32
$\begingroup$

What is a simple, fast way to test whether an expression is a real-valued number? I ask since there is no RealQ function.

If we call this test realQ, it should satisfy these constraints:

  • realQ["text"] is False (non numerics are all false)
  • realQ[0] is True (integers are true)
  • realQ[3.0] is True (reals are true)
  • realQ[1/2] is True (rationals are true)
  • realQ[I] is False (anything with an imaginary component is false)
Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ Background: I thought of this as I was writing code to validate xmin or xmax for Plot[f[x],{,xmin,xmax}] in a web form. Even though I could write something, I thought there's probably more than one way to skin this cat, and it would make a good question for this site. You guys did not disappoint! $\endgroup$
    – Joel Klein
    Commented Jan 29, 2013 at 14:37

7 Answers 7

Reset to default
28
$\begingroup$

Update: 

 Internal`RealValuedNumericQ /@ {1, N[Pi], 1/2, Sin[1.], Pi, 3/4, aa,  I}
 (* {True, True, True, True, True, True, False, False} *)

or

 Internal`RealValuedNumberQ /@ {1, N[Pi], 1/2, Sin[1.], Pi, 3/4, aa, I}
 (* {True, True, True, True, False, True, False, False} *)

Using @RM's test list

 listRM = With[{n = 10^5},
  RandomSample[Flatten[{RandomChoice[CharacterRange["A", "z"], n],
  RandomInteger[100, n],
  RandomReal[1, n],
  RandomComplex[1, n],
  RandomInteger[100, n]/RandomInteger[{1, 100}, n],
  Unevaluated@Pause@5}], 5 n + 1]];

and his realQ

 ClearAll@realQrm
 SetAttributes[realQrm, Listable] 
 realQrm[_Real | _Integer | _Rational] := True 
 realQrm[_] := False

timings 

 realQrm@listRM; // AbsoluteTiming
 (* {0.458046, Null}  *)

 Internal`RealValuedNumericQ /@ listRM; // AbsoluteTiming
 (* {0.247025, Null} *)

 Internal`RealValuedNumberQ /@ listRM; // AbsoluteTiming
 (* {0.231023, Null} *)

 realQ = NumberQ[#] && ! MatchQ[#, _Complex] &
 realQ /@ {1, N[Pi], 1/2, Sin[1.], 3/4, aa, I}
 (* {True, True, True, True, True, False, False} *)

or

realQ2 = NumericQ[#] && ! MatchQ[#, _Complex] &
realQ3 = NumericQ[#] && FreeQ[#, _Complex] &
Improve this answer
$\endgroup$
4
  • $\begingroup$ This is pretty close to what I started with, although I put everything into a single MatchQ. $\endgroup$
    – Joel Klein
    Commented Jan 28, 2013 at 23:44
  • $\begingroup$ Ad update: nice find! $\endgroup$
    – sebhofer
    Commented Jan 29, 2013 at 9:23
  • $\begingroup$ Accepting, since this fulfills my original request for simple and fast. It was hard to decide, what we have now are all very good answers. $\endgroup$
    – Joel Klein
    Commented Feb 1, 2013 at 18:19
  • $\begingroup$ @Joel, thanks for the accept. $\endgroup$
    – kglr
    Commented Feb 2, 2013 at 2:30
23
$\begingroup$

As the responses show, there are a number of quick "probably real" tests. In general, the problem is undecidable, however. This is an easy corollary of Richardson's theorem, which says that it is impossible to decide if two real expressions $x$ and $y$ are equal. Assuming Richardson's theorem, note that $(x-y)i$ is real if and only if $x=y$.

As a more mundane example, that arises in common practice with Mathematica, consider the polynomial $p(x)=13x^3-13x-1$. It's easy to see that all three roots are real (even if they don't look it), yet they don't pass any of the test here.

roots = x /. Solve[13 x^3 - 13 x - 1 == 0, x]
Internal`RealValuedNumericQ /@ roots

enter image description here

Improve this answer
$\endgroup$
20
$\begingroup$ 
RealQ[x_] := Element[x, Reals] === True

It fulfills all your samples and I think is generally correct.

Improve this answer
$\endgroup$
4
  • 1
    $\begingroup$ This is probably the most readable so far. $\endgroup$
    – Joel Klein
    Commented Jan 28, 2013 at 23:43
  • 11
    $\begingroup$ TrueQ@Element[x, Reals] $\endgroup$
    – Rojo
    Commented Jan 29, 2013 at 0:50
  • 2
    $\begingroup$ In case it's not clear -- this proposed test, too, fails with Mark McClure's example of the roots of $13x^3-13x-1$. $\endgroup$
    – murray
    Commented Sep 10, 2013 at 15:10
  • 1
    $\begingroup$ Why do you compare the result of Element[...] to True? $\endgroup$
    – Qbyte
    Commented Mar 26, 2020 at 16:25
15
$\begingroup$

I think a solution based on pattern matching will be much faster than using Element (which is more mathematical in nature) or only pattern tests or anything else that forces evaluation, since we can bypass the main evaluator. However, it is not possible to completely escape evaluation, because there can be infinitely large number of possibilities for a real number that cannot be matched solely by pattern matching. Hence, the following tries to delegate as much as possible to the pattern matcher and evaluates only what's necessary. The unfortunate consequence is that it is no longer immune to prank entries such as Unevaluated@Pause@10.

ClearAll@realQ
SetAttributes[realQ, Listable]
realQ[_Real | _Integer | _Rational] := True

realQ[Catalan | ChampernowneNumber | Degree | E | EulerGamma | Internal`Euler2Gamma |
      Glaisher | GoldenRatio | Khinchin | MachinePrecision | Pi] := True

realQ[Complex[_, 0.]] := True
realQ[x_] := NumericQ[x]

realQ[{"text", 0, 3.0, 1/2, I, Pi, 1 + 0. I}]
(* {False, True, True, True, False, True, True} *)

The list in the second definition was obtained using 

Select[Names["*`*"], MemberQ[Attributes@#, Constant] &]
Improve this answer
$\endgroup$
10
  • $\begingroup$ Pretty direct translation of the spec I gave, and listable to boot. I wish it didn't have the 3 alternatives of Real, Integer, and Rational though, I kind of like starting with the knowledge that we have a numeric (eliminate strings and such) and then exclude complex. $\endgroup$
    – Joel Klein
    Commented Jan 28, 2013 at 23:46
  • $\begingroup$ @Joel You mentioned expression generally and wanted it to also be fast & simple and I think this satisfies both. This is faster than the others and also doesn't evaluate anything. For example, compare the speeds with list = With[{n = 10^5}, RandomSample[ Flatten[{RandomChoice[CharacterRange["A", "z"], n], RandomInteger[100, n], RandomReal[1, n], RandomComplex[1, n], RandomInteger[100, n]/RandomInteger[{1, 100}, n], Unevaluated@Pause@5}], 5 n + 1]];. Evaluation in the other solutions makes their timing on the order of 5s, compared to 0.3s for mine (on my machine). $\endgroup$
    – rm -rf ♦
    Commented Jan 28, 2013 at 23:56
  • $\begingroup$ Good explanation. $\endgroup$
    – Joel Klein
    Commented Jan 29, 2013 at 2:27
  • $\begingroup$ What about Pi? $\endgroup$
    – Mark McClure
    Commented Jan 29, 2013 at 12:24
  • $\begingroup$ You can add realQ[c_Complex] := PossibleZeroQ[Im[c]] if you want 0.I etc to be real $\endgroup$
    – ssch
    Commented Jan 29, 2013 at 13:29
7
$\begingroup$

I wasn't planning to add an answer, but this now seems like it has its place in this fine list of answers:

realQ[x_?NumericQ] := Head[x] =!= Complex
realQ[_] := False

While maybe not the absolute fastest, it is fast and also relatively simple and uses only System` functions.

Improve this answer
$\endgroup$
1
  • 1
    $\begingroup$ This works only until quaternions are not properly integrated into Mathematica :) $\endgroup$
    – István Zachar
    Commented Sep 10, 2013 at 12:54
5
$\begingroup$

Since V 13.3 we have RealValuedNumberQ and RealValuedNumericQ

col[dt_, f_] := ToString[f] -> AssociationThread[Map[HoldForm, dt] -> Map[f, dt]]

head[dt_] := "Head" -> AssociationThread[Map[HoldForm, dt] -> Map[Head, dt]]

test = {Pi, x[1], "text", 0, 3.0, 1/2, I, 1 + 0. I}

 <|head[test], col[test, RealValuedNumberQ], 
   col[test, RealValuedNumericQ]|> // Transpose // Dataset

enter image description here

Improve this answer
$\endgroup$
1
$\begingroup$

Surely the simplest way to do this is just to check whether the imaginary part is zero:

Im[z] == 0

This will return true if z is a real number and false if z is complex.

Improve this answer
$\endgroup$
1
  • 3
    $\begingroup$ Try it on the OP's first example Im["text"] == 0. You might want to consider Mark McClure's example, too. $\endgroup$
    – Michael E2
    Commented Apr 27, 2016 at 15:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.