Implement mpf.format #819

javierelpianista · 2024-07-19T22:28:42Z

Implemented the __format__ method for the _mpf type.

Currently, f, F, g, G, e, and E formats are implemented, following the same specifications as specified for regular floats here.
Only nearest rounding is implemented.

Partially resolves #337.

The method is only implemented for the _mpf type, and it follows the same conventions as in https://docs.python.org/3/library/string.html#formatstrings, but extended to mpf floats. Also added tests to this function.

skirpichev

Very good start!

Please take look on comments. Also, it seems the code is not fully covered by tests (see CI failure).

mpmath/tests/test_format.py

mpmath/libmp/libfmt.py

mpmath/libmp/__init__.py

mpmath/libmp/libfmt.py

….mpf formatting

…haracters after type specification

… test case for when both 0-padding and fill char are selected.

The 'z' feature was included in Python 3.11, so it makes no sense to test for it if we are using an older Python version.

mpmath/tests/test_format.py

mpmath/tests/test_str.py

mpmath/libmp/libmpf.py

mpmath/tests/test_str.py

* change imports * move helper up * spaces

* unident some code * kill some comments * move up _FLOAT_FORMAT_SPECIFICATION_MATCHER

skirpichev

LGTM, with one comment on round_digits(). Please address this.

I took liberty to do some stylistic adjustments. Please take look, lets be sure your are OK with this.

mpmath/libmp/libmpf.py

skirpichev

Ah, you should test also subnormal values for floats. Either in test_mpf_float() or in test_mpf_fmt_cpython().

mpf has no restrictions on exponent range, yet I think that output should here match Python's as well.

javierelpianista · 2024-07-22T18:28:54Z

Ah, you should test also subnormal values for floats. Either in test_mpf_float() or in test_mpf_fmt_cpython().

mpf has no restrictions on exponent range, yet I think that output should here match Python's as well.

If I understand you correctly, you mean mpf initialized from floats? Such as

num = 2e-315
print('{:22.15f}'.format(num))
print('{:22.15f}'.format(mpf(num)))
print('{:22.15g}'.format(num))
print('{:22.15g}'.format(mpf(num)))
print('{:22.15e}'.format(num))
print('{:22.15e}'.format(mpf(num)))

resulting in

     0.000000000000000
     0.000000000000000
 2.00000000190402e-315
 2.00000000190402e-315
2.000000001904024e-315
2.000000001904024e-315

But I would expect

num_str = '2e-315'
print('{:22.15f}'.format(mpf(num_str)))
print('{:22.15g}'.format(mpf(num_str)))
print('{:22.15e}'.format(mpf(num_str)))

to result into

     0.000000000000000
                2e-315
2.000000000000000e-315

because in this case the mpf is correctly initialized.

skirpichev · 2024-07-23T03:47:34Z

If I understand you correctly, you mean mpf initialized from floats?

Yes, as in other your compatibility tests. And with subnormal values, like in shown example.

But I would expect [...] because in this case the mpf is correctly initialized.

Nothing incorrect in using mpf(float) for default settings.

* Added this functionality to format_scientific and format_fixed functions * Added random tests for very small numbers to test_format.py

* Added 10 digits to the calls to to_digits_exp, this fixes wrong formatting for some subnormal floats. * Also removed code for no_neg_0 checking in format_scientific, since it was never called.

mpmath/tests/test_format.py

skirpichev · 2024-07-24T05:02:44Z

Formatting support must be documented, probably here: https://mpmath.readthedocs.io/en/latest/general.html#conversion-and-printing

But maybe we can do this later in a separate pr.

skirpichev

LGTM with few comments

mpmath/tests/test_format.py

skirpichev · 2024-07-24T06:42:42Z

mpmath/libmp/libmpf.py

+#                              String formatting                             #
+#----------------------------------------------------------------------------#
+
+blog2_10 = 3.3219280948873626


This constant should be moved up (near math_float_inf, I think) and reused in a couple of places:

$ git grep 3.321928 mpmath/libmp/libmpf.py mpmath/libmp/libmpf.py: return max(1, int(round(int(n)/3.3219280948873626)-1)) mpmath/libmp/libmpf.py: return max(1, int(round((int(n)+1)*3.3219280948873626))) mpmath/libmp/libmpf.py: prec = int(len(x.as_tuple()[1])*3.3219280948873626) mpmath/libmp/libmpf.py: blog2 = 3.3219280948873626

mpmath/libmp/libmpf.py

Co-authored-by: Sergey B Kirpichev <skirpichev@gmail.com>

skirpichev · 2024-07-24T16:38:58Z

@javierelpianista, thank you!

skirpichev · 2024-07-25T03:59:22Z

@javierelpianista, are you planing to work on mpc.__format__?

javierelpianista · 2024-07-25T12:44:08Z

@javierelpianista, are you planing to work on mpc.__format__?

I could give it a try in the future, but not right now.

javierelpianista added 4 commits July 19, 2024 17:56

Add __format__ function to the _mpf type.

8858240

The method is only implemented for the _mpf type, and it follows the same conventions as in https://docs.python.org/3/library/string.html#formatstrings, but extended to mpf floats. Also added tests to this function.

Fixed test_format.py calling for files that were not there

c34c56d

Additional fixes for test_format.py

8631449

Removed unnecesary imports from ctx_mp_python and libmp/__init__.py

bc3e71a

skirpichev reviewed Jul 20, 2024

View reviewed changes

javierelpianista added 11 commits July 21, 2024 16:24

Moved string formatting functions to libmpf.py

98b8e98

Made to_str to use the new round_digits function

048dba5

Made function test_mpf_fmt_cpython really test with mp.mpf numbers.

c3b5fce

Added an assortment of random tests that compare mp.fp.mpf with mp.mp…

5d42158

….mpf formatting

Corrected tests for nan and inf when formatting with the + option

2840f97

Improved read_format_spec to include 0- padding and to detect wrong c…

85765cf

…haracters after type specification

Added some tests for instances where the format specification is wrong.

118c860

Modified test_str.py. 0.65625 should round to 0.6562, not 0.6563.

8c5e683

Changed the read_format_spec function to use the regex from fractions.py

c182d09

Modified code to remove parts that were never executed. Also, added a…

679434f

… test case for when both 0-padding and fill char are selected.

Modified test_format.py to exclude future format features.

9757a1c

The 'z' feature was included in Python 3.11, so it makes no sense to test for it if we are using an older Python version.

skirpichev reviewed Jul 22, 2024

View reviewed changes

mpmath/tests/test_format.py Show resolved Hide resolved

mpmath/tests/test_format.py Outdated Show resolved Hide resolved

mpmath/tests/test_format.py Outdated Show resolved Hide resolved

mpmath/tests/test_format.py Outdated Show resolved Hide resolved

skirpichev reviewed Jul 22, 2024

View reviewed changes

mpmath/tests/test_str.py Outdated Show resolved Hide resolved

skirpichev reviewed Jul 22, 2024

View reviewed changes

mpmath/libmp/libmpf.py Outdated Show resolved Hide resolved

skirpichev reviewed Jul 22, 2024

View reviewed changes

mpmath/libmp/libmpf.py Outdated Show resolved Hide resolved

skirpichev reviewed Jul 22, 2024

View reviewed changes

skirpichev added 7 commits July 22, 2024 08:02

Apply suggestions from code review

b474e0e

Run isort

d0c0798

Slight refactoring for tests

50b9ea6

* change imports * move helper up * spaces

Minor reformatting for libmpf.py

0e5b458

* unident some code * kill some comments * move up _FLOAT_FORMAT_SPECIFICATION_MATCHER

And small cleanup in test_errors()

fc9d662

Use formatted string literals

86e76cd

+ fix fstrings for < 3.12

aeb6c56

skirpichev approved these changes Jul 22, 2024

View reviewed changes

mpmath/libmp/libmpf.py Outdated Show resolved Hide resolved

skirpichev reviewed Jul 22, 2024

View reviewed changes

mpmath/libmp/libmpf.py Show resolved Hide resolved

mpmath/libmp/libmpf.py Outdated Show resolved Hide resolved

Apply suggestions from code review

2e095ab

skirpichev requested changes Jul 22, 2024

View reviewed changes

javierelpianista added 5 commits July 23, 2024 11:55

Programmed the no_neg_0 option for mpf.__format__.

d8e99c4

* Added this functionality to format_scientific and format_fixed functions * Added random tests for very small numbers to test_format.py

Added format tests for mpf initialized from subnormal floats

bb90d12

Added tests to test_format.py that were only covered by random tests

256373a

Fix for wrong formatting in some subnormal floats

3fdc261

* Added 10 digits to the calls to to_digits_exp, this fixes wrong formatting for some subnormal floats. * Also removed code for no_neg_0 checking in format_scientific, since it was never called.

Moved the subnormal tests to test_mpf_float

8107b8f

skirpichev reviewed Jul 24, 2024

View reviewed changes

mpmath/tests/test_format.py Outdated Show resolved Hide resolved

Update mpmath/tests/test_format.py

76050d9

skirpichev approved these changes Jul 24, 2024

View reviewed changes

Update mpmath/libmp/libmpf.py

44813bb

Co-authored-by: Sergey B Kirpichev <skirpichev@gmail.com>

skirpichev merged commit cead3b2 into mpmath:master Jul 24, 2024

skirpichev added this to the 1.4 milestone May 9, 2025

Implement mpf.__format__ #819

Implement mpf.__format__ #819

Uh oh!

Conversation

javierelpianista commented Jul 19, 2024

Uh oh!

skirpichev left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

skirpichev left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

skirpichev left a comment

Choose a reason for hiding this comment

Uh oh!

javierelpianista commented Jul 22, 2024

Uh oh!

skirpichev commented Jul 23, 2024

Uh oh!

Uh oh!

skirpichev commented Jul 24, 2024

Uh oh!

skirpichev left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

skirpichev Jul 24, 2024

Choose a reason for hiding this comment

Uh oh!

Uh oh!

skirpichev commented Jul 24, 2024

Uh oh!

skirpichev commented Jul 25, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

javierelpianista commented Jul 25, 2024

Uh oh!

Uh oh!

Implement mpf.format #819

Implement mpf.format #819

skirpichev commented Jul 25, 2024 •

edited

Loading