Left bracket remains in format string result when '\' preceeds it

[DimitrisJim]

BPO	29104
Nosy	@vstinner, @ericvsmith, @serhiy-storchaka, @refi64, @DimitrisJim
PRs	bpo-29104: Fixed parsing backslashes in f-strings. #490 [3.6] bpo-29104: Fixed parsing backslashes in f-strings. (GH-490) #1812
Files	fstring_backslash.patch fstring_backslash_2.patch

^{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 = 'https://github.com/ericvsmith'
closed_at = <Date 2017-05-25.11:19:44.542>
created_at = <Date 2016-12-29.11:56:25.482>
labels = ['interpreter-core', 'type-bug', '3.7']
title = "Left bracket remains in format string result when '\\' preceeds it"
updated_at = <Date 2017-05-25.11:19:44.541>
user = 'https://github.com/DimitrisJim'

bugs.python.org fields:

activity = <Date 2017-05-25.11:19:44.541>
actor = 'serhiy.storchaka'
assignee = 'eric.smith'
closed = True
closed_date = <Date 2017-05-25.11:19:44.542>
closer = 'serhiy.storchaka'
components = ['Interpreter Core']
creation = <Date 2016-12-29.11:56:25.482>
creator = 'Jim Fasarakis-Hilliard'
dependencies = []
files = ['46080', '46127']
hgrepos = []
issue_num = 29104
keywords = ['patch']
message_count = 26.0
messages = ['284249', '284256', '284257', '284258', '284259', '284260', '284261', '284262', '284266', '284271', '284558', '284578', '284579', '284580', '285949', '288511', '288979', '289645', '289689', '289690', '289694', '290733', '293604', '293608', '294458', '294469']
nosy_count = 5.0
nosy_names = ['vstinner', 'eric.smith', 'serhiy.storchaka', 'refi64', 'Jim Fasarakis-Hilliard']
pr_nums = ['490', '1812']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue29104'
versions = ['Python 3.6', 'Python 3.7']

[DimitrisJim]

In short:

>>> f"\{10}"

yields:

"\\{10"

This is reproducible only when \ precedes the opening bracket, that is:

>>> f"\ {10}"

results in: "\\ 10"

[ericvsmith]

I'm not sure this counts as an error. The backslash means to treat the next character literally, which this does. And since \{ is not a valid escape sequence, it keeps both characters, exactly like:

>>> '\c'
'\\c'

Furthermore, since unknown escape sequences are deprecated, this usage is going to be an error in the future. You can see this now with -Werror:

$ python.exe  -Werror
Python 3.6.0b1+ (3.6:87de1f12c41c+, Sep 16 2016, 07:05:57) [MSC v.1900 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> f'\{10}'
DeprecationWarning: invalid escape sequence '\{'
>>> '\c'
DeprecationWarning: invalid escape sequence '\c'
>>>

Since this works as I expect it to now, and since it will become an error in the future, I don't think any change is warranted.

[ericvsmith]

But I admit that dropping the trailing } seems odd. I think this particular usage should be an error: no bare } allowed, similar to:

>>> f'10}'
  File "<stdin>", line 1
SyntaxError: f-string: single '}' is not allowed

[DimitrisJim]

I see, the original "complaint" about this behavior on stack overflow was made due to the discrepancy between the f-string the the format "equivalent": '\{}'.format(10)'.

[ericvsmith]

Do you have a link to the SO question?

[DimitrisJim]

Yes, should've attached in my previous message. See http://stackoverflow.com/questions/41330097/why-does-the-symbol-remain-when-f-10-is-evaluated-in-python-3-6

[serhiy-storchaka]

The problem is not that the trailing } is dropped, but that the starting { starts an f-string expression.

>>> f'\{2*5}'
'\\{10'

I expected either '\\10' as in '\{}'.format(2*5), or at least '\\{2*5}'.

There is other f-string parsing error:

>>> f'\\N{2*5}'
'\\N{2*5}'

I expected '\\N10'. '\\N' doesn't start a unicode name escape. This is a legitimate expression.

[ericvsmith]

Yes:

>>> f'\{2*5}'
'\\{10'

I agree '\\10' would make sense.

[ericvsmith]

This problem was no doubt introduced after 3.6a1 when I changed the parsing to disallow backslashes inside {}.

[serhiy-storchaka]

Proposed patch fixes parsing backslashes in f-strings.

P.S. Definitely we should find other name for this term. I hate misleading "f-string".

[vstinner]

The PEP-498 should be updated to define the expected behaviour of f'\{5}':
https://www.python.org/dev/peps/pep-0498/#escape-sequences

+ self.assertEqual(f'\{6*7}', '\\42')
+ self.assertEqual(f'\\{6*7}', '\\42')
+ self.assertEqual(fr'\{6*7}', '\\42')
+
+ AMPERSAND = 123
+ self.assertEqual(f'\N{AMPERSAND}', '&')
+ self.assertEqual(f'\\N{AMPERSAND}', '\\N123')
+ self.assertEqual(fr'\N{AMPERSAND}', '\\N123')
+ self.assertEqual(f'\\\N{AMPERSAND}', '\\&')

I'm not sure that I like this behaviour.

f'\\N{AMPERSAND}': reading a local variable looks like a typo or a security vulnerability, rather than a nice feature.

IMHO if you want an anti-slash (\) in the output string, it *must* be written "\\" since we *do* interpret other escape sequences:

• f'\n' returns '\n', the newline character U+000A
• f'\N{Snowman}' returns the nice snowman character U+2603 (☃)
• etc.

What is the issue with having to write "\\{10}" to get "\\10" string? It's less error prone.