers whose exponents are positive integers should be treated like products. Actually, ORDMEXPT does this already but it isn't always called by ORDFN, for whatever reason. Anyway, here is the patch I'm currently experimenting with (it solves all the issues reported by Stavros and also the last example above, but in light of the other example it is certainly far from being a complete solution. It might even be totally wrong since I have no reason to believe that it is more than a simple palliative and that it would make things more consistent). ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Index: simp.lisp =================================================================== RCS file: /cvsroot/maxima/maxima/src/simp.lisp,v retrieving revision 1.5 diff C2 r1.5 simp.lisp *** simp.lisp 5 Mar 2003 01:36:26 0000 1.5  simp.lisp 8 Aug 2003 19:10:57 0000 *************** *** 1848,1854 **** ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT) (MPLUSP (CADR Y))) (NOT (ORDMEXPT Y X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X)))  1848,1854  ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT)) (NOT (ORDMEXPT Y X))) + ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X))) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~  Comment By: Stavros Macrakis (macrakis) Date: 20030804 22:35 Message: Logged In: YES user_id=588346 More amusing consequences: q+r+s => (x+1)^2+x^2+x*(x+2) expand(%,0,0) => x^2+x*(x+2)+(x+1)^2 expand(%,0,0) => x*(x+2)+(x+1)^2+x^2 expand(%,0,0) => (x+1)^2+x^2+x*(x+2) q+r+srqs => (x+1)^2+x^2+x*(x+2)(x+1)^2x^2x* (x+2) expand(%,0,0) => x^2x^2 expand(%,0,0) => 0 I haven't found an example where simptimes fails, though. Fateman reports that this bug is also found in commercial Macsyma 2.4, and calls it a Methuselah bug because it has persisted for so long  presumably it has been around for 30+ years.  Comment By: Stavros Macrakis (macrakis) Date: 20030725 08:26 Message: Logged In: YES user_id=588346 This not only screws up SORT etc., but even basic simplification, since simplus, simptimes, etc. depend on great: q+r+s => (x+1)^2+x^2+x*(x+2) q+s+r => x^2+x*(x+2)+(x+1)^2 (q+s+r)(q+s+r) => x^2x^2 (q+s+r)(s+q+r) => x^2x^2 (q+r+s)(q+s+r) => x (x + 2)  x (x + 2)  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 
From: SourceForge.net <noreply@so...>  20091214 00:29:08

Bugs item #776441, was opened at 20030723 14:25 Message generated for change (Settings changed) made by dgildea You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&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 Group: None >Status: Pending >Resolution: Fixed Priority: 7 Private: No Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  >Comment By: Dan Gildea (dgildea) Date: 20091213 19:29 Message: I think these issues are resolved in simp.lisp rev 1.93.  Comment By: Robert Dodier (robert_dodier) Date: 20060708 13:13 Message: Logged In: YES user_id=501686 Increasing the priority on this one  potential for subtle breakage in various contexts.  Comment By: Wolfgang Jenkner (wjenkner) Date: 20030808 19:50 Message: Logged In: YES user_id=581700 This one doesn't even involve MEXPT (I found it while checking one of the cases needed for proving that ORDLIST implements a consistent way of extending a given total order on a set of simplified expressions to their simplified sums and products. So it doesn't...) (C1) orderlessp(t/2,t); (D1) TRUE (C2) orderlessp(t,t+1/4); (D2) TRUE (C3) orderlessp(t/2,t+1/4); (D3) FALSE The point is that t/2 is ((MTIMES SIMP) ((RAT SIMP) 1 2) $t) and, lexicographically, we have (t, 1/2) < (t, 1), (t, 0) < (t, 1/4) and (t, 1/2, *) > (t, 1/4, +). So t corresponds to (t, 1) in the first comparison and to (t, 0) in the second comparison. Trouble. Floats instead of rational numbers give the same results, by the way. This one is more like Stavros's examples. (C1) orderlessp((x+1)^2,x^21); (D1) TRUE (C2) orderlessp(x^21,x^2); (D2) TRUE (C3) orderlessp((x+1)^2,x^2); (D3) FALSE Maybe powers whose exponents are positive integers should be treated like products. Actually, ORDMEXPT does this already but it isn't always called by ORDFN, for whatever reason. Anyway, here is the patch I'm currently experimenting with (it solves all the issues reported by Stavros and also the last example above, but in light of the other example it is certainly far from being a complete solution. It might even be totally wrong since I have no reason to believe that it is more than a simple palliative and that it would make things more consistent). ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Index: simp.lisp =================================================================== RCS file: /cvsroot/maxima/maxima/src/simp.lisp,v retrieving revision 1.5 diff C2 r1.5 simp.lisp *** simp.lisp 5 Mar 2003 01:36:26 0000 1.5  simp.lisp 8 Aug 2003 19:10:57 0000 *************** *** 1848,1854 **** ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT) (MPLUSP (CADR Y))) (NOT (ORDMEXPT Y X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X)))  1848,1854  ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT)) (NOT (ORDMEXPT Y X))) + ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X))) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~  Comment By: Stavros Macrakis (macrakis) Date: 20030805 00:35 Message: Logged In: YES user_id=588346 More amusing consequences: q+r+s => (x+1)^2+x^2+x*(x+2) expand(%,0,0) => x^2+x*(x+2)+(x+1)^2 expand(%,0,0) => x*(x+2)+(x+1)^2+x^2 expand(%,0,0) => (x+1)^2+x^2+x*(x+2) q+r+srqs => (x+1)^2+x^2+x*(x+2)(x+1)^2x^2x* (x+2) expand(%,0,0) => x^2x^2 expand(%,0,0) => 0 I haven't found an example where simptimes fails, though. Fateman reports that this bug is also found in commercial Macsyma 2.4, and calls it a Methuselah bug because it has persisted for so long  presumably it has been around for 30+ years.  Comment By: Stavros Macrakis (macrakis) Date: 20030725 10:26 Message: Logged In: YES user_id=588346 This not only screws up SORT etc., but even basic simplification, since simplus, simptimes, etc. depend on great: q+r+s => (x+1)^2+x^2+x*(x+2) q+s+r => x^2+x*(x+2)+(x+1)^2 (q+s+r)(q+s+r) => x^2x^2 (q+s+r)(s+q+r) => x^2x^2 (q+r+s)(q+s+r) => x (x + 2)  x (x + 2)  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 
From: SourceForge.net <noreply@so...>  20091228 02:20:38

Bugs item #776441, was opened at 20030723 18:25 Message generated for change (Comment added) made by sfrobot You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&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 Group: None >Status: Closed Resolution: Fixed Priority: 7 Private: No Submitted By: Stavros Macrakis (macrakis) Assigned to: Nobody/Anonymous (nobody) Summary: orderlessp not transitive Initial Comment: l: [z+x*(x+2)+v+1,z+x^2+x+v+1,z+(x+1)^2+v]; orderlessp(l[1],l[2]) => True orderlessp(l[2],l[3]) => True orderlessp(l[1],l[3]) => False !!! More concise example: q: x^2; r: (x+1)^2; s: x*(x+2); orderlessp(q,r) => true orderlessp(r,s) => true orderlessp(s,q) => true That is, s<q<r<s. The problem is somewhere in the internal great function, which by the way does some strange things, in particular: why does ordlist have an explicit check for mplus: (RETURN (COND ((= L2 0) (EQ CX 'MPLUS)) (Thanks to Barton for his contributions to tracking this down.) Maxima 5.9.0 GCL 2.5.0  >Comment By: SourceForge Robot (sfrobot) Date: 20091228 02:20 Message: This Tracker item was closed automatically by the system. It was previously set to a Pending status, and the original submitter did not respond within 14 days (the time period specified by the administrator of this Tracker).  Comment By: Dan Gildea (dgildea) Date: 20091214 00:29 Message: I think these issues are resolved in simp.lisp rev 1.93.  Comment By: Robert Dodier (robert_dodier) Date: 20060708 17:13 Message: Logged In: YES user_id=501686 Increasing the priority on this one  potential for subtle breakage in various contexts.  Comment By: Wolfgang Jenkner (wjenkner) Date: 20030808 23:50 Message: Logged In: YES user_id=581700 This one doesn't even involve MEXPT (I found it while checking one of the cases needed for proving that ORDLIST implements a consistent way of extending a given total order on a set of simplified expressions to their simplified sums and products. So it doesn't...) (C1) orderlessp(t/2,t); (D1) TRUE (C2) orderlessp(t,t+1/4); (D2) TRUE (C3) orderlessp(t/2,t+1/4); (D3) FALSE The point is that t/2 is ((MTIMES SIMP) ((RAT SIMP) 1 2) $t) and, lexicographically, we have (t, 1/2) < (t, 1), (t, 0) < (t, 1/4) and (t, 1/2, *) > (t, 1/4, +). So t corresponds to (t, 1) in the first comparison and to (t, 0) in the second comparison. Trouble. Floats instead of rational numbers give the same results, by the way. This one is more like Stavros's examples. (C1) orderlessp((x+1)^2,x^21); (D1) TRUE (C2) orderlessp(x^21,x^2); (D2) TRUE (C3) orderlessp((x+1)^2,x^2); (D3) FALSE Maybe powers whose exponents are positive integers should be treated like products. Actually, ORDMEXPT does this already but it isn't always called by ORDFN, for whatever reason. Anyway, here is the patch I'm currently experimenting with (it solves all the issues reported by Stavros and also the last example above, but in light of the other example it is certainly far from being a complete solution. It might even be totally wrong since I have no reason to believe that it is more than a simple palliative and that it would make things more consistent). ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Index: simp.lisp =================================================================== RCS file: /cvsroot/maxima/maxima/src/simp.lisp,v retrieving revision 1.5 diff C2 r1.5 simp.lisp *** simp.lisp 5 Mar 2003 01:36:26 0000 1.5  simp.lisp 8 Aug 2003 19:10:57 0000 *************** *** 1848,1854 **** ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT) (MPLUSP (CADR Y))) (NOT (ORDMEXPT Y X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X)))  1848,1854  ((MEMQ CX '(MPLUS MTIMES)) (COND ((MEMQ CY '(MPLUS MTIMES)) (ORDLIST (CDR X) (CDR Y) CX CY)) ! ((AND (EQ CX 'MPLUS) (EQ CY 'MEXPT)) (NOT (ORDMEXPT Y X))) + ((ALIKE1 (SETQ U (CAR (LAST X))) Y) (NOT (ORDHACK X))) (T (GREAT U Y)))) ((MEMQ CY '(MPLUS MTIMES)) (NOT (ORDFN Y X))) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cut ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~  Comment By: Stavros Macrakis (macrakis) Date: 20030805 04:35 Message: Logged In: YES user_id=588346 More amusing consequences: q+r+s => (x+1)^2+x^2+x*(x+2) expand(%,0,0) => x^2+x*(x+2)+(x+1)^2 expand(%,0,0) => x*(x+2)+(x+1)^2+x^2 expand(%,0,0) => (x+1)^2+x^2+x*(x+2) q+r+srqs => (x+1)^2+x^2+x*(x+2)(x+1)^2x^2x* (x+2) expand(%,0,0) => x^2x^2 expand(%,0,0) => 0 I haven't found an example where simptimes fails, though. Fateman reports that this bug is also found in commercial Macsyma 2.4, and calls it a Methuselah bug because it has persisted for so long  presumably it has been around for 30+ years.  Comment By: Stavros Macrakis (macrakis) Date: 20030725 14:26 Message: Logged In: YES user_id=588346 This not only screws up SORT etc., but even basic simplification, since simplus, simptimes, etc. depend on great: q+r+s => (x+1)^2+x^2+x*(x+2) q+s+r => x^2+x*(x+2)+(x+1)^2 (q+s+r)(q+s+r) => x^2x^2 (q+s+r)(s+q+r) => x^2x^2 (q+r+s)(q+s+r) => x (x + 2)  x (x + 2)  You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=104933&aid=776441&group_id=4933 