homepage

ConfigParser.add_section() raises DuplicateSectionError if add_section()
is called with the name of a section that is already present. How ever,
if a section is present multiple times in a file, readfp() does not
raise any exception. Instead, the sections are "combined" with values in
latter sections overwriting previous values (if any). This should be
documented in readfp(). I tried few alternatives but the sentences don't
look right. I am creating the issue in the hope that some one else may
give it a try while I continue to come up with a doc patch.

Should this 'undesirable' behavior be documented? My thoughts are
a) All the key,value pairs in the named section be retrived from entire
file. Just appending the dict.
Otherwise, 
b) An Exception can be thrown if it encounters an invalid Configfile
having multiple same sections which can cause problem, but I cannot
readily think of one such case.
What do others think?
Thanks,
Senthil

> Should this 'undesirable' behavior be documented? My thoughts are
Document should definitely reflect the code whether the behaviour is
desirable or not.
> a) All the key,value pairs in the named section be retrived from entire
> file. Just appending the dict.
Isn't this what is happening now unless I am not following you?
> Otherwise,
> b) An Exception can be thrown if it encounters an invalid Configfile
> having multiple same sections which can cause problem, but I cannot
> readily think of one such case.
I have no problem with throwing an exception in case of duplicate
sections but I don't think that can be done by default (it may break
some programs).

The attached patch is a proof of concept for throwing an exception.
As discussed on the mailing list [1], it has some shortcomings that
should be addressed before it is merged.
[1]: http://mail.python.org/pipermail/python-dev/2009- 

> The attached patch is a proof of concept for throwing an exception.
If it is not too much of a problem, can you upload the patch to
http://codereview.appspot.com? Reviewing there is simpler. Also, you
will have to include some test cases in the patch.
> As discussed on the mailing list [1], it has some shortcomings that
> should be addressed before it is merged.
I think that the builder interface already has correct behaviour in
that duplicate sections are not allowed. So you can have
'unique_sections' control only parser interface and can set it to
False by default (to be compatible with current behaviour).
> [1]: http://mail.python.org/pipermail/python-dev/2009-
Correct discussion link:
http://mail.python.org/pipermail/python-dev/2009-February/085838.html 

I am using configparser after combining 3 or so ini files, I would like
to merge similar sections rather than throwing an exception. Right now
I have to make sure the ini files all have unique sections, but I would
like to able to use same section names, then combine the files and have
the sections merged. Even merging the variables within sections would
be nice.
Throwing an exception, imho, is no the right direction. Even in same
name sections are in the same file, the variables where introduced for a
reason.
Or maybe a trigger option ie action=merge or action=flag.

I personally would keep the current behaviour so that:
- creating a duplicate section programatically invokes an error
- reading a file which has duplicate sections causes an implicit merge without any exceptions
This is quite understandable, more so when you consider how ConfigParser merges configuration from multiple sources as well.
Brett, the only thing I guess we could do here is to adjust the documentation so it's more clear on the default behaviour.

@Łukasz: could you provide a patch that clarifies the default behaviour.

This bug is superseded by #9452 where the patch includes a `strict=` argument to __init__ which enables checking for section duplicates when reading from a single source.