modernize sage.rings.real_mpfr

[vdelecroix]

The current sage.rings.real_mpfr deals with arbitrary precision floating-point numbers (using mpfr). This is the default "real field" in many places in Sage even though it is often wiser to use intervals or balls. We perform some cosmetic changes to sage.rings.real_mpfr to emphasize the fact that the module deals with floating-point approximations of real numbers and not genuine reals. These modifications touch a lot of files and involve deprecation of widely used names.

See also the introduction of a genuine real field class in #24456 for another motivation.

step 1

• #24511: move the class factory create_RealField in real_field.py
• #24523: change string representation
• #24524:
  • RealField -> RealFloatingPointField_factory
  • RealField_class -> RealFloatingPointField
  • RealNumber -> RealFloatingPoint
• #24525: deprecate is_RealNumber(x) and is_RealField(x) in favour of isinstance(x, RealFloatingPoint) and isinstance(x, RealFloatingPointField)

step 2

• Deprecate RR and Reals and use RFF (or RFPF?) as the standard name for RealFloatingPointField(53)?
• more documentation.

---

for complex numbers, see #24489.

[rws]

In real_mpfr.pyx do you mean to rename cpdef RealField(int prec=53,...) or cdef class RealField_class?

[rws]

This commit changes 164 files. Preliminary testing passes in rings and src/doc. There might be subtle errors introduced in the documentation. Let's try the patchbots.

---

New commits:

​62d2d21	24456: class for the field of real numbers
​71f9fd0	24457: bulk rename RealField --> RealFloatingPointField

[rws]

Ah I forgot the string representation part.

[rws]

Sorry, I was pretty sure I excluded real_field.py.

[rws]

After changing the string representation as well: 223 files changed, 1493 insertions(+), 1493 deletions(-). I don't have hopes this gets through, except maybe in a beta of its own.

---

New commits:

​62d2d21	24456: class for the field of real numbers
​8bdf66b	24457: rename sage.rings.real_mpfr.RealField

[jdemeyer]

Replying to vdelecroix:

• change the name of RealField/RealField_class to RealFloatingPointField/RealFloatingPointField_class

Do we still need the old RealField given that it's no longer used after #24456?

[rws]

New commits:

​3c7f16e

24457: rename sage.rings.real_mpfr.RealField

[vdelecroix]

Replying to jdemeyer:

Replying to vdelecroix:

• change the name of RealField/RealField_class to RealFloatingPointField/RealFloatingPointField_class

Do we still need the old RealField given that it's no longer used after #24456?

It is. The old real field is the field of floating point numbers that is intensively used (and renamed in this ticket). The new real field is the field of real numbers (the real one).

[jdemeyer]

Replying to vdelecroix:

The old real field is the field of floating point numbers that is intensively used (and renamed in this ticket).

I was talking about the RealField() function. Is that still used?

If so, I would rather rename it to create_RealFloatingPointField.

[vdelecroix]

Replying to jdemeyer:

Replying to vdelecroix:

The old real field is the field of floating point numbers that is intensively used (and renamed in this ticket).

I was talking about the RealField() function. Is that still used?

If so, I would rather rename it to create_RealFloatingPointField.

I think that we should get rid of this function and the hand made cache. The caching would better be handled by UniqueRepresentation by appropriately writing __classcall__/__init__. This class is very old.

[vdelecroix]

Replying to vdelecroix:

Replying to jdemeyer:

Replying to vdelecroix:

The old real field is the field of floating point numbers that is intensively used (and renamed in this ticket).

I was talking about the RealField() function. Is that still used?

If so, I would rather rename it to create_RealFloatingPointField.

I think that we should get rid of this function and the hand made cache. The caching would better be handled by UniqueRepresentation by appropriately writing __classcall__/__init__. This class is very old.

But this is rather disjoint from this ticket.

[vdelecroix]

Replying to jdemeyer:

Replying to vdelecroix:

The old real field is the field of floating point numbers that is intensively used (and renamed in this ticket).

I was talking about the RealField() function. Is that still used?

If so, I would rather rename it to create_RealFloatingPointField.

For the sake of this ticket +1.

[rws]

Replying to vdelecroix:

Replying to jdemeyer:

If so, I would rather rename it to create_RealFloatingPointField.

For the sake of this ticket +1.

You mean "in this ticket"?

[vdelecroix]

Replying to rws:

Replying to vdelecroix:

Replying to jdemeyer:

If so, I would rather rename it to create_RealFloatingPointField.

For the sake of this ticket +1.

You mean "in this ticket"?

Sorry for my english. Yes, this is what I meant.

[rws]

Replying to jdemeyer:

If so, I would rather rename it to create_RealFloatingPointField.

Ah I see, in rings/all.py: import create_RealFloatingPointField as RealFloatingPointField.

[git]

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

​9b2a5c2

24457: missing changes in interpreters

[tmonteil]

Some quick comments:

• why renaming RealField and not RR at the same time ?
• i am not sure that completely changing the behaviour of RealField without any kind of deprecation period is wise, how will the user's code survive to this change ?
• do you really plan to change a behaviour that implies modifying 164 files without even an email to sage-devel ?
• Regarding the naming, i would personally prefer Floating-point over Floating Point (as it is usually), moreover, it has the advantage that the shortcut will be RFF (not RFPF), consistently with other representations of real numbers.
• where is the doc ?

[rws]

It's in the experimental phase. I will need an IDE with refactoring capabilities to do this reliably and quickly, or develop scripts. Also, I think it will need its own beta.

[vdelecroix]

Replying to tmonteil:

Some quick comments:

• why renaming RealField and not RR at the same time ?

why not. But it will

• be a much more delicate backward incompatibility.
• make the ticket even bigger

The 2 steps transition seems simpler to me.

• i am not sure that completely changing the behaviour of RealField without any kind of deprecation period is wise, how will the user's code survive to this change ?

There must be deprecation about the name changes (which is not the case in the current branch). RealField shoud still point to the FloatingPointField (with deprecation) and RR should still point to RFF (with deprecation).

I don't think there is a need for deprecation about the change of printing.

• do you really plan to change a behaviour that implies modifying 164 files without even an email to sage-devel ?
• Regarding the naming, i would personally prefer Floating-point over Floating Point (as it is usually), moreover, it has the advantage that the shortcut will be RFF (not RFPF), consistently with other representations of real numbers.

+1 for Floating-point.

• where is the doc ?

[rws]

I'm confused. You deleted the description item that renames RealField/RealField_class in comment:29. Does that mean those refactorings are not planned in this ticket? On the other hand you write isinstance(x, RealFloatingPointField) so the new name is used?

As to automating this and other big refactorings in Python Sage I just learned that 1. Pycharm refactoring apparently cannot be scripted (not to be expected anyway in an IDE), and 2. the rope tool is not up to the task.

[rws]

The deprecation of is_RealNumber(x)/is_RealField(x) in favor of isinstance can be done independently of the renaming. Separate ticket to come.

[vdelecroix]

Replying to rws:

The deprecation of is_RealNumber(x)/is_RealField(x) in favor of isinstance can be done independently of the renaming. Separate ticket to come.

Good idea!

[rws]

Note that is_RealField(X) then becomes isinstance(X, RealField_class).

[vdelecroix]

Let us do that incrementally (ie each subtask in ticket). It will simplify review and merging. Though I would feel much more confortable if the whole task would be actually closed within the next stable release.

[vdelecroix]

Replying to vdelecroix:

Replying to rws:

The deprecation of is_RealNumber(x)/is_RealField(x) in favor of isinstance can be done independently of the renaming. Separate ticket to come.

Good idea!

Though it would be better to first change the names:

• RealField -> RealFloatingPointField_factory
• RealField_class -> RealFloatingPointField
• RealNumber -> RealFloatingPoint

to avoid passing twice on the same lines.

[rws]

Rebased on the remaining dependency.

[rws]

OOps wrong ticket.

[vdelecroix]

update milestone 8.3 -> 8.4