Keep trailing zeros in precision for string format option g

[sk1d]

BPO	32790
Nosy	@mdickinson, @ericvsmith, @matrixise, @miss-islington
PRs	bpo-32790: Added info about alt format using # for 'g' in chart #6624 [3.8] bpo-32790: Add info about alt format using GH- for 'g' in chart (GH-6624) #16121 [3.7] bpo-32790: Add info about alt format using GH- for 'g' in chart (GH-6624) #16122

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2019-09-13.17:29:44.215>
created_at = <Date 2018-02-08.02:08:36.788>
labels = ['easy', 'type-bug', '3.8', '3.9', '3.7', 'docs']
title = 'Keep trailing zeros in precision for string format option g'
updated_at = <Date 2019-09-13.17:29:44.214>
user = 'https://bugs.python.org/sk1d'

bugs.python.org fields:

activity = <Date 2019-09-13.17:29:44.214>
actor = 'matrixise'
assignee = 'docs@python'
closed = True
closed_date = <Date 2019-09-13.17:29:44.215>
closer = 'matrixise'
components = ['Documentation']
creation = <Date 2018-02-08.02:08:36.788>
creator = 'sk1d'
dependencies = []
files = []
hgrepos = []
issue_num = 32790
keywords = ['patch', 'easy']
message_count = 8.0
messages = ['311813', '311816', '311831', '311847', '352376', '352378', '352379', '352380']
nosy_count = 6.0
nosy_names = ['mark.dickinson', 'eric.smith', 'docs@python', 'matrixise', 'miss-islington', 'sk1d']
pr_nums = ['6624', '16121', '16122']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue32790'
versions = ['Python 3.7', 'Python 3.8', 'Python 3.9']

[sk1d]

The documentation starts the the string format parameter 'g':

General format. For a given precision p >= 1, this rounds the number to **p significant digits** and then formats the result in either fixed-point format or in scientific notation, depending on its magnitude.

I think the behavior of format is inconsistent here:
>>> format(0.1949, '.2g')
returns '0.19' as expected but
>>> format(0.1950, '.2g')
returns '0.2' instead of '0.20'

This behavior for float is in my opinion the correct one here
>>> format(0.1950, '.2f')
returns '0.20'

[mdickinson]

The behaviour here is intentional, though the reasons for doing it this way are at least partly historical: it's the way that %g formatting works in C's *printf functions (see C99 7.19.6.1p8 for details), and as a direct result of that it's also the way that old-style %-based formatting works in Python. That behaviour then got transferred to the new-style .format-based formatting for consistency.

I don't think we can or should change the current behaviour here: there's a significant risk of breaking existing code.

However, note that C does offer an *alternate* mode for .g-style formatting, using the '#' character, and this is also available in Python's formatting, both %-based and format-based:

>>> "%.2g" % 0.1950
'0.2'
>>> "%#.2g" % 0.1950
'0.20'

and

>>> format(0.1950, '.2g')
'0.2'
>>> format(0.1950, '#.2g')
'0.20'

Does this meet your needs?

[sk1d]

This meet my needs.

Maybe the documentation could also add this information in the chart for 'g' here: https://docs.python.org/3.7/library/string.html#format-specification-mini-language as only into the running text. As I did not notice it.

As the text in the table for 'g' is already long, if you do not want to add all the same information again, add at least that trailing zeros will get removed.

[mdickinson]

Maybe the documentation could also add this information in the chart for 'g' here:

Agreed that that would make sense. All the information is *there* in the docs (the alternate format is described a bit above that table), but it's not very easy to extract the information.

And that table entry is misleading as it currently is, since it suggests that trailing zeros are removed unconditionally: "In both cases insignificant trailing zeros are removed from the significand, and the decimal point is also removed if there are no remaining digits following it."

Perhaps we could adopt something like the wording in the C standard, where it says:

Finally, unless the # flag is used, any trailing zeros are removed
from the fractional portion of the result and the decimal-point
character is removed if there is no fractional portion remaining.

Re-classifying as a documentation issue.

[matrixise]

New changeset d44542f by Stéphane Wirtel (bchhabra2490) in branch 'master':
bpo-32790: Add info about alt format using # for 'g' in chart (GH-6624)
d44542f

[miss-islington]

New changeset 77878ca by Miss Islington (bot) in branch '3.7':
bpo-32790: Add info about alt format using GH- for 'g' in chart (GH-6624)
77878ca

[miss-islington]

New changeset e6b14c0 by Miss Islington (bot) in branch '3.8':
bpo-32790: Add info about alt format using GH- for 'g' in chart (GH-6624)
e6b14c0

[matrixise]

Thank you for your PR and this issue, the PR is merged into master, 3.8 and 3.7.