With Python 2.7 the ConfigParser was enriched with the allow_no_value option which resulted in a much more usable configparser for MySQL configs.

It can now parse configs like this:
[mysqld]
log_bin
innodb_file_per_table
innodb_io_capacity=800

However it can't parse this config:
[mysqld]
log_bin
innodb_file_per_table
innodb_io_capacity=800
replicate-do-db="test1"
replicate-do-db="test2"

A comma separated config might have been better, but that's not the case with many MySQL config (and probably doesn't even work for this option).

From http://dev.mysql.com/doc/refman/5.6/en/replication-options-slave.html#option_mysqld_replicate-do-db
"To specify more than one database, use this option multiple times, once for each database"


The dict/orderedDict which is used by configparser doesn't easlily allow to store a config like this.

The MySQL config file is used as an example in the documentation:
https://docs.python.org/3/library/configparser.html#customizing-parser-behaviour

This could be solved by having a list of values if the key exists more than once.

Python 3 improves this a bit by giving a DuplicateOptionError by default.

The data structure you are asking for bears a resemblance to the data structure used by the email package to record message headers.  Email uses a list with a dict-like API cobbled together on top.  The difference with your suggestion is that email also needs to record the order of the headers, so just having a list-of-values for each key isn't adequate.  Any bets on whether or not we eventually run across a use of .ini format where the exact order of the duplicated keys within the whole list of keys matters? :)

With two use cases that could be served by the same data structure in the stdlib, I wonder if it is worth actually building the abstract data type.  Probably not, but I thought I'd raise the possibility anyway :)

See Issue 12662 for a related earlier proposal. (Although the multimap proposal here sounds better than concatenating the values as proposed there.)