Python list of floats ends up with a ':' when converting to a string

Question

I have a list of floats read in from a text file. After some data processing I write the list to a file using:

for val in flist:
    sa = '{0}'.format(val)
    fout.write(sa)

For specific input files, the output file will have a ':' in the string. I have run debug and stopped the script at the point of failure. The value should be 58710000.0

[Dbg]>>> print val[464]
   5870:000.0
[Dbg]>>> fa = val[464]
[Dbg]>>> print fa
  5870:000.0
[Dbg]>>> 
[Dbg]>>> fa = fa + 1
[Dbg]>>> print fa
   58710001.0
[Dbg]>>> fa = fa - 1
[Dbg]>>> print fa
   5870:000.0

This happens only for certain files and floats

Any suggestions?

do me a favour and give the output of `print repr(val[464])` — Jakob Bowyer, Jan 09 '13 at 00:32
Also https://stackoverflow.com/questions/45871269/why-can-10f-decimalu-emit-a-string-with-a-literal-colon ... it appears to be platform specific. — Yann Vernier, Sep 11 '17 at 10:23

pashute · Answer 1 · 2017-09-11T09:54:17.703

4

It's a bug in Python 2.7.3 or perhaps earlier, with certain environments.

User @ecatmur pointed out in a different post with a similar question, that '9' + 1 = ':' in ASCII

This has been Fixed in later versions of Python.
Specifically, the problem disappeared in Python 2.7.5 so the issue has been fixed.

See Gord Thompson's accepted answer on:

Python inserts a colon in a decimal number from Access via PyoDBc

Similar questions were closed or not answered:

edited Sep 11 '17 at 09:54

answered Sep 11 '17 at 09:48

pashute

3,965
3
38
65

2

Would be nice with a reference to the Python change that did fix something. AFAICT this is a C library bug, not internal to CPython. – Yann Vernier Sep 11 '17 at 10:48

score 0 · Answer 2 · answered Jan 09 '13 at 00:39

0

Your values are probably not actually floats. Anyhow, sa = '{0}'.format(val) is silly; you might as well do just sa = str(val). But if you do sa = '%f' % (val,) instead, you will get an exception when val is not a float.

answered Jan 09 '13 at 00:39

Phil Frost

3,668
21
29

1

You can still use the newer string formatting, just do `sa = '{0:f}'.format(val)` – Gavin H Jan 09 '13 at 00:50

Yann Vernier · Answer 3 · 2017-09-11T10:52:15.577

This appears to be highly platform specific. The repr for float calls PyOS_double_to_string which ends up calling PyOS_snprintf, which wraps snprintf with some code to make that function more consistent across platforms. It appears on some version of AIX in particular, snprintf can produce 0: instead of 10.

Could you share your sys.platform and sys.version values? (The version header printed when you start an interactive python interpreter should do.)

Relevant source files: Python/mysnprintf.c for PyOS_snprintf, Objects/floatobject.c for float_repr, Python/pystrtod.c for PyOS_string_to_double.

Python list of floats ends up with a ':' when converting to a string

3 Answers3

Linked