Developer Guide

This document describes the internal architecture and the main concepts behind ini2toml.

Please notice this document does not target the end users, but instead people that are involved in the development of the project or that wish to write ini2toml plugins.

How it works

ini2toml converts your .ini/.cfg files into a data structure with an intermediate representation and then serialise this data structure using a TOML library.

This transformation works as a 5-stage data pipeline:

  1. The original .ini/.cfg text file is pre-processed.

  2. ini2toml parses the .ini/.cfg file contents using ConfigParser (“lite” flavour) or ConfigUpdater (“full” flavour) and transforms it into an intermediate data structure.

  3. This intermediate object is further processed.

  4. ini2toml convert the intermediate representation into a string that uses TOML syntax.

  5. The resulting TOML text file is post-processed.

Core Concepts

The INI syntax (as supported by Python’s standard library) is not very expressive or even clearly defined, at least in terms of data types. For example, while TOML have clear ways of indicating whether values are strings and numbers (or even more complex lists and associative tables), most of the times it is up to the INI syntax user the responsibility of identifying the correct data type of a stored value and convert it according to the context.

Since, over the time different communities of .ini/.cfg file users ended up creating different conventions on how to “encode” complex values, it is very hard for a generic conversion tool to get every single document right.

ini2toml address this challenge by relying on a system of Profiles and allowing third-party Plugins, as documented in the next sections.


A profile is a simple collection of text and intermediate representation transformations, responsible for adjusting or correcting any non-trivial translation between the original .ini/.cfg file format and the resulting TOML (such as converting values to specific data types or changing field names or configuration keys).

This collection of transformations is identified by a string (the profile name), which in general corresponds to a file naming convention. This is motivated by the tradition of different communities using specific file names for their use cases.

For example, the Python community uses the setup.cfg file to store packaging metadata. Therefore, ini2toml built-in profile named "setup.cfg" is responsible for converting "setup.cfg" files into PEP 621-compliant TOML documents.

Each profile will correspond to a specific pipeline being selected for execution. When using the ini2toml command line tool without explicitly specifying a profile, the basename of the input file will be used if it is implemented, falling back to "setup.cfg".

Pre-processing and Post-processing

Pre-processing and post-processing are simple text-processing transformations (i.e. the text contents are transformed from a string object to another string object). The difference is that pre-processors will receive as input a text following the INI syntax, while post-processors will receive as input a text with the converted result, following the TOML syntax.

Each text-processor is a simple Python function with the following signature:

def text_process(file_contents: str) -> str: ...


All processors are called in sequence, so the output of one is the input of the following (also working as a pipeline). Ideally processor implementations should be idempotent.

Intermediate representation processing

Processing the intermediate representation allows more powerful transformations, including converting stored values to specific types (e.g. a INI string value to a TOML list) or combining several INI options into a nested TOML table.

Each intermediate-processor is a simple Python function with the following signature:

def intermediate_process(intermediate: IntermediateRepr) -> IntermediateRepr: ...

IntermediateRepr is a special kind of Python object with characteristics of both dict and list. It respects the MutableMapping protocol, but also adds some handy position-dependent methods - such as insert(), index(), append() - and the very useful rename() method.

IntermediateRepr can contain any kind of built-in Python object supported by TOML (e.g. numbers, strings, lists, booleans…) and also a few other special objects that carry comments along:

  • Commented: A wrapper around a Python built-in value carrying an in-line comment:

    ir = Commented(42, "comment")  # => represents `42 # comment`
  • CommentedList: A wrapper around a list of elements. The elements are organised in groups (that are equivalent to a single line in the TOML document), with each group being a Commented[list]. For example:

    ir = IntermediateRepr()
    ir["x"] = CommentedList(
            Commented([0, 1], "comment"),
            Commented([2], "other"),

    is equivalent to the TOML:

    x = [
        0, 1, # comment
        2, # other
  • CommentedKV: similar to CommentedList, but each element is a key-value pair. For example:

    ir = IntermediateRepr()
    ir["x"] = CommentedKV([Commented([("a", 1), ("b", 2)], "comment")])
    ir["y"] = CommentedKV(
        [Commented([("a", 1), ("b", 2)], "comment"), Commented([("c", 3)], "other")]

    is equivalent to the TOML:

    x = {a = 1, b = 2} # comment
    a = 1
    b = 2 # comment
    c = 3 # other

    Due to TOML limitations, you can only have “one-line” inline-tables, therefore CommentedKV objects with more than one group are automatically converted to full-blown TOML tables.

A comment or newline can be added directly to the intermediate representation, by using a HiddenKey:

ir = IntermediateRepr()
ir[CommentKey()] = "comment"  # => represents `# comment`
ir[WhitespaceKey()] = ""  # => represents a `"\n"` in the TOML

Also notice that IntermediateRepr objects can be nested.


Plugins are a way of extending the built-in ini2toml functionality, by adding processors to specific profiles using the Python programming language.

The implementation requirement for a ini2toml plugin is a function that accepts a Translator object. Using this object it is possible to register new processors for different profiles, as shown in the example below.

from ini2toml.types import Translator

def activate(translator: Translator):
    profile = translator["setup.cfg"]  # will be ``setup.cfg``
    desc = "Convert 'setup.cfg' files to 'pyproject.toml' based on PEP 621"
    profile.description = desc

Profile-independent processing via profile augmentation

Sometimes it might be useful to implement generic processing tasks that do not depend on the nature/focus of the file being converted and therefore do not belong to a specific profile (e.g. removing trailing newline, blank lines, …). The augment_profiles() mechanism in ini2toml allow plugins to include such processing tasks, by enabling them to modify the profile after it is selected.

An example of these - here called “profile augmentation functions” - is shown in the following example:

from ini2toml.types import Translator, Profile

def activate(translator: Translator):
    translator.augment_profiles(strip_trailing_newline, active_by_default=True)

def strip_trailing_newline(profile: Profile):
    """Remove trailing newline from the generated TOML file"""

Customising the CLI help text

ini2toml will try to automatically generate a help text to be displayed in the CLI for the registered profiles based on the name and help_text properties of the Profile objects. If help_text is blank, the profile will not be featured in the CLI description (i.e. it will be a hidden profile).

ini2toml will also generate a “on/off”-style CLI option flag (depending on the active_by_default value) for each “profile-augmentation function”. By default, the name and docstring of the function registered with augment_profiles() will be used to create the CLI help text, but this can also be customised via optional keyword arguments name and help_text. Differently from profiles, these flags will always be visible in the CLI, independently of the values of help_text.

Distributing Plugins

To distribute ini2toml plugins, it is necessary to create a Python package with a ini2toml.processing entry-point.

For the time being, if using setuptools, this can be achieved by adding the following to your setup.cfg file:

# in setup.cfg
ini2toml.processing =
    your_plugin = your_package.your_module:your_activate_function

When using a PEP 621-compliant backend, the following can be add to your pyproject.toml file:

# in pyproject.toml
"ini2toml.processing" = {your_plugin = "your_package.your_module:activate"}

It is recommended that plugins created by the community and meant to be publicly shared are distributed via PyPI under a name that adheres to the following convention:

ini2toml-contrib-<your specific name>

with <your specific name> being the same string identifier used as entry-point.

Please notice plugins are activated in a specific order, which can interfere with the order that the processors run. They are sorted using Python’s built-in sorted function.