SR.var('x', domain='real') and assume(x, 'real') are very slow

[mkoeppe]

(from #30061)

See also:

• https://trac.sagemath.org/wiki/symbolics#variabledomainsandassumptions

CC: @egourgoulhon @sagetrac-tmonteil @orlitzky

Component: symbolics

Author: Matthias Koeppe

Branch/Commit: 1fbfb68

Reviewer: Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/30065

[mkoeppe]

comment:2

sage: %prun for i in range(10000): x = SR.var('x', domain='real')
   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
  1030001   79.176    0.000   87.914    0.000 maxima_lib.py:412(_eval_line)
   500000    5.826    0.000    6.854    0.000 maxima_lib.py:274(max_to_string)
   230000    1.408    0.000   11.558    0.000 interface.py:1146(_repr_)
   500000    1.027    0.000    1.027    0.000 {method 'python' of 'sage.libs.ecl.EclObject' objects}
   250000    0.944    0.000   14.659    0.000 maxima_abstract.py:1610(__len__)

[mkoeppe]

comment:3

sage: %prun for i in range(3000): x = SR.var('x', domain='real')
   309000   23.496    0.000   26.177    0.000 maxima_lib.py:412(_eval_line)
sage: %prun for i in range(3000): assume(x, 'real')
   309000   29.508    0.000   32.219    0.000 maxima_lib.py:412(_eval_line)

Gets slower every time.

sage: %prun for i in range(3000): assume(x, 'real')
   309000   39.085    0.000   42.002    0.000 maxima_lib.py:412(_eval_line)
sage: %prun for i in range(3000): assume(x, 'real')
   309000   54.613    0.000   57.968    0.000 maxima_lib.py:412(_eval_line)

[mkoeppe]

Branch: u/mkoeppe/faster_maxima_assumptions

[mkoeppe]

Commit: 134da39

[mkoeppe]

comment:5

Some low-hanging fruit here.

---

New commits:

134da39

sage.symbolic.assumptions.GenericDeclaration.assume: Validate against cached valid features first

[mkoeppe]

comment:6

sage: %prun for i in range(10000): x = SR.var('x', domain='real')
   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    70097    8.210    0.000    8.683    0.000 maxima_lib.py:412(_eval_line)
       25    0.651    0.026    0.651    0.026 {built-in method sage.libs.ecl.ecl_eval}
    10000    0.298    0.000   12.105    0.001 {method 'var' of 'sage.symbolic.ring.SymbolicRing' objects}
    20048    0.273    0.000    0.325    0.000 maxima_lib.py:274(max_to_string)

[mkoeppe]

comment:7

More can be done by caching GenericDeclaration:

diff --git a/src/sage/symbolic/assumptions.py b/src/sage/symbolic/assumptions.py
index 3e0ca736ba..827ab61b67 100644
--- a/src/sage/symbolic/assumptions.py
+++ b/src/sage/symbolic/assumptions.py
@@ -71,11 +71,13 @@ Assumptions are added and in some cases checked for consistency::
 from sage.structure.sage_object import SageObject
 from sage.rings.all import ZZ, QQ, RR, CC
 from sage.symbolic.ring import is_SymbolicVariable
+from sage.structure.unique_representation import UniqueRepresentation
+
 _assumptions = []
 
 _valid_feature_strings = set()
 
-class GenericDeclaration(SageObject):
+class GenericDeclaration(UniqueRepresentation):
     """
     This class represents generic assumptions, such as a variable being
     an integer or a function being increasing. It passes such

... but the semantics of the assume and forget methods are not very clear...

[mkoeppe]

comment:9

I'll investigate further improvements on follow-up tickets. Needs review

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

comment:10

Follow-up: #30074

[tscrim]

comment:11

Perhaps it would be better to do

-_valid_feature_strings = set(repr(x).strip() for x in list(maxima("features")))
+_valid_feature_strings.update(repr(x).strip() for x in list(maxima("features")))

so it doesn't overwrite the _valid_feature_strings but instead just adds to it?

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

1fbfb68

Update instead of overwriting

[sagetrac-git]

Changed commit from 134da39 to 1fbfb68

[mkoeppe]

comment:13

Good idea, done.

[tscrim]

Reviewer: Travis Scrimshaw

[tscrim]

comment:14

Thanks.

[mkoeppe]

comment:15

Thank you!

[mjungmath]

comment:16

Patchbot is complaining. Apparently, this is due to the preexisting code (line 971):

https://www.flake8rules.com/rules/E701.html

Should this be fixed right on the fly?

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

999852d

src/sage/symbolic/assumptions.py: Fix style warning

[sagetrac-git]

Changed commit from 1fbfb68 to 999852d

[sagetrac-git]

Changed commit from 999852d to 1fbfb68

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

[mkoeppe]

comment:19

Actually no, this ticket is already on Volker's branch.

[mkoeppe]

comment:20

I've put the commit on #30074 instead

[vbraun]

Changed branch from u/mkoeppe/faster_maxima_assumptions to 1fbfb68