This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author sunfinite
Recipients Arfrever, adlaiff6, amaury.forgeotdarc, belopolsky, benjamin.peterson, christian.heimes, eric.araujo, eric.smith, eric.snow, georg.brandl, giampaolo.rodola, gjb1002, jackdied, lemburg, rhettinger, salty-horse, santoso.wijaya, sunfinite, terry.reedy, vstinner
Date 2013-11-07.12:40:48
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <>
New patch for 3.4 adds the following:

1. _fields
2. _replace()
3. _asdict()
4. eval(repr(s)) == s

Now the issues:

1. _asdict() returns a normal dictionary. I don't know if this is what
   is required.

2. Both _asdict() and _replace() assume that unnamed visible fields are
   at the end of the visible sequence.  A comment at the beginning says
   they are allowed for indices < n_visible_fields.
   Is there another way to map members to values? Because tp->members
   is (visible named fields + invisible named fields) whereas values
   array is (visible named fields + visible unnamed fields + invisible
   named fields)

3. The mismatch mentioned above is present in the current
   implementation of repr:

   In os.stat_result, the last three visible fields are unnamed
   (i.e integer a_time, m_time and c_time). However they are present 
   in the repr output with keys which are the first three keys from the
   invisible part(float a_time, m_time and c_time).
   Was this intentional?

   Also, the above logic causes duplicate keys when invisible fields 
   are included in the repr output as per issue11629.
   In my patch for that issue, i put invisible fields under the 'dict'
   keyword argument. This output format utilises code already present 
   in structseq_new and makes eval work as expected when invisible 
   fields are present in repr. Also, it sidesteps the question of
   duplicated keys because they are present inside a dict.

4. Raymond stated that _fields should contain only the visible 
   positional fields. So _fields of os.stat_result contains only 7
   fields because of the three unnamed fields. Is this the expected 

5. Is there a way to declare a member function in C that accepts only 
   keyword arguments(like _replace())? I could not find one.

This is my first real C patch. So some of the issues might just be a
lack of understanding. Apologies.
Date User Action Args
2013-11-07 12:40:50sunfinitesetrecipients: + sunfinite, lemburg, georg.brandl, rhettinger, terry.reedy, amaury.forgeotdarc, gjb1002, belopolsky, salty-horse, vstinner, jackdied, eric.smith, giampaolo.rodola, christian.heimes, benjamin.peterson, adlaiff6, eric.araujo, Arfrever, santoso.wijaya, eric.snow
2013-11-07 12:40:50sunfinitesetmessageid: <>
2013-11-07 12:40:50sunfinitelinkissue1820 messages
2013-11-07 12:40:49sunfinitecreate