fix docstring printing of constants

[rws]

Other constants give more or less specific help but

sage: log2?

Type:           Expression
String form:    log2
Length:         0
File:           /home/ralf/sage/local/lib/python2.7/site-packages/sage/symbolic/expression.so
Docstring:
   Nearly all expressions are created by calling
   new_Expression_from_*, but we need to make sure this at least does

This confused me because I first thought I had found the Sage binary log function.

[kcrisman]

That's interesting, because ​here exactly this constant is used to show how awesome the interactive help is! So something is messed up.

[deinst]

This appears to be broken for most constants declared in sage/symbolic/constants.py with the exception of e.

I would expect that Expression would have a custom _sage_doc function that would reroute the help request to the appropriate class, but no such function seems to exist, or to ever have existed. If someone indicates how this should work, I'll be glad to try to fix it.

On a not entirely unrelated note: The format of the help has changed somewhat from what is written in the tutorial.

[kcrisman]

On a not entirely unrelated note: The format of the help has changed somewhat from what is written in the tutorial.

Can you be more specific?

[deinst]

Replying to kcrisman:

On a not entirely unrelated note: The format of the help has changed somewhat from what is written in the tutorial.

Can you be more specific?

I thought I had answered this a while ago. Obviously I got distracted.

The tutorial help for tan (gotten locally from file:///Users/davideinstein/projects/sage/src/doc/output/html/en/tutorial/tour_help.html looks like:

sage: tan?
Type:        <class 'sage.calculus.calculus.Function_tan'>
Definition:  tan( [noargspec] )
Docstring:

    The tangent function

    EXAMPLES:
        sage: tan(pi)
        0
        sage: tan(3.1415)
        -0.0000926535900581913
        sage: tan(3.1415/4)
        0.999953674278156
        sage: tan(pi/4)
        1
        sage: tan(1/2)
        tan(1/2)
        sage: RR(tan(1/2))
        0.546302489843790

If I run tan? in sage 6.7 I get

Type:           Function_tan
String form:    tan
File:           ~/projects/sage/local/lib/python2.7/site-packages/sage/functions/trig.py
Docstring:
   The tangent function.

   EXAMPLES:

      sage: tan(pi)
      0
      sage: tan(3.1415)
      -0.0000926535900581913
      sage: tan(3.1415/4)
      0.999953674278156
      sage: tan(pi/4)
      1
      sage: tan(1/2)
      tan(1/2)
      sage: RR(tan(1/2))
      0.546302489843790

   We can prevent evaluation using the "hold" parameter:

      sage: tan(pi/4,hold=True)
      tan(1/4*pi)

   To then evaluate again, we currently must use Maxima via
   "sage.symbolic.expression.Expression.simplify()":

      sage: a = tan(pi/4,hold=True); a.simplify()
      1

   TESTS:

      sage: conjugate(tan(x))
      tan(conjugate(x))
      sage: tan(complex(1,1))     # rel tol 1e-15
      (0.2717525853195118+1.0839233273386946j)

Init docstring:
   The tangent function.

   EXAMPLES:

      sage: tan(pi)
      0
      sage: tan(3.1415)
      -0.0000926535900581913
      sage: tan(3.1415/4)
      0.999953674278156
      sage: tan(pi/4)
      1
      sage: tan(1/2)
      tan(1/2)
      sage: RR(tan(1/2))
      0.546302489843790

   We can prevent evaluation using the "hold" parameter:

      sage: tan(pi/4,hold=True)
      tan(1/4*pi)

   To then evaluate again, we currently must use Maxima via
   "sage.symbolic.expression.Expression.simplify()":

      sage: a = tan(pi/4,hold=True); a.simplify()
      1

   TESTS:

      sage: conjugate(tan(x))
      tan(conjugate(x))
      sage: tan(complex(1,1))     # rel tol 1e-15
      (0.2717525853195118+1.0839233273386946j)
Call docstring:
   Wrapper around "BuiltinFunction.__call__()" which converts Python
   >>``<<int``s which are returned by Ginac to Sage Integers.

   This is needed to fix http://trac.sagemath.org/10133, where Ginac
   evaluates "sin(0)" to the Python int "0":

      sage: from sage.symbolic.function import BuiltinFunction
      sage: out = BuiltinFunction.__call__(sin, 0)
      sage: out, parent(out)
      (0, <type 'int'>)

   With this wrapper we have:

      sage: out = sin(0)
      sage: out, parent(out)
      (0, Integer Ring)

   However, if all inputs are Python types, we do not convert:

      sage: out = sin(int(0))
      sage: (out, parent(out))
      (0, <type 'int'>)
      sage: out = arctan2(int(0), float(1))
      sage: (out, parent(out))
      (0, <type 'int'>)
      sage: out = arctan2(int(0), RR(1))
      sage: (out, parent(out))
      (0, Integer Ring)

The Type: is shorter. The module information is now in the new File: line. The Definition: has been kidnapped by aliens, possibly replaced by the less informative String Form:. The Init Docstring and Call Docstring have been added.

[rws]

Replying to deinst:

This appears to be broken for most constants declared in sage/symbolic/constants.py with the exception of e.

This works because e is not a PyObject embedded in an Expression like pi (where pi.pyobject() returns that object). e has its own class, and so its own docstring. e.is_constant() returns True because is_constant() just checks that there are no variables present.

I would expect that Expression would have a custom _sage_doc function that would reroute the help request to the appropriate class, but no such function seems to exist, or to ever have existed. If someone indicates how this should work, I'll be glad to try to fix it.

Well, _sage_doc() apparently(?) has nothing to do with what IPython prints when appending ? to an object (same as giving the %pdoc magic). This is some other help mechanism.

[rws]

Okay, there is now sage.docs.instancedoc. This will work.

[rws]

The uninformative docstring from the Expression class that follows the constant docstring will be improved in another ticket.

Please review.

---

New commits:

​d080e9e

18077: fix docstring printing of constants

[jdemeyer]

What does the following sentence mean:

Instances of constants are usually embedded as ``PyObject`` in
a symbolic expression.

[jdemeyer]

Can you move the doctest to the _instancedoc_ method? It seems like it would best fit there. I would also want to see a doctest for the __doc__ of an Expression which is not a constant, something like (sin(x) + 1).__doc__

[jdemeyer]

I'm not convinced by the except AttributeError. It seems that every object has a __doc__, but it could be empty. So instead of the AttributeError, it would be better to do something like:

doc = self.pyobject().__doc__
if doc:
    return doc
return type(self).__doc__   # Isn't self.__doc__ an infinite recursion?

[rws]

Replying to jdemeyer:

What does the following sentence mean:

Instances of constants are usually embedded as ``PyObject`` in
a symbolic expression.

Does this make more sense? Instances of constants are usually embedded in a symbolic expression.

[rws]

I'll also move the orphaned docstring for I to the file where it is defined, because at the moment the reference manual associates it with the wrong object.

[rws]

Note that I also improved docstrings for the class, init, and call slots. Suggestions welcome.

---

New commits:

​6103879

18077: fix docstring printing of constants

[jdemeyer]

Replying to rws:

I'll also move the orphaned docstring for I to the file where it is defined, because at the moment the reference manual associates it with the wrong object.

You can just remove that docstring. For some reason, it is an identical copy of a docstring in init_pynac_I().

[rws]

Improved branch uploaded.

[git]

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

​281eb93

remove duplicate docstring

[jdemeyer]

Please rebase to latest master.

[jdemeyer]

I'm also wondering whether this could help with #26492 and/or #26297