Reading and Writing Config Files

ConfigObj 4 Introduction and Reference

Authors: Michael Foord
Nicola Larosa
Version: ConfigObj 4.7.2
Date: 2010/02/27
Homepage:ConfigObj Homepage
PyPI Entry:ConfigObj on PyPI
Development:Google Code Homepage
License:BSD License
Support:Mailing List

ConfigObj Manual


The best introduction to working with ConfigObj, including the powerful configuration validation system, is the article:

1   Introduction

ConfigObj is a simple but powerful config file reader and writer: an ini file round tripper. Its main feature is that it is very easy to use, with a straightforward programmer's interface and a simple syntax for config files. It has lots of other features though :

For support and bug reports please use the ConfigObj Mailing List or the issue tracker on the Google Code Homepage.

2   Downloading

The current version is 4.7.2, dated 27th February 2010. ConfigObj 4 is stable and mature. We still expect to pick up a few bugs along the way though [1].

You can get ConfigObj in the following ways :

2.1   Installing

ConfigObj has a source distribution on PyPI. If you unzip the archive you can install it with: install

Alternatively, you can install with easy install or pip:

easy_install configobj

2.2   Documentation also contains this document.

2.3   Development Version

It is sometimes possible to get the latest development version of ConfigObj from the Subversion Repository maintained on the Google Code Homepage.

3   ConfigObj in the Real World

ConfigObj is widely used. Projects using it include:

4   Getting Started

The outstanding feature of using ConfigObj is simplicity. Most functions can be performed with single line commands.

4.1   Reading a Config File

The normal way to read a config file, is to give ConfigObj the filename :

from configobj import ConfigObj
config = ConfigObj(filename)

You can also pass the config file in as a list of lines, or a StringIO instance, so it doesn't matter where your config data comes from.

You can then access members of your config file as a dictionary. Subsections will also be dictionaries.

from configobj import ConfigObj
config = ConfigObj(filename)
value1 = config['keyword1']
value2 = config['keyword2']
section1 = config['section1']
value3 = section1['keyword3']
value4 = section1['keyword4']
# you could also write
value3 = config['section1']['keyword3']
value4 = config['section1']['keyword4']

4.2   Writing a Config File

Creating a new config file is just as easy as reading one. You can specify a filename when you create the ConfigObj, or do it later [2].

If you don't set a filename, then the write method will return a list of lines instead of writing to file. See the write method for more details.

Here we show creating an empty ConfigObj, setting a filename and some values, and then writing to file :

from configobj import ConfigObj
config = ConfigObj()
config.filename = filename
config['keyword1'] = value1
config['keyword2'] = value2
config['section1'] = {}
config['section1']['keyword3'] = value3
config['section1']['keyword4'] = value4
section2 = {
    'keyword5': value5,
    'keyword6': value6,
    'sub-section': {
        'keyword7': value7
config['section2'] = section2
config['section3'] = {}
config['section3']['keyword 8'] = [value8, value9, value10]
config['section3']['keyword 9'] = [value11, value12, value13]


Keywords and section names can only be strings [3]. Attempting to set anything else will raise a ValueError.

See String Interpolation and List Values for an important note on using lists in combination with String Interpolation.

4.3   Config Files

The config files that ConfigObj will read and write are based on the 'INI' format. This means it will read and write files created for ConfigParser [4].

Keywords and values are separated by an '=', and section markers are between square brackets. Keywords, values, and section names can be surrounded by single or double quotes. Indentation is not significant, but can be preserved.

Subsections are indicated by repeating the square brackets in the section marker. You nest levels by using more brackets.

You can have list values by separating items with a comma, and values spanning multiple lines by using triple quotes (single or double).

For full details on all these see the config file format. Here's an example to illustrate:

# This is the 'initial_comment'
# Which may be several lines
keyword1 = value1
'keyword 2' = 'value 2'

[ "section 1" ]
# This comment goes with keyword 3
keyword 3 = value 3
'keyword 4' = value4, value 5, 'value 6'

    [[ sub-section ]]    # an inline comment
    # sub-section is inside "section 1"
    'keyword 5' = 'value 7'
    'keyword 6' = '''A multiline value,
that spans more than one line :-)
The line breaks are included in the value.'''

        [[[ sub-sub-section ]]]
        # sub-sub-section is *in* 'sub-section'
        # which is in 'section 1'
        'keyword 7' = 'value 8'

[section 2]    # an inline comment
keyword8 = "value 9"
keyword9 = value10     # an inline comment
# The 'final_comment'
# Which also may be several lines

5   ConfigObj specifications

config = ConfigObj(infile=None, options=None, configspec=None, encoding=None,
                   interpolation=True, raise_errors=False, list_values=True,
                   create_empty=False, file_error=False, stringify=True,
                   indent_type=None, default_encoding=None, unrepr=False,
                   write_empty_values=False, _inspec=False)

Many of the keyword arguments are available as attributes after the config file has been parsed.


New in ConfigObj 4.7.0: Instantiating ConfigObj with an options dictionary is now deprecated. To modify code that used to do this simply unpack the dictionary in the constructor call:

config = ConfigObj(filename, **options)

ConfigObj takes the following arguments (with the default values shown) :

5.1   Methods

The ConfigObj is a subclass of an object called Section, which is itself a subclass of dict, the builtin dictionary type. This means it also has all the normal dictionary methods.

In addition, the following Section Methods may be useful :

  • 'restore_default'
  • 'restore_defaults'
  • 'walk'
  • 'merge'
  • 'dict'
  • 'as_bool'
  • 'as_float'
  • 'as_int'
  • 'as_list'

Read about Sections for details of all the methods.


The merge method of sections is a recursive update.

You can use this to merge sections, or even whole ConfigObjs, into each other.

You would typically use this to create a default ConfigObj and then merge in user settings. This way users only need to specify values that are different from the default. You can use configspecs and validation to achieve the same thing of course.

The public methods available on ConfigObj are :

  • 'write'
  • 'validate'
  • 'reset'
  • 'reload'

5.1.1   write


This method writes the current ConfigObj and takes a single, optional argument [9].

If you pass in a file like object to the write method, the config file will be written to this. (The only method of this object that is used is its write method, so a StringIO instance, or any other file like object will work.)

Otherwise, the behaviour of this method depends on the filename attribute of the ConfigObj.

ConfigObj will write the configuration to the file specified.
write returns a list of lines. (Not '\n' terminated)

First the 'initial_comment' is written, then the config file, followed by the 'final_comment'. Comment lines and inline comments are written with each key/value.

5.1.2   validate

validate(validator, preserve_errors=False, copy=False)
# filename is the config file
# filename2 is the configspec
# (which could also be hardcoded into your program)
config = ConfigObj(filename, configspec=filename2)
from validate import Validator
val = Validator()
test = config.validate(val)
if test == True:
    print 'Succeeded.'

The validate method uses the validate module to do the validation.

This method validates the ConfigObj against the configspec. By doing type conversion as well it can abstract away the config file altogether and present the config data to your application (in the types it expects it to be).

If the configspec attribute of the ConfigObj is None, it raises a ValueError.

If the stringify attribute is set, this process will convert values to the type defined in the configspec.

The validate method uses checks specified in the configspec and defined in the Validator object. It is very easy to extend.

The configspec looks like the config file, but instead of the value, you specify the check (and any default value). See the validation section for details.


The system of configspecs can seem confusing at first, but is actually quite simple and powerful. The best guide to them is this article on ConfigObj:

The copy parameter fills in missing values from the configspec (default values), without marking the values as defaults. It also causes comments to be copied from the configspec into the config file. This allows you to use a configspec to create default config files. (Normally default values aren't written out by the write method.)

As of ConfigObj 4.3.0 you can also pass in a ConfigObj instance as your configspec. This is especially useful if you need to specify the encoding of your configspec file. When you read your configspec file, you must specify list_values=False. If you need to support hashes inside the configspec values then you must also pass in _inspec=True. This is because configspec files actually use a different syntax to config files and inline comment support must be switched off to correctly read configspec files with hashes in the values.

from configobj import ConfigObj
configspec = ConfigObj(configspecfilename, encoding='UTF8',
                       list_values=False, _inspec=True)
config = ConfigObj(filename, configspec=configspec)   Return Value

By default, the validate method either returns True (everything passed) or a dictionary of True / False representing pass/fail. The dictionary follows the structure of the ConfigObj.

If a whole section passes then it is replaced with the value True. If a whole section fails, then it is replaced with the value False.

If a value is missing, and there is no default in the check, then the check automatically fails.

The validate method takes an optional keyword argument preserve_errors. If you set this to True, instead of getting False for failed checks you get the actual error object from the validate module. This usually contains useful information about why the check failed.

See the flatten_errors function for how to turn your results dictionary into a useful list of error messages.

Even if preserve_errors is True, missing keys or sections will still be represented by a False in the results dictionary.   Mentioning Default Values

In the check in your configspec, you can specify a default to be used - by using the default keyword. E.g.

key1 = integer(0, 30, default=15)
key2 = integer(default=15)
key3 = boolean(default=True)
key4 = option('Hello', 'Goodbye', 'Not Today', default='Not Today')

If the configspec check supplies a default and the value is missing in the config, then the default will be set in your ConfigObj. (It is still passed to the Validator so that type conversion can be done: this means the default value must still pass the check.)

ConfigObj keeps a record of which values come from defaults, using the defaults attribute of sections. Any key in this list isn't written out by the write method. If a key is set from outside (even to the same value) then it is removed from the defaults list.

There is additionally a special case default value of None. If you set the default value to None and the value is missing, the value will always be set to None. As the other checks don't return None (unless you implement your own that do), you can tell that this value came from a default value (and was missing from the config file). It allows an easy way of implementing optional values. Simply check (and ignore) members that are set to None.


If stringify is False then default=None returns '' instead of None. This is because setting a value to a non-string raises an error if stringify is unset.

The default value can be a list. See List Values for the way to do this.

Writing invalid default values is a guaranteed way of confusing your users. Default values must pass the check.   Mentioning Repeated Sections and Values

In the configspec it is possible to cause every sub-section in a section to be validated using the same configspec. You do this with a section in the configspec called __many__. Every sub-section in that section has the __many__ configspec applied to it (without you having to explicitly name them in advance).

Your __many__ section can have nested subsections, which can also include __many__ type sections.

You can also specify that all values should be validated using the same configspec, by having a member with the name __many__. If you want to use repeated values along with repeated sections then you can call one of them ___many___ (triple underscores).

Sections with repeated sections or values can also have specifically named sub-sections or values. The __many__ configspec will only be used to validate entries that don't have an explicit configspec.

See Repeated Sections for examples.   Mentioning SimpleVal

If you just want to check if all members are present, then you can use the SimpleVal object that comes with ConfigObj. It only fails members if they are missing.

Write a configspec that has all the members you want to check for, but set every section to ''.

val = SimpleVal()
test = config.validate(val)
if test is True:
    print 'Succeeded.'   Mentioning copy Mode

As discussed in Mentioning Default Values, you can use a configspec to supply default values. These are marked in the ConfigObj instance as defaults, and not written out by the write mode. This means that your users only need to supply values that are different from the defaults.

This can be inconvenient if you do want to write out the default values, for example to write out a default config file.

If you set copy=True when you call validate, then no values are marked as defaults. In addition, all comments from the configspec are copied into your ConfigObj instance. You can then call write to create your config file.

There is a limitation with this. In order to allow String Interpolation to work within configspecs, DEFAULT sections are not processed by validation; even in copy mode.

5.1.3   reload

If a ConfigObj instance was loaded from the filesystem, then this method will reload it. It will also reuse any configspec you supplied at instantiation (including reloading it from the filesystem if you passed it in as a filename).

If the ConfigObj does not have a filename attribute pointing to a file, then a ReloadError will be raised.

5.1.4   reset

This method takes no arguments and doesn't return anything. It restores a ConfigObj instance to a freshly created state.

5.2   Attributes

A ConfigObj has the following attributes :

  • indent_type
  • interpolation
  • stringify
  • BOM
  • initial_comment
  • final_comment
  • list_values
  • encoding
  • default_encoding
  • unrepr
  • write_empty_values
  • newlines


This doesn't include comments, inline_comments, defaults, or configspec. These are actually attributes of Sections.

It also has the following attributes as a result of parsing. They correspond to options when the ConfigObj was created, but changing them has no effect.

  • raise_errors
  • create_empty
  • file_error

5.2.1   interpolation

ConfigObj can perform string interpolation in a similar way to ConfigParser. See the String Interpolation section for full details.

If interpolation is set to False, then interpolation is not done when you fetch values.

5.2.2   stringify

If this attribute is set (True) then the validate method changes the values in the ConfigObj. These are turned back into strings when write is called.

If stringify is unset (False) then attempting to set a value to a non string (or a list of strings) will raise a TypeError.

5.2.3   BOM

If the initial config file started with the UTF8 Unicode signature (known slightly incorrectly as the BOM - Byte Order Mark), or the UTF16 BOM, then this attribute is set to True. Otherwise it is False.

If it is set to True when write is called then, if encoding is set to None or to utf_8 (and variants) a UTF BOM will be written.

For UTF16 encodings, a BOM is always written.

5.2.4   initial_comment

This is a list of lines. If the ConfigObj is created from an existing file, it will contain any lines of comments before the start of the members.

If you create a new ConfigObj, this will be an empty list.

The write method puts these lines before it starts writing out the members.

5.2.5   final_comment

This is a list of lines. If the ConfigObj is created from an existing file, it will contain any lines of comments after the last member.

If you create a new ConfigObj, this will be an empty list.

The write method puts these lines after it finishes writing out the members.

5.2.6   list_values

This attribute is True or False. If set to False then values are not parsed for list values. In addition single line values are not unquoted.

This allows you to do your own parsing of values. It exists primarily to support the reading of the configspec - but has other use cases.

For example you could use the LineParser from the listquote module to read values for nested lists.

Single line values aren't quoted when writing - but multiline values are handled as normal.


Because values aren't quoted, leading or trailing whitespace can be lost. This behaviour was changed in version 4.0.1. Prior to this, single line values might have been quoted; even with list_values=False. This means that files written by earlier versions of ConfigObj could now be incompatible and need the quotes removing by hand.

5.2.7   encoding

This is the encoding used to encode the output, when you call write. It must be a valid encoding recognised by Python.

If this value is None then no encoding is done when write is called.

5.2.8   default_encoding

If encoding is set, any byte-strings in your ConfigObj instance (keys or members) will first be decoded to Unicode using the encoding specified by the default_encoding attribute. This ensures that the output is in the encoding specified.

If this value is None then sys.defaultencoding is used instead.

5.2.9   unrepr

Another boolean value. If this is set, then repr(value) is used to write values. This writes values in a slightly different way to the normal ConfigObj file syntax.

This preserves basic Python data-types when read back in. See unrepr mode for more details.

5.2.10   write_empty_values

Also boolean. If set, values that are an empty string ('') are written as empty values. See Empty Values for more details.

5.2.11   newlines

When a config file is read, ConfigObj records the type of newline separators in the file and uses this separator when writing. It defaults to None, and ConfigObj uses the system default (os.sep) if write is called without newlines having been set.

6   The Config File Format

You saw an example config file in the Config Files section. Here is a fuller specification of the config files used and created by ConfigObj.

The basic pattern for keywords is:

# comment line
# comment line
keyword = value # inline comment

Both keyword and value can optionally be surrounded in quotes. The equals sign is the only valid divider.

Values can have comments on the lines above them, and an inline comment after them. This, of course, is optional. See the comments section for details.

If a keyword or value starts or ends with whitespace, or contains a quote mark or comma, then it should be surrounded by quotes. Quotes are not necessary if whitespace is surrounded by non-whitespace.

Values can also be lists. Lists are comma separated. You indicate a single member list by a trailing comma. An empty list is shown by a single comma:

keyword1 = value1, value2, value3
keyword2 = value1, # a single member list
keyword3 = , # an empty list

Values that contain line breaks (multi-line values) can be surrounded by triple quotes. These can also be used if a value contains both types of quotes. List members cannot be surrounded by triple quotes:

keyword1 = ''' A multi line value
on several
lines'''     # with a comment
keyword2 = '''I won't be "afraid".'''
keyword3 = """ A multi line value
on several
lines"""     # with a comment
keyword4 = """I won't be "afraid"."""


There is no way of safely quoting values that contain both types of triple quotes.

A line that starts with a '#', possibly preceded by whitespace, is a comment.

New sections are indicated by a section marker line. That is the section name in square brackets. Whitespace around the section name is ignored. The name can be quoted with single or double quotes. The marker can have comments before it and an inline comment after it:

# The First Section
[ section name 1 ] # first section
keyword1 = value1

# The Second Section
[ "section name 2" ] # second section
keyword2 = value2

Any subsections (sections that are inside the current section) are designated by repeating the square brackets before and after the section name. The number of square brackets represents the nesting level of the sub-section. Square brackets may be separated by whitespace; such whitespace, however, will not be present in the output config written by the write method.

Indentation is not significant, but can be preserved. See the description of the indent_type option, in the ConfigObj specifications chapter, for the details.

A NestingError will be raised if the number of the opening and the closing brackets in a section marker is not the same, or if a sub-section's nesting level is greater than the nesting level of it parent plus one.

In the outer section, single values can only appear before any sub-section. Otherwise they will belong to the sub-section immediately before them:

# initial comment
keyword1 = value1
keyword2 = value2

[section 1]
keyword1 = value1
keyword2 = value2

    # this is in section 1
    keyword1 = value1
    keyword2 = value2

        [[[nested section]]]
        # this is in sub section
        keyword1 = value1
        keyword2 = value2

    # this is in section 1 again
    keyword1 = value1
    keyword2 = value2

# this is also in section 1, indentation is misleading here
keyword1 = value1
keyword2 = value2

# final comment

When parsed, the above config file produces the following data structure:

    'keyword1': 'value1',
    'keyword2': 'value2',
    'section 1': {
        'keyword1': 'value1',
        'keyword2': 'value2',
        'sub-section': {
            'keyword1': 'value1',
            'keyword2': 'value2',
            'nested section': {
                'keyword1': 'value1',
                'keyword2': 'value2',
        'sub-section2': {
            'keyword1': 'value1',
            'keyword2': 'value2',
        'sub-section3': {
            'keyword1': 'value1',
            'keyword2': 'value2',

Sections are ordered: note how the structure of the resulting ConfigObj is in the same order as the original file.


In ConfigObj 4.3.0 empty values became valid syntax. They are read as the empty string. There is also an option/attribute (write_empty_values) to allow the writing of these.

This is mainly to support 'legacy' config files, written from other applications. This is documented under Empty Values.

unrepr mode introduces another syntax variation, used for storing basic Python datatypes in config files.

7   Sections

Every section in a ConfigObj has certain properties. The ConfigObj itself also has these properties, because it too is a section (sometimes called the root section).

Section is a subclass of the standard new-class dictionary, therefore it has all the methods of a normal dictionary. This means you can update and clear sections.


You create a new section by assigning a member to be a dictionary.

The new Section is created from the dictionary, but isn't the same thing as the dictionary. (So references to the dictionary you use to create the section aren't references to the new section).

Note the following.

config = ConfigObj()
vals = {'key1': 'value 1',
        'key2': 'value 2'
config['vals'] = vals
config['vals'] == vals
config['vals'] is vals

If you now change vals, the changes won't be reflected in config['vals'].

A section is ordered, following its scalars and sections attributes documented below. This means that the following dictionary attributes return their results in order.

7.1   Section Attributes

  • main

    A reference to the main ConfigObj.

  • parent

    A reference to the 'parent' section, the section that this section is a member of.

    On the ConfigObj this attribute is a reference to itself. You can use this to walk up the sections, stopping when section.parent is section.

  • depth

    The nesting level of the current section.

    If you create a new ConfigObj and add sections, 1 will be added to the depth level between sections.

  • defaults

    This attribute is a list of scalars that came from default values. Values that came from defaults aren't written out by the write method. Setting any of these values in the section removes them from the defaults list.

  • default_values

    This attribute is a dictionary mapping keys to the default values for the keys. By default it is an empty dictionary and is populated when you validate the ConfigObj.

  • scalars, sections

    These attributes are normal lists, representing the order that members, single values and subsections appear in the section. The order will either be the order of the original config file, or the order that you added members.

    The order of members in this lists is the order that write creates in the config file. The scalars list is output before the sections list.

    Adding or removing members also alters these lists. You can manipulate the lists directly to alter the order of members.


    If you alter the scalars, sections, or defaults attributes so that they no longer reflect the contents of the section, you will break your ConfigObj.

    See also the rename method.

  • comments

    This is a dictionary of comments associated with each member. Each entry is a list of lines. These lines are written out before the member.

  • inline_comments

    This is another dictionary of comments associated with each member. Each entry is a string that is put inline with the member.

  • configspec

    The configspec attribute is a dictionary mapping scalars to checks. A check defines the expected type and possibly the allowed values for a member.

    The configspec has the same format as a config file, but instead of values it has a specification for the value (which may include a default value). The validate method uses it to check the config file makes sense. If a configspec is passed in when the ConfigObj is created, then it is parsed and broken up to become the configspec attribute of each section.

    If you didn't pass in a configspec, this attribute will be None on the root section (the main ConfigObj).

    You can set the configspec attribute directly on a section.

    See the validation section for full details of how to write configspecs.

  • extra_values

    By default an empty list. After validation this is populated with any members of the section that don't appear in the configspec (i.e. they are additional values). Rather than accessing this directly it may be more convenient to get all the extra values in a config file using the get_extra_values function.

    New in ConfigObj 4.7.0.

7.2   Section Methods

  • dict

    This method takes no arguments. It returns a deep copy of the section as a dictionary. All subsections will also be dictionaries, and list values will be copies, rather than references to the original [10].

  • rename

    rename(oldkey, newkey)

    This method renames a key, without affecting its position in the sequence.

  • merge


    This method is a recursive update method. It allows you to merge two config files together.

    You would typically use this to create a default ConfigObj and then merge in user settings. This way users only need to specify values that are different from the default.

    For example :

    # def_cfg contains your default config settings
    # user_cfg contains the user settings
    cfg = ConfigObj(def_cfg)
    usr = ConfigObj(user_cfg)
    cfg now contains a combination of the default settings and the user
    The user settings will have overwritten any of the default ones.
  • walk

    This method can be used to transform values and names. See walking a section for examples and explanation.

  • as_bool


    Returns True if the key contains a string that represents True, or is the True object.

    Returns False if the key contains a string that represents False, or is the False object.

    Raises a ValueError if the key contains anything else.

    Strings that represent True are (not case sensitive):

    true, yes, on, 1

    Strings that represent False are:

    false, no, off, 0
  • as_int


    This returns the value contained in the specified key as an integer.

    It raises a ValueError if the conversion can't be done.

  • as_float


    This returns the value contained in the specified key as a float.

    It raises a ValueError if the conversion can't be done.

  • as_list


    This returns the value contained in the specified key as a list.

    If it isn't a list it will be wrapped as a list so that you can guarantee the returned value will be a list.

  • restore_default


    Restore (and return) the default value for the specified key.

    This method will only work for a ConfigObj that was created with a configspec and has been validated.

    If there is no default value for this key, KeyError is raised.

  • restore_defaults


    Recursively restore default values to all members that have them.

    This method will only work for a ConfigObj that was created with a configspec and has been validated.

    It doesn't delete or modify entries without default values.

7.3   Walking a Section


The walk method allows you to call a function on every member/name.

walk(function, raise_errors=True,
     call_on_sections=False, **keywargs)

walk is a method of the Section object. This means it is also a method of ConfigObj.

It walks through every member and calls a function on the keyword and value. It walks recursively through subsections.

It returns a dictionary of all the computed values.

If the function raises an exception, the default is to propagate the error, and stop. If raise_errors=False then it sets the return value for that keyword to False instead, and continues. This is similar to the way validation works.

Your function receives the arguments (section, key). The current value is then section[key] [11]. Any unrecognised keyword arguments you pass to walk, are passed on to the function.

Normally walk just recurses into subsections. If you are transforming (or checking) names as well as values, then you want to be able to change the names of sections. In this case set call_on_sections to True. Now, on encountering a sub-section, first the function is called for the whole sub-section, and then it recurses into it's members. This means your function must be able to handle receiving dictionaries as well as strings and lists.

If you are using the return value from walk and call_on_sections, note that walk discards the return value when it calls your function.


You can use walk to transform the names of members of a section but you mustn't add or delete members.

7.4   Examples

You can use this for transforming all values in your ConfigObj. For example you might like the nested lists from ConfigObj 3. This was provided by the listquote module. You could switch off the parsing for list values (list_values=False) and use listquote to parse every value.

Another thing you might want to do is use the Python escape codes in your values. You might be used to using \n for line feed and \t for tab. Obviously we'd need to decode strings that come from the config file (using the escape codes). Before writing out we'll need to put the escape codes back in encode.

As an example we'll write a function to use with walk, that encodes or decodes values using the string-escape codec.

The function has to take each value and set the new value. As a bonus we'll create one function that will do decode or encode depending on a keyword argument.

We don't want to work with section names, we're only transforming values, so we can leave call_on_sections as False. This means the two datatypes we have to handle are strings and lists, we can ignore everything else. (We'll treat tuples as lists as well).

We're not using the return values, so it doesn't need to return anything, just change the values if appropriate.

def string_escape(section, key, encode=False):
    A function to encode or decode using the 'string-escape' codec.
    To be passed to the walk method of a ConfigObj.
    By default it decodes.
    To encode, pass in the keyword argument ``encode=True``.
    val = section[key]
    # is it a type we can work with
    # NOTE: for platforms where Python > 2.2
    # you can use basestring instead of (str, unicode)
    if not isinstance(val, (str, unicode, list, tuple)):
        # no !
    elif isinstance(val, (str, unicode)):
        # it's a string !
        if not encode:
            section[key] = val.decode('string-escape')
            section[key] = val.encode('string-escape')
        # it must be a list or tuple!
        # we'll be lazy and create a new list
        newval = []
        # we'll check every member of the list
        for entry in val:
            if isinstance(entry, (str, unicode)):
                if not encode:
        # done !
        section[key] =  newval

# assume we have a ConfigObj called ``config``
# To decode
# To encode.
# Because ``walk`` doesn't recognise the ``encode`` argument
# it passes it to our function.
config.walk(string_escape, encode=True)

Here's a simple example of using walk to transform names and values. One usecase of this would be to create a standard config file with placeholders for section and keynames. You can then use walk to create new config files and change values and member names :

# We use 'XXXX' as a placeholder
config = '''
XXXXkey1 = XXXXvalue1
XXXXkey2 = XXXXvalue2
XXXXkey3 = XXXXvalue3
XXXXkey1 = XXXXvalue1
XXXXkey2 = XXXXvalue2
XXXXkey3 = XXXXvalue3
XXXXkey1 = XXXXvalue1
XXXXkey2 = XXXXvalue2
XXXXkey3 = XXXXvalue3
    XXXXkey1 = XXXXvalue1
    XXXXkey2 = XXXXvalue2
    XXXXkey3 = XXXXvalue3
cfg = ConfigObj(config)
def transform(section, key):
    val = section[key]
    newkey = key.replace('XXXX', 'CLIENT1')
    section.rename(key, newkey)
    if isinstance(val, (tuple, list, dict)):
        val = val.replace('XXXX', 'CLIENT1')
        section[newkey] = val
cfg.walk(transform, call_on_sections=True)
print cfg
ConfigObj({'CLIENT1key1': 'CLIENT1value1', 'CLIENT1key2': 'CLIENT1value2',
'CLIENT1key3': 'CLIENT1value3',
'CLIENT1section1': {'CLIENT1key1': 'CLIENT1value1',
    'CLIENT1key2': 'CLIENT1value2', 'CLIENT1key3': 'CLIENT1value3'},
'CLIENT1section2': {'CLIENT1key1': 'CLIENT1value1',
    'CLIENT1key2': 'CLIENT1value2', 'CLIENT1key3': 'CLIENT1value3',
    'CLIENT1section1': {'CLIENT1key1': 'CLIENT1value1',
        'CLIENT1key2': 'CLIENT1value2', 'CLIENT1key3': 'CLIENT1value3'}}})

8   Exceptions

There are several places where ConfigObj may raise exceptions (other than because of bugs).

  1. If a configspec filename you pass in doesn't exist, or a config file

    filename doesn't exist and file_error=True, an IOError will be raised.

  2. If you try to set a non-string key, or a non string value when

    stringify=False, a TypeError will be raised.

  3. A badly built config file will cause parsing errors.

  4. A parsing error can also occur when reading a configspec.

  5. In string interpolation you can specify a value that doesn't exist, or

    create circular references (recursion).

Number 5 (which is actually two different types of exceptions) is documented in String Interpolation.

This section is about errors raised during parsing.

The base error class is ConfigObjError. This is a subclass of SyntaxError, so you can trap for SyntaxError without needing to directly import any of the ConfigObj exceptions.

The following other exceptions are defined (all deriving from ConfigObjError) :

When parsing a configspec, ConfigObj will stop on the first error it encounters. It will raise a ConfigspecError. This will have an error attribute, which is the actual error that was raised.

Behaviour when parsing a config file depends on the option raise_errors. If ConfigObj encounters an error while parsing a config file:

If raise_errors=True then ConfigObj will raise the appropriate error and parsing will stop.

If raise_errors=False (the default) then parsing will continue to the end and all errors will be collected.

If raise_errors is False and multiple errors are found a ConfigObjError is raised. The error raised has a config attribute, which is the parts of the ConfigObj that parsed successfully. It also has an attribute errors, which is a list of all the errors raised. Each entry in the list is an instance of the appropriate error type. Each one has the following attributes (useful for delivering a sensible error message to your user) :

If only one error is found, then that error is re-raised. The error still has the config and errors attributes. This means that your error handling code can be the same whether one error is raised in parsing , or several.

It also means that in the most common case (a single error) a useful error message will be raised.


One wrongly written line could break the basic structure of your config file. This could cause every line after it to flag an error, so having a list of all the lines that caused errors may not be as useful as it sounds.

9   Validation


The system of configspecs can seem confusing at first, but is actually quite simple and powerful. The best reference is my article on ConfigObj:

Validation is done through a combination of the configspec and a Validator object. For this you need [12]. See downloading if you don't have a copy.

Validation can perform two different operations :

  1. Check that a value meets a specification. For example, check that a value

    is an integer between one and six, or is a choice from a specific set of options.

  2. It can convert the value into the type required. For example, if one of

    your values is a port number, validation will turn it into an integer for you.

So validation can act as a transparent layer between the datatypes of your application configuration (boolean, integers, floats, etc) and the text format of your config file.

9.1   configspec

The validate method checks members against an entry in the configspec. Your configspec therefore resembles your config file, with a check for every member.

In order to perform validation you need a Validator object. This has several useful built-in check functions. You can also create your own custom functions and register them with your Validator object.

Each check is the name of one of these functions, including any parameters and keyword arguments. The configspecs look like function calls, and they map to function calls.

The basic datatypes that an un-extended Validator can test for are :

  • boolean values (True and False)
  • integers (including minimum and maximum values)
  • floats (including min and max)
  • strings (including min and max length)
  • IP addresses (v4 only)

It can also handle lists of these types and restrict a value to being one from a set of options.

An example configspec is going to look something like:

port = integer(0, 100)
user = string(max=25)
mode = option('quiet', 'loud', 'silent')

You can specify default values, and also have the same configspec applied to several sections. This is called repeated sections.

For full details on writing configspecs, please refer to the documentation.


Your configspec is read by ConfigObj in the same way as a config file.

That means you can do interpolation within your configspec.

In order to allow this, checks in the 'DEFAULT' section (of the root level of your configspec) are not used.

If you want to use a configspec without interpolation being done in it you can create your configspec manually and switch off interpolation:

from configobj import ConfigObj

configspec = ConfigObj(spec_filename, interpolation=False, list_values=False,
conf = ConfigObj(config_filename, configspec=configspec)

If you need to specify the encoding of your configspec, then you can pass in a ConfigObj instance as your configspec. When you read your configspec file, you must specify list_values=False. If you need to support hashes in configspec values then you must also pass in _inspec=True.

from configobj import ConfigObj
configspec = ConfigObj(configspecfilename, encoding='UTF8',
                       list_values=False, _inspec=True)
config = ConfigObj(filename, configspec=configspec)

9.2   Type Conversion

By default, validation does type conversion. This means that if you specify integer as the check, then calling validate will actually change the value to an integer (so long as the check succeeds).

It also means that when you call the write method, the value will be converted back into a string using the str function.

To switch this off, and leave values as strings after validation, you need to set the stringify attribute to False. If this is the case, attempting to set a value to a non-string will raise an error.

9.3   Default Values

You can set a default value in your check. If the value is missing from the config file then this value will be used instead. This means that your user only has to supply values that differ from the defaults.

If you don't supply a default then for a value to be missing is an error, and this will show in the return value from validate.

Additionally you can set the default to be None. This means the value will be set to None (the object) whichever check is used. (It will be set to '' rather than None if stringify is False). You can use this to easily implement optional values in your config files.

port = integer(0, 100, default=80)
user = string(max=25, default=0)
mode = option('quiet', 'loud', 'silent', default='loud')
nick = string(default=None)


Because the default goes through type conversion, it also has to pass the check.

Note that default=None is case sensitive.

9.3.1   List Values

It's possible that you will want to specify a list as a default value. To avoid confusing syntax with commas and quotes you use a list constructor to specify that keyword arguments are lists. This includes the default value. This makes checks look something like:

checkname(default=list('val1', 'val2', 'val3'))

This works with all keyword arguments, but is most useful for default values.

9.4   Repeated Sections

Repeated sections are a way of specifying a configspec for a section that should be applied to all unspecified subsections in the same section.

The easiest way of explaining this is to give an example. Suppose you have a config file that describes a dog. That dog has various attributes, but it can also have many fleas. You don't know in advance how many fleas there will be, or what they will be called, but you want each flea validated against the same configspec.

We can define a section called fleas. We want every flea in that section (every sub-section) to have the same configspec applied to it. We do this by defining a single section called __many__.

name = string(default=Rover)
age = float(0, 99, default=0)


        bloodsucker = boolean(default=True)
        children = integer(default=10000)
        size = option(small, tiny, micro, default=tiny)

Every flea on our dog will now be validated using the __many__ configspec.

__many__ sections can have sub-sections, including their own __many__ sub-sections. Defaults work in the normal way in repeated sections.

9.5   Repeated Values

As well as using __many__ to validate unspecified sections you can use it to validate values. For example, to specify that all values in a section should be integers:

    __many__ = integer

If you want to use repeated values alongside repeated sections you can call one __many__ and the other ___many___ (with three underscores).

9.6   Copy Mode

Because you can specify default values in your configspec, you can use ConfigObj to write out default config files for your application.

However, normally values supplied from a default in a configspec are not written out by the write method.

To do this, you need to specify copy=True when you call validate. As well as not marking values as default, all the comments in the configspec file will be copied into your ConfigObj instance.

from configobj import ConfigObj
from validate import Validator
vdt = Validator()
config = ConfigObj(configspec='default.ini')
config.filename = 'new_default.ini'
config.validate(vdt, copy=True)

If you need to support hashes in the configspec values then you must create it with _inspec=True. This has the side effect of switching off the parsing of inline comments, meaning that they won't be copied into the new config file. (ConfigObj syntax is slightly different from configspec syntax and the parser can't support both inline comments and hashes in configspec values.)

9.7   Validation and Interpolation

String interpolation and validation don't play well together. When validation changes type it sets the value. If the value uses interpolation, then the interpolation reference would normally be overwritten. Calling write would then use the absolute value and the interpolation reference would be lost.

As a compromise - if the value is unchanged by validation then it is not reset. This means strings that pass through validation unmodified will not be overwritten. If validation changes type - the value has to be overwritten, and any interpolation references are lost.

9.8   Extra Values

After validation the extra_values member of every section that is listed in the configspec will be populated with the names of members that are in the config file but not in the configspec.

If you are reporting configuration errors to your user this information can be useful, for example some missing entries may be due to misspelt entries that appear as extra values.

See the get_extra_values function

New in ConfigObj 4.7.0.

9.9   SimpleVal

You may not need a full validation process, but still want to check if all the expected values are present.

Provided as part of the ConfigObj module is the SimpleVal object. This has a dummy test method that always passes.

The only reason a test will fail is if the value is missing. The return value from validate will either be True, meaning all present, or a dictionary with False for all missing values/sections.

To use it, you still need to pass in a valid configspec when you create the ConfigObj, but just set all the values to ''. Then create an instance of SimpleVal and pass it to the validate method.

As a trivial example if you had the following config file:

# config file for an application
port = 80
protocol = http
domain = voidspace
top_level_domain =

You would write the following configspec:

port = ''
protocol = ''
domain = ''
top_level_domain = ''
config = Configobj(filename, configspec=configspec)
val = SimpleVal()
test = config.validate(val)
if test == True:
    print 'All values present.'
elif test == False:
    print 'No values present!'
    for entry in test:
        if test[entry] == False:
            print '"%s" missing.' % entry

10   Empty values

Many config files from other applications allow empty values. As of version 4.3.0, ConfigObj will read these as an empty string.

A new option/attribute has been added (write_empty_values) to allow ConfigObj to write empty strings as empty values.

from configobj import ConfigObj
cfg = '''
    key =
    key2 = # a comment
config = ConfigObj(cfg)
print config
ConfigObj({'key': '', 'key2': ''})

config.write_empty_values = True
for line in config.write():
    print line

key =
key2 =     # a comment

11   unrepr mode

The unrepr option allows you to store and retrieve the basic Python data-types using config files. It has to use a slightly different syntax to normal ConfigObj files. Unsurprisingly it uses Python syntax.

This means that lists are different (they are surrounded by square brackets), and strings must be quoted.

The types that unrepr can work with are :

strings, lists tuples
None, True, False
dictionaries, integers, floats
longs and complex numbers

You can't store classes, types or instances.

unrepr uses repr(object) to write out values, so it currently doesn't check that you are writing valid objects. If you attempt to read an unsupported value, ConfigObj will raise a configobj.UnknownType exception.

Values that are triple quoted cased. The triple quotes are removed before converting. This means that you can use triple quotes to write dictionaries over several lines in your config files. They won't be written like this though.

If you are writing config files by hand, for use with unrepr, you should be aware of the following differences from normal ConfigObj syntax :

List : ['A List', 'With', 'Strings']
Strings : "Must be quoted."
Backslash : "The backslash must be escaped \\"

These all follow normal Python syntax.

In unrepr mode inline comments are not saved. This is because lines are parsed using the compiler package which discards comments.

12   String Interpolation


String interpolation can slow down (slightly) the fetching of values from your config object. If you aren't using interpolation and it is performance critical then create your instance with interpolation=False.

ConfigObj allows string interpolation similar to the way ConfigParser or string.Template work. The value of the interpolation attribute determines which style of interpolation you want to use. Valid values are "ConfigParser" or "Template" (case-insensitive, so "configparser" and "template" will also work). For backwards compatibility reasons, the value True is also a valid value for the interpolation attribute, and will select ConfigParser-style interpolation. At some undetermined point in the future, that default may change to Template-style interpolation.

For ConfigParser-style interpolation, you specify a value to be substituted by including %(name)s in the value.

For Template-style interpolation, you specify a value to be substituted by including ${cl}name{cr} in the value. Alternately, if 'name' is a valid Python identifier (i.e., is composed of nothing but alphanumeric characters, plus the underscore character), then the braces are optional and the value can be written as $name.

Note that ConfigParser-style interpolation and Template-style interpolation are mutually exclusive; you cannot have a configuration file that's a mix of one or the other. Pick one and stick to it. Template-style interpolation is simpler to read and write by hand, and is recommended if you don't have a particular reason to use ConfigParser-style.

Interpolation checks first the current section to see if name is the key to a value. ('name' is case sensitive).

If it doesn't find it, next it checks the 'DEFAULT' sub-section of the current section.

If it still doesn't find it, it moves on to check the parent section and the parent section's 'DEFAULT' subsection, and so on all the way up to the main section.

If the value specified isn't found in any of these locations, then a MissingInterpolationOption error is raised (a subclass of ConfigObjError).

If it is found then the returned value is also checked for substitutions. This allows you to make up compound values (for example directory paths) that use more than one default value. It also means it's possible to create circular references. If there are any circular references which would cause an infinite interpolation loop, an InterpolationLoopError is raised.

Both of these errors are subclasses of InterpolationError, which is a subclass of ConfigObjError.

String interpolation and validation don't play well together. This is because validation overwrites values - and so may erase the interpolation references. See Validation and Interpolation. (This can only happen if validation has to change the value).

New in ConfigObj 4.7.0: String interpolation is now done in members of list values.

12.1   String Interpolation and List Values

Since version 4.7 string interpolation is done on string members of list values. If interpolation changes any members of the list then what you get back is a copy of the list rather than the original list.

This makes fetching list values slightly slower when interpolation is on, it also means that if you mutate the list changes won't be reflected in the original list:

>>> c = ConfigObj()
>>> c['foo'] = 'boo'
>>> c['bar'] = ['%(foo)s']
>>> c['bar']
>>> c['bar'].append('fish')
>>> c['bar']

Instead of mutating the list you must create a new list and reassign it.


Any line that starts with a '#', possibly preceded by whitespace, is a comment.

If a config file starts with comments then these are preserved as the initial_comment.

If a config file ends with comments then these are preserved as the final_comment.

Every key or section marker may have lines of comments immediately above it. These are saved as the comments attribute of the section. Each member is a list of lines.

You can also have a comment inline with a value. These are saved as the inline_comments attribute of the section, with one entry per member of the section.

Subsections (section markers in the config file) can also have comments.

See Section Attributes for more on these attributes.

These comments are all written back out by the write method.

14   flatten_errors

flatten_errors(cfg, res)

Validation is a powerful way of checking that the values supplied by the user make sense.

The validate method returns a results dictionary that represents pass or fail for each value. This doesn't give you any information about why the check failed.

flatten_errors is an example function that turns a results dictionary into a flat list, that only contains values that failed.

cfg is the ConfigObj instance being checked, res is the results dictionary returned by validate.

It returns a list of keys that failed. Each member of the list is a tuple:

([list of sections...], key, result)

If validate was called with preserve_errors=False (the default) then result will always be False.

list of sections is a flattened list of sections that the key was found in.

If the section was missing then key will be None.

If the value (or section) was missing then result will be False.

If validate was called with preserve_errors=True and a value was present, but failed the check, then result will be the exception object returned. You can use this as a string that describes the failure.

For example :

The value "3" is of the wrong type.

14.1   Example Usage

The output from flatten_errors is a list of tuples.

Here is an example of how you could present this information to the user.

vtor = validate.Validator()
# ini is your config file - cs is the configspec
cfg = ConfigObj(ini, configspec=cs)
res = cfg.validate(vtor, preserve_errors=True)
for entry in flatten_errors(cfg, res):
    # each entry is a tuple
    section_list, key, error = entry
    if key is not None:
        section_list.append('[missing section]')
    section_string = ', '.join(section_list)
    if error == False:
        error = 'Missing value or section.'
    print section_string, ' = ', error

15   get_extra_values


New in ConfigObj 4.7.0.

Find all the values and sections not in the configspec from a validated ConfigObj.

get_extra_values returns a list of tuples where each tuple represents either an extra section, or an extra value.

The tuples contain two values, a tuple representing the section the value is in and the name of the extra values. For extra values in the top level section the first member will be an empty tuple. For values in the 'foo' section the first member will be ('foo',). For members in the 'bar' subsection of the 'foo' section the first member will be ('foo', 'bar').

Extra sections will only have one entry. Values and subsections inside an extra section aren't listed separately.

NOTE: If you call get_extra_values on a ConfigObj instance that hasn't been validated it will return an empty list.

15.1   Example Usage

The output from get_extra_values is a list of tuples.

Here is an example of how you could present this information to the user.

vtor = validate.Validator()
# ini is your config file - cs is the configspec
cfg = ConfigObj(ini, configspec=cs)
cfg.validate(vtor, preserve_errors=True)

for sections, name in get_extra_values(cfg):

    # this code gets the extra values themselves
    the_section = cfg
    for section in sections:
        the_section = cfg[section]

    # the_value may be a section or a value
    the_value = the_section[name]

    section_or_value = 'value
    if isinstance(the_value, dict):
        # Sections are subclasses of dict
        section_or_value = 'section'

    section_string = ', '.join(sections) or "top level"
    print 'Extra entry in section: %s. Entry %r is a %s' % (section_string, name, section_or_value)


ConfigObj 4 is written by (and copyright) Michael Foord and Nicola Larosa.

Particularly thanks to Nicola Larosa for help on the config file spec, the validation system and the doctests. was originally written by Michael Foord and Mark Andrews.

Thanks to many others for input, patches and bugfixes.


ConfigObj, and related files, are licensed under the BSD license. This is a very unrestrictive license, but it comes with the usual disclaimer. This is free software: test it, break it, just don't blame us if it eats your data ! Of course if it does, let us know and we'll fix the problem so it doesn't happen to anyone else:

Copyright (c) 2004 - 2010, Michael Foord & Nicola Larosa
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are

    * Redistributions of source code must retain the above copyright
      notice, this list of conditions and the following disclaimer.

    * Redistributions in binary form must reproduce the above
      copyright notice, this list of conditions and the following
      disclaimer in the documentation and/or other materials provided
      with the distribution.

    * Neither the name of Michael Foord nor Nicola Larosa
      may be used to endorse or promote products derived from this
      software without specific prior written permission.


You should also be able to find a copy of this license at : BSD License

18   TODO

Better support for configuration from multiple files, including tracking where the original file came from and writing changes to the correct file.

Make newline a keyword argument (as well as an attribute) ?

UTF16 encoded files, when returned as a list of lines, will have the BOM at the start of every line. Should this be removed from all but the first line ?

Option to set warning type for unicode decode ? (Defaults to strict).

A method to optionally remove uniform indentation from multiline values. (do as an example of using walk - along with string-escape)

Should the results dictionary from validate be an ordered dictionary if odict is available ?

Implement some of the sequence methods (which include slicing) from the newer odict ?

Preserve line numbers of values (and possibly the original text of each value).



Please file any bug reports to Michael Foord or the ConfigObj Mailing List.

There is currently no way to specify the encoding of a configspec file.

As a consequence of the changes to configspec handling in version 4.6.0, when you create a ConfigObj instance and provide a configspec, the configspec attribute is only set on the ConfigObj instance - it isn't set on the sections until you validate. You also can't set the configspec attribute to be a dictionary. This wasn't documented but did work previously.

In order to fix the problem with hashes in configspecs I had to turn off the parsing of inline comments in configspecs. This will only affect you if you are using copy=True when validating and expecting inline comments to be copied from the configspec into the ConfigObj instance (all other comments will be copied as usual).

If you create the configspec by passing in a ConfigObj instance (usual way is to pass in a filename or list of lines) then you should pass in _inspec=True to the constructor to allow hashes in values. This is the magic that switches off inline comment parsing.

When using copy mode for validation, it won't copy DEFAULT sections. This is so that you can use interpolation in configspec files. This is probably true even if interpolation is off in the configspec.

You can't have a keyword with the same name as a section (in the same section). They are both dictionary keys - so they would overlap.

ConfigObj doesn't quote and unquote values if list_values=False. This means that leading or trailing whitespace in values will be lost when writing. (Unless you manually quote).

Interpolation checks first the current section, then the 'DEFAULT' subsection of the current section, before moving on to the current section's parent and so on up the tree.

Does it matter that we don't support the ':' divider, which is supported by ConfigParser ?

String interpolation and validation don't play well together. When validation changes type it sets the value. This will correctly fetch the value using interpolation - but then overwrite the interpolation reference. If the value is unchanged by validation (it's a string) - but other types will be.


This is an abbreviated changelog showing the major releases up to version 4. From version 4 it lists all releases and changes.

20.1   2010/02/27 - Version 4.7.2

  • BUGFIX: Restore Python 2.3 compatibility
  • BUGFIX: Members that were lists were being returned as copies due to interpolation introduced in 4.7. Lists are now only copies if interpolation changes a list member.
  • BUGFIX: pop now does interpolation in list values as well.
  • BUGFIX: where interpolation matches a section name rather than a value it is ignored instead of raising an exception on fetching the item.
  • BUGFIX: values that use interpolation to reference members that don't exist can now be repr'd.
  • BUGFIX: Fix to avoid writing '\r\r\n' on Windows when given a file opened in text write mode ('w').

See String Interpolation and List Values for information about the problem with lists and interpolation.

20.2   2010/02/06 - Version 4.7.1

  • Fix bug in options deprecation warning added in 4.7.0

20.3   2010/01/09 - Version 4.7.0

  • Minimum supported version of Python is now 2.3
  • ~25% performance improvement thanks to Christian Heimes
  • String interpolation now works in list value members
  • After validation any additional entries not in the configspec are listed in the extra_values section member
  • Addition of the get_extra_values function for finding all extra values in a validated ConfigObj instance
  • Deprecated the use of the options dictionary in the ConfigObj constructor and added explicit keyword arguments instead. Use **options if you want to initialise a ConfigObj instance from a dictionary
  • Constructing a ConfigObj from an existing ConfigObj instance now preserves the order of values and sections from the original instance in the new one
  • BUGFIX: Checks that failed validation would not populate default_values and restore_default_value() wouldn't work for those entries
  • BUGFIX: clear() now clears 'defaults'
  • BUGFIX: empty values in list values were accidentally valid syntax. They now raise a ParseError. e.g. "value = 1, , 2"
  • BUGFIX: Change to the result of a call to validate when preserve_errors is True. Previously sections where all values failed validation would return False for the section rather than preserving the errors. False will now only be returned for a section if it is missing
  • Distribution includes version 1.0.1 of
  • Removed __revision__ and __docformat__

20.4   2009/04/13 - Version 4.6.0

  • Pickling of ConfigObj instances now supported (thanks to Christian Heimes)
  • Hashes in confgspecs are now allowed (see note below)
  • Replaced use of hasattr (which can swallow exceptions) with getattr
  • __many__ in configspecs can refer to scalars (ordinary values) as well as sections
  • You can use ___many___ (three underscores!) where you want to use __many__ as well
  • You can now have normal sections inside configspec sections that use __many__
  • You can now create an empty ConfigObj with a configspec, programmatically set values and then validate
  • A section that was supplied as a value (or vice-versa) in the actual config file would cause an exception during validation (the config file is still broken of course, but it is now handled gracefully)
  • Added as_list method
  • Removed the deprecated istrue, encode and decode methods
  • Running now also runs the doctests in the configobj module

As a consequence of the changes to configspec handling, when you create a ConfigObj instance and provide a configspec, the configspec attribute is only set on the ConfigObj instance - it isn't set on the sections until you validate. You also can't set the configspec attribute to be a dictionary. This wasn't documented but did work previously.

In order to fix the problem with hashes in configspecs I had to turn off the parsing of inline comments in configspecs. This will only affect you if you are using copy=True when validating and expecting inline comments to be copied from the configspec into the ConfigObj instance (all other comments will be copied as usual).

If you create the configspec by passing in a ConfigObj instance (usual way is to pass in a filename or list of lines) then you should pass in _inspec=True to the constructor to allow hashes in values. This is the magic that switches off inline comment parsing.

20.5   2008/06/27 - Version 4.5.3

BUGFIX: fixed a problem with copy=True when validating with configspecs that use __many__ sections.

20.6   2008/02/05 - Version 4.5.2

Distribution updated to include version 0.3.2 of validate. This means that None as a default value in configspecs works.

20.7   2008/02/05 - Version 4.5.1

Distribution updated to include version 0.3.1 of validate. This means that Unicode configspecs now work.

20.8   2008/02/05 - Version 4.5.0

ConfigObj will now guarantee that files will be written terminated with a newline.

ConfigObj will no longer attempt to import the validate module, until/unless you call ConfigObj.validate with preserve_errors=True. This makes it faster to import.

New methods restore_default and restore_defaults. restore_default resets an entry to its default value (and returns that value). restore_defaults resets all entries to their default value. It doesn't modify entries without a default value. You must have validated a ConfigObj (which populates the default_values dictionary) before calling these methods.

BUGFIX: Proper quoting of keys, values and list values that contain hashes (when writing). When list_values=False, values containing hashes are triple quoted.

Added the reload method. This reloads a ConfigObj from file. If the filename attribute is not set then a ReloadError (a new exception inheriting from IOError) is raised.

BUGFIX: Files are read in with 'rb' mode, so that native/non-native line endings work!

Minor efficiency improvement in unrepr mode.

Added missing docstrings for some overidden dictionary methods.

Added the reset method. This restores a ConfigObj to a freshly created state.

Removed old CHANGELOG file.

20.9   2007/02/04 - Version 4.4.0

Official release of 4.4.0

20.10   2006/12/17 - Version 4.3.3-alpha4

By Nicola Larosa

Allowed arbitrary indentation in the indent_type parameter, removed the NUM_INDENT_SPACES and MAX_INTERPOL_DEPTH (a leftover) constants, added indentation tests (including another docutils workaround, sigh), updated the documentation.

By Michael Foord

Made the import of compiler conditional so that ConfigObj can be used with IronPython.

20.11   2006/12/17 - Version 4.3.3-alpha3

By Nicola Larosa

Added a missing self. in the _handle_comment method and a related test, per Sourceforge bug #1523975.

20.12   2006/12/09 - Version 4.3.3-alpha2

By Nicola Larosa

Changed interpolation search strategy, based on this patch by Robin Munn:

20.13   2006/12/09 - Version 4.3.3-alpha1

By Nicola Larosa

Added Template-style interpolation, with tests, based on this patch by Robin Munn: (awful archives, bad Sourceforge, bad).

20.14   2006/06/04 - Version 4.3.2

Changed error handling, if parsing finds a single error then that error will be re-raised. That error will still have an errors and a config attribute.

Fixed bug where '\n' terminated files could be truncated.

Bugfix in unrepr mode, it couldn't handle '#' in values. (Thanks to Philippe Normand for the report.)

As a consequence of this fix, ConfigObj doesn't now keep inline comments in unrepr mode. This is because the parser in the compiler package doesn't keep comments.

Error messages are now more useful. They tell you the number of parsing errors and the line number of the first error. (In the case of multiple errors.)

Line numbers in exceptions now start at 1, not 0.

Errors in unrepr mode are now handled the same way as in the normal mode. The errors stored will be an UnreprError.

20.15   2006/04/29 - Version 4.3.1

Added back into (Thanks to Stewart Midwinter)

Updated to 0.2.2.

Preserve tuples when calling the dict method. (Thanks to Gustavo Niemeyer.)

Changed __repr__ to return a string that contains ConfigObj({ ... }).

Change so that an options dictionary isn't modified by passing it to ConfigObj. (Thanks to Artarious.)

Added ability to handle negative integers in unrepr. (Thanks to Kevin Dangoor.)

20.16   2006/03/24 - Version 4.3.0

Moved the tests and the CHANGELOG (etc) into a separate file. This has reduced the size of by about 40%.

Added the unrepr mode to reading and writing config files. Thanks to Kevin Dangoor for this suggestion.

Empty values are now valid syntax. They are read as an empty string ''. (key =, or key = # comment.)

validate now honours the order of the configspec.

Added the copy mode to validate. Thanks to Louis Cordier for this suggestion.

Fixed bug where files written on windows could be given '\r\r\n' line terminators.

Fixed bug where last occurring comment line could be interpreted as the final comment if the last line isn't terminated.

Fixed bug where nested list values would be flattened when write is called. Now sub-lists have a string representation written instead.

Deprecated encode and decode methods instead.

You can now pass in a ConfigObj instance as a configspec (remember to read the configspec file using list_values=False).

Sorted footnotes in the docs.

20.17   2006/02/16 - Version 4.2.0

Removed BOM_UTF8 from __all__.

The BOM attribute has become a boolean. (Defaults to False.) It is only True for the UTF16/UTF8 encodings.

File like objects no longer need a seek attribute.

Full unicode support added. New options/attributes encoding, default_encoding.

ConfigObj no longer keeps a reference to file like objects. Instead the write method takes a file like object as an optional argument. (Which will be used in preference of the filename attribute if that exists as well.)

utf16 files decoded to unicode.

If BOM is True, but no encoding specified, then the utf8 BOM is written out at the start of the file. (It will normally only be True if the utf8 BOM was found when the file was read.)

Thanks to Aaron Bentley for help and testing on the unicode issues.

File paths are not converted to absolute paths, relative paths will remain relative as the filename attribute.

Fixed bug where final_comment wasn't returned if write is returning a list of lines.

Deprecated istrue, replaced it with as_bool.

Added as_int and as_float.

20.18   2005/12/14 - Version 4.1.0

Added merge, a recursive update.

Added preserve_errors to validate and the flatten_errors example function.

Thanks to Matthew Brett for suggestions and helping me iron out bugs.

Fixed bug where a config file is all comment, the comment will now be initial_comment rather than final_comment.

Validation no longer done on the 'DEFAULT' section (only in the root level). This allows interpolation in configspecs.

Also use the new list syntax in validate 0.2.1. (For configspecs).

20.19   2005/12/02 - Version 4.0.2

Fixed bug in create_empty. Thanks to Paul Jimenez for the report.

20.20   2005/11/05 - Version 4.0.1

Fixed bug in Section.walk when transforming names as well as values.

Added the istrue method. (Fetches the boolean equivalent of a string value).

Fixed list_values=False - they are now only quoted/unquoted if they are multiline values.

List values are written as item, item rather than item,item.

20.21   2005/10/17 - Version 4.0.0

ConfigObj 4.0.0 Final

Fixed bug in setdefault. When creating a new section with setdefault the reference returned would be to the dictionary passed in not to the new section. Bug fixed and behaviour documented.

Obscure typo/bug fixed in write. Wouldn't have affected anyone though.

20.22   2005/09/09 - Version 4.0.0 beta 5

Removed PositionError.

Allowed quotes around keys as documented.

Fixed bug with commas in comments. (matched as a list value)

20.23   2005/09/07 - Version 4.0.0 beta 4

Fixed bug in __delitem__. Deleting an item no longer deletes the inline_comments attribute.

Fixed bug in initialising ConfigObj from a ConfigObj.

Changed the mailing list address.

20.24   2005/08/28 - Version 4.0.0 beta 3

Interpolation is switched off before writing out files.

Fixed bug in handling StringIO instances. (Thanks to report from Gustavo Niemeyer.)

Moved the doctests from the __init__ method to a separate function. (For the sake of IDE calltips).

20.25   2005/08/25 - Version 4.0.0 beta 2

Amendments to

First public release.

20.26   2005/08/21 - Version 4.0.0 beta 1

Reads nested subsections to any depth.

Multiline values.

Simplified options and methods.

New list syntax.

Faster, smaller, and better parser.

Validation greatly improved. Includes:

  • type conversion
  • default values
  • repeated sections

Improved error handling.

Plus lots of other improvements.

20.27   2004/05/24 - Version 3.0.0

Several incompatible changes: another major overhaul and change. (Lots of improvements though).

Added support for standard config files with sections. This has an entirely new interface: each section is a dictionary of values.

Changed the update method to be called writein: update clashes with a dict method.

Made various attributes keyword arguments, added several.

Configspecs and orderlists have changed a great deal.

Removed support for adding dictionaries: use update instead.

Now subclasses a new class called caselessDict. This should add various dictionary methods that could have caused errors before.

It also preserves the original casing of keywords when writing them back out.

Comments are also saved using a caselessDict.

Using a non-string key will now raise a TypeError rather than converting the key.

Added an exceptions keyword for much better handling of errors.

Made creatempty=False the default.

Now checks indict and any keyword args. Keyword args take precedence over indict.

' ', ':', '=', ',' and '\t' are now all valid dividers where the keyword is unquoted.

ConfigObj now does no type checking against configspec when you set items.

delete and add methods removed (they were unnecessary).

Docs rewritten to include all this gumph and more; actually ConfigObj is really easy to use.

Support for stdout was removed.

A few new methods added.

Charmap is now incorporated into ConfigObj.

20.28   2004/03/14 - Version 2.0.0 beta

Re-written it to subclass dict. My first forays into inheritance and operator overloading.

The config object now behaves like a dictionary.

I've completely broken the interface, but I don't think anyone was really using it anyway.

This new version is much more 'classy'.

It will also read straight from/to a filename and completely parse a config file without you having to supply a config spec.

Uses listparse, so can handle nested list items as values.

No longer has getval and setval methods: use normal dictionary methods, or add and delete.

20.29   2004/01/29 - Version 1.0.5

Version 1.0.5 has a couple of bugfixes as well as a couple of useful additions over previous versions.

Since 1.0.0 the buildconfig function has been moved into this distribution, and the methods reset, verify, getval and setval have been added.

A couple of bugs have been fixed.

20.30   Origins

ConfigObj originated in a set of functions for reading config files in the atlantibots project. The original functions were written by Rob McNeur.

21   Footnotes

[1]And if you discover any bugs, let us know. We'll fix them quickly.
[2]If you specify a filename that doesn't exist, ConfigObj will assume you are creating a new one. See the create_empty and file_error options.
[3]They can be byte strings (ordinary strings) or Unicode.
[4]Except we don't support the RFC822 style line continuations, nor ':' as a divider.
[5]This is a change in ConfigObj 4.2.0. Note that ConfigObj doesn't call the seek method of any file like object you pass in. You may want to call yourself, first.

A side effect of this is that it enables you to copy a ConfigObj :

# only copies members
# not attributes/comments
config2 = ConfigObj(config1)

Since ConfigObj 4.7.0 the order of members and sections will be preserved when copying a ConfigObj instance.

[7]Other than lists of strings.
[8]The exception is if it detects a UTF16 encoded file which it must decode before parsing.
[9]The method signature shows that this method takes two arguments. The second is the section to be written. This is because the write method is called recursively.
[10]The dict method doesn't actually use the deepcopy mechanism. This means if you add nested lists (etc) to your ConfigObj, then the dictionary returned by dict may contain some references. For all normal ConfigObjs it will return a deepcopy.
[11]Passing (section, key) rather than (value, key) allows you to change the value by setting section[key] = newval. It also gives you access to the rename method of the section.
[12]Minimum required version of 0.2.0 .