#2400 closed defect (fixed)
[with patch, positive review] maxima automatically simplifies floats to rationals
Reported by: | jason | Owned by: | was |
---|---|---|---|
Priority: | major | Milestone: | sage-2.10.4 |
Component: | interfaces | Keywords: | |
Cc: | Merged in: | ||
Authors: | Reviewers: | ||
Report Upstream: | N/A | Work issues: | |
Branch: | Commit: | ||
Dependencies: | Stopgaps: |
Description
I tried doing some integrals today and the output doesn't make much sense to me: sage: f = e^(-x2) sage: f.integrate(x, 0, 0.1) 2066*sqrt(pi)/36741 sage: f.integrate(x, 0, 1/10) sqrt(pi)*erf(1/10)/2 Hmmmm. Does this mean erf(1/10) is a rational number? That's a little surprising to me. In fact: sage: RR(f.integrate(x, 0, 0.1)) 0.0996676643523801 sage: RR(f.integrate(x, 0, 1/10)) 0.0996676642903363 What's going on here? david
Attachments (2)
Change History (16)
comment:1 Changed 9 years ago by
comment:2 Changed 9 years ago by
Jason, please improve the comment "# no ascii art output" to indicate what the new flag does.
Changed 9 years ago by
comment:3 Changed 9 years ago by
- Summary changed from maxima automatically simplifies floats to rationals to [with patch, needs review] maxima automatically simplifies floats to rationals
I replaced the patch with an updated one addressing dmharvey's concerns. The patch is ready to be reviewed again.
comment:4 Changed 9 years ago by
This patch fails doctests on
sage -t devel/sage-2400/sage/matrix/matrix_symbolic_dense.pyx
(There might be more; I don't know, I killed it.)
comment:5 Changed 9 years ago by
- Summary changed from [with patch, needs review] maxima automatically simplifies floats to rationals to [with patch, Negative review] maxima automatically simplifies floats to rationals
The code itself is fine but symbolic matrices with keepfloat appear to be broken in Maxima. When/if symbolic matrices do not use maxima this will be a good patch.
TypeError: Error executing code in Maxima CODE: sage173 : matrixexp(sage172)$ Maxima ERROR: `rat' replaced 1.0 by 1/1 = 1.0 `rat' replaced -5.2 by -26/5 = -5.2 `rat' replaced -12.0 by -12/1 = -12.0 `rat' replaced 2.0 by 2/1 = 2.0 `rat' replaced -5.2 by -26/5 = -5.2 `rat' replaced 5.0 by 5/1 = 5.0 `rat' replaced -26.0 by -26/1 = -26.0 `rat' replaced -60.0 by -60/1 = -60.0 `rat' replaced 10.0 by 10/1 = 10.0 `rat' replaced -26.0 by -26/1 = -26.0 `rat' replaced -12.0 by -12/1 = -12.0 `rat' replaced -5.2 by -26/5 = -5.2 `rat' replaced 1.0 by 1/1 = 1.0 Unable to find the spectral representation
comment:6 Changed 9 years ago by
- Summary changed from [with patch, Negative review] maxima automatically simplifies floats to rationals to [with patch, under discussion] maxima automatically simplifies floats to rationals
The matrixexp command seems to work fine in maxima (see below). How do you reproduce your error? (can you give a short sage session that gives the error above?)
sage: maxima.interact() --> Switching to Maxima <-- maxima: keepfloat: true true maxima: matrixexp(matrix([x,1],[1,x])); matrix([(%e^2+1)*%e^(x-1)/2,(%e^2-1)*%e^(x-1)/2],[(%e^2-1)*%e^(x-1)/2,(%e^2+1)*%e^(x-1)/2]) maxima: keepfloat: false false maxima: matrixexp(matrix([x,1],[1,x])); matrix([(%e^2+1)*%e^(x-1)/2,(%e^2-1)*%e^(x-1)/2],[(%e^2-1)*%e^(x-1)/2,(%e^2+1)*%e^(x-1)/2]) maxima: keepfloat: false false maxima: matrixexp(matrix([3,1],[1,2])); matrix([%e^(5/2-sqrt(5)/2)*((sqrt(5)+5)*%e^sqrt(5)-sqrt(5)+5)/10,%e^(5/2-sqrt(5)/2)*(sqrt(5)*%e^sqrt(5)-sqrt(5))/5],[%e^(5/2-sqrt(5)/2)*(sqrt(5)*%e^sqrt(5)-sqrt(5))/5,-%e^(5/2-sqrt(5)/2)*((sqrt(5)-5)*%e^sqrt(5)-sqrt(5)-5)/10]) maxima: keepfloat: true true maxima: matrixexp(matrix([3,1],[1,2])); matrix([%e^(5/2-sqrt(5)/2)*((sqrt(5)+5)*%e^sqrt(5)-sqrt(5)+5)/10,%e^(5/2-sqrt(5)/2)*(sqrt(5)*%e^sqrt(5)-sqrt(5))/5],[%e^(5/2-sqrt(5)/2)*(sqrt(5)*%e^sqrt(5)-sqrt(5))/5,-%e^(5/2-sqrt(5)/2)*((sqrt(5)-5)*%e^sqrt(5)-sqrt(5)-5)/10])
comment:7 Changed 9 years ago by
Here is a way to reproduce the above error:
maxima: keepfloat: true true maxima: matrixexp(matrix([3.0,1.0],[1.0,2.0])); --------------------------------------------------------------------------- <type 'exceptions.TypeError'> Traceback (most recent call last) /home/grout/sage/local/lib/python2.5/site-packages/sage/misc/interpreter.py in sage_prefilter(self, block, continuation) 406 for i in range(len(B)): 407 L = B[i] --> 408 M = do_prefilter_paste(L, continuation or (not first)) 409 first = False 410 # The L[:len(L)-len(L.lstrip())] business here preserves /home/grout/sage/local/lib/python2.5/site-packages/sage/misc/interpreter.py in do_prefilter_paste(line, continuation) 347 348 if len(line) > 0: --> 349 line = preparser_ipython.preparse_ipython(line, not continuation) 350 return line 351 /home/grout/sage/local/lib/python2.5/site-packages/sage/misc/preparser_ipython.py in preparse_ipython(line, reset) 104 # TODO: do sage substitutions here 105 #t = interface._eval_line(line) --> 106 t = interface.eval(line) 107 108 import sage.misc.interpreter /home/grout/sage/local/lib/python2.5/site-packages/sage/interfaces/expect.py in eval(self, code, strip, **kwds) 712 raise 713 except TypeError, s: --> 714 raise TypeError, 'error evaluating "%s":\n%s'%(code,s) 715 716 def execute(self, *args, **kwds): <type 'exceptions.TypeError'>: error evaluating "matrixexp(matrix([3.0,1.0],[1.0,2.0]));": Error executing code in Maxima CODE: matrixexp(matrix([3.0,1.0],[1.0,2.0])); Maxima ERROR: `rat' replaced 1.0 by 1/1 = 1.0 `rat' replaced -5.0 by -5/1 = -5.0 `rat' replaced 5.0 by 5/1 = 5.0 `rat' replaced 2.0 by 2/1 = 2.0 `rat' replaced -5.0 by -5/1 = -5.0 `rat' replaced 1.0 by 1/1 = 1.0 `rat' replaced -5.0 by -5/1 = -5.0 `rat' replaced 5.0 by 5/1 = 5.0 `rat' replaced 2.0 by 2/1 = 2.0 `rat' replaced -5.0 by -5/1 = -5.0 `rat' replaced 5.0 by 5/1 = 5.0 `rat' replaced -5.0 by -5/1 = -5.0 `rat' replaced 1.0 by 1/1 = 1.0 Unable to find the spectral representation maxima: keepfloat: false false maxima: matrixexp(matrix([3.0,1.0],[1.0,2.0])); matrix([%e^(5/2-sqrt(5)/2)*((sqrt(5)+5)*%e^sqrt(5)-sqrt(5)+5)/10,%e^(5/2-sqrt(5)/2)*(sqrt(5)*%e^sqrt(5)-sqrt(5))/5],[%e^(5/2-sqrt(5)/2)*(sqrt(5)*%e^sqrt(5)-sqrt(5))/5,-%e^(5/2-sqrt(5)/2)*((sqrt(5)-5)*%e^sqrt(5)-sqrt(5)-5)/10]) maxima:
comment:8 Changed 9 years ago by
This post talks about the weakness of the maxima linear algebra functionality and explicitly includes the above error: http://www.ma.utexas.edu/pipermail/maxima/2006/003031.html
comment:9 Changed 9 years ago by
The last post at http://www.nabble.com/matrix-exponential--td6816907.html suggests using the "diag" package instead of the "linearalgebra" package for calculating the matrix exponential. This works with keepfloat:
sage: maxima.interact() --> Switching to Maxima <-- maxima: keepfloat: true true maxima: load("diag") ?\/home\/grout\/sage\/local\/share\/maxima\/5\.13\.0\/share\/contrib\/diag\.mac maxima: mat_function(exp,matrix([3.0,1.0],[1.0,2.0])); matrix([(sqrt(5)+1)*%e^((sqrt(5)+5)/2)/(2*sqrt(5))+(sqrt(5)-1)*%e^-((sqrt(5)-5)/2)/(2*sqrt(5)),%e^((sqrt(5)+5)/2)/sqrt(5)-%e^-((sqrt(5)-5)/2)/sqrt(5)],[%e^((sqrt(5)+5)/2)/sqrt(5)-%e^-((sqrt(5)-5)/2)/sqrt(5),2*%e^((sqrt(5)+5)/2)/(sqrt(5)*(sqrt(5)+1))+2*%e^-((sqrt(5)-5)/2)/(sqrt(5)*(sqrt(5)-1))])
comment:10 Changed 9 years ago by
- Summary changed from [with patch, under discussion] maxima automatically simplifies floats to rationals to [with patch, needs review] maxima automatically simplifies floats to rationals
The maxima matrixexp function has problems with floating point entries. I don't think that should keep us from fixing *wrong* output (or at least, unnecessarily rounded) from maxima due to their rounding floating points to rationals.
I've attached a patch which works around the problems in the matrixexp function by specifying keepfloat: false for that one command. I've also added a note talking about maxima automatically rounding floating point numbers to the docstring for matrixexp.
comment:11 Changed 9 years ago by
- Milestone changed from sage-2.11 to sage-2.10.4
comment:12 Changed 9 years ago by
- Summary changed from [with patch, needs review] maxima automatically simplifies floats to rationals to [with patch, positive review] maxima automatically simplifies floats to rationals
These two patches apply against 2.10.4.alpha0, fix the problems, and pass tests for me.
comment:13 Changed 9 years ago by
- Resolution set to fixed
- Status changed from new to closed
Merged both patches in Sage 2.10.4.rc0
comment:14 Changed 3 years ago by
- Report Upstream set to N/A
For some reason this hack doesn't work anymore in Maxima 5.30, see #13973.
Following a lead from the mailing list, I added "keepfloat: true" to the init_code for the maxima interfaces. The patch is attached.