docformatter Requirements¶
The goal of docformatter is to be an autoformatting tool for producing
PEP 257 compliant docstrings. This document provides a discussion of the
requirements from various sources for docformatter. Every effor will be
made to keep this document up to date, but this is not a formal requirements
document and shouldn’t be construed as such.
PEP 257 Requirements¶
PEP 257 provides conventions for docstrings. Conventions are general agreements or customs of usage rather than strict engineering requirements. This is appropriate for providing guidance to a broad community. In order to provide a tool for automatically formatting or style checking docstrings, however, some objective criteria is needed. Fortunately, the language of PEP 257 lends itself to defining objective criteria, or requirements, for such tools.
The conventions in PEP 257 define the high-level structure of docstrings:
How the docstring needs to be formatted.
What information needs to be in a docstring.
PEP 257 explicitly ignores markup syntax in the docstring; these are style choices left to the individual or organization to enforce. This gives us two categories of requirements in PEP 257. Let’s call them convention requirements and methodology requirements to be consistent with PEP 257 terminology.
An autoformatter should produce docstrings with the proper convention so tools
such as Docutils or pydocstyle can process them properly. The
contents of a docstring are irrelevant to tools like Docutils or
pydocstyle. An autoformatter may be able to produce some content, but
much of the content requirements would be difficult at best to satisfy
automatically.
Requirements take one of three types, shall, should, and may. Various sources provide definitions of, and synonyms for, these words. But generally:
Shall represents an absolute.
Should represents a goal.
May represents an option.
Thus, an autoformatting tool:
Must produce output that satisfies all the convention shall requirements.
Ought to provide arguments to allow the user to dictate how each convention should or may requirement is interpreted.
Would be nice to produce as much output that satisfies the methodology requirements.
Would be nice to provide arguments to allow the user to turn on/off each methodology requirement the tool supports.
Docstring Style¶
There are at least four “flavors” of docstrings in common use today; Epytext, Sphinx, NumPy, and Google. Each of these docstring flavors follow the PEP 257 convention requirements. What differs between the three docstring flavors is the reST syntax used in the field list of the multi-line docstring.
For example, here is how each syntax documents function arguments.
Epytext syntax:
@type num_dogs: int
@param num_dogs: the number of dogs
Sphinx syntax:
:param param1: The first parameter, defaults to 1.
:type: int
Google syntax:
Args:
param1 (int): The first parameter.
NumPy syntax:
Parameters
----------
param1 : int
The first parameter.
Syntax is also important to Docutils. An autoformatter should be aware of
syntactical directives so they can be placed properly in the structure of the
docstring. To accommodate the various syntax flavors used in docstrings, a
third requirement category is introduced, style.
Another consideration in the style category is line wrapping. According to PEP
257, splitting a one-line docstring is to allow “Emacs’ fill-paragraph command”
to be used. The fill-paragraph command is a line-wrapping command. Additionally,
it would be desirable to wrap docstrings for visual continuity with the code.
NumPy makes a stylistic decision to place a blank line after the long description.
Some code formatting tools also format docstrings. For example, black places a space before a one-line or the summary line when that line begins with a double quote (“). It would be desirable to provide the user an option to have docformatter also insert this space for compatibility.
Thus, an autoformatting tool:
Ought to provide arguments to allow the user to select the style or “flavor” of their choice.
Ought to provide arguments to allow the user to, as seamlessly as possible, produce output of a compatible style with other formatting tools in the eco-system.
Would be nice to to provide short cut arguments that represent aliases for a commonly used group of style arguments.
Program Control¶
Finally, how the docformatter tool is used should have some user-defined
options to accommodate various use-cases. These could best be described as
stakeholder requirements. An autoformatting tool:
Ought to provide arguments to allow the user to integrate it into their existing workflow.
Exceptions and Interpretations¶
As anyone who’s ever been involved with turning a set of engineering requirements into a real world product knows, they’re never crystal clear and they’re always revised along the way. Interpreting and taking exception to the requirements for an aerospace vehicle would be frowned upon without involving the people who wrote the requirements. However, the consequences for a PEP 257 autoformatting tool doing this are slightly less dire. We have confidence the GitHub issue system is the appropriate mechanism if there’s a misinterpretation or inappropriate exception taken.
The following items are exceptions or interpretations of the PEP 257 requirements:
One-line and summary lines can end with any punctuation.
docformatterwill recognize any of [. ! ?]. Exception to requirement PEP_257_4.5; consistent with Google style. See also #56 for situations when this is not desired.One-line and summary lines will have the first word capitalized.
docformatterwill capitalize the first word for grammatical correctness. Interpretation of requirement PEP_257_4.5. Some proper nouns are explicitly spelled using a lowercase letter (e.g.,docformatter). A user option is provided for a list of words to maintain lower case.PEP 257 discusses placing closing quotes on a new line in the multi-line section. However, it really makes no sense here as there is no way this condition could be met for a multi-line docstring. Given the basis provided in PEP 257, this requirement really applies to wrapped one-liners. Thus, this is assumed to apply to wrapped one-liners and the closing quotes will be placed on a line by themselves in this case. However, an argument will be provided to allow the user to select their desired behavior. Interpretation of requirement PEP_257_5.5.
These give rise to the derived requirement category which would also cover any requirements that must be met for a higher level requirement to be met.
The table below summarizes the requirements for docformatter. It
includes an ID for reference, the description from PEP 257, which category
the requirement falls in, the type of requirement, and whether
docformatter has implemented the requirement.
|
|
|
|
|
|---|---|---|---|---|
|
Always use “””triple double quotes””” |
|
|
|
|
Use r”””raw triple double quotes””” if you use backslashes. |
|
|
|
|
Use u”””unicode triple double quotes””” for unicode docstrings. |
|
|
|
|
One-line docstrings: |
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Multi-line docstrings: |
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Class docstrings: |
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Script docstring: |
|||
|
|
|
|
|
|
|
|
|
|
|
Module and Package docstrings: |
|||
|
|
|
|
|
|
Function and Method docstrings: |
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docstring Syntax |
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Shall |
|
|
|
|
|
|
|
Program Control |
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Requirement ID’s that begin with PEP_257 are taken from PEP 257. Those prefaced with docformatter are un-related to PEP 257.
Test Suite¶
Each requirement in the table above should have one or more test in the test suite to verify compliance. Ideally the test docstring will reference the requirement(s) it is verifying to provide traceability.
Current Implementation¶
docformatter currently provides the following arguments for interacting
with convention requirements.
--pre-summary-newline [boolean, default False]
Boolean to indicate whether to place the summary line on the line after
the opening quotes in a multi-line docstring. See requirement
PEP_257_5.2.
docformatter currently provides these arguments for style requirements.
-s, --style [string, default sphinx]
name of the docstring syntax style to use for formatting parameter
lists.
--rest-section-adorns [REGEX, default [!\"#$%&'()*+,-./\\:;<=>?@[]^_`{|}~]{4,}]
regular expression for identifying reST section adornments
-n, --non-cap [string, default []]
list of words not to capitalize when they appear as the first word in the
summary
--black [boolean, default False]
Boolean to indicate whether to format docstrings to be compatible
with black.
--blank [boolean, default False]
Boolean to indicate whether to add a blank line after the
elaborate description.
--close-quotes-on-newline [boolean, default False]
Boolean to indicate whether to place closing triple quotes on new line
for wrapped one-line docstrings.
--make-summary-multi-line [boolean, default False]
Boolean to indicate whether to add a newline before and after a
one-line docstring. This option results in non-conventional
docstrings; violates requirements PEP_257_4.1 and PEP_257_4.3.
--non-strict [boolean, default False]
Boolean to indicate whether to ignore strict compliance with reST list
syntax (see issue #67).
--pre-summary-space [boolean, default False]
Boolean to indicate whether to add a space between the opening triple
quotes and the first word in a one-line or summary line of a
multi-line docstring.
--tab-width [integer, defaults to 1]
Sets the number of characters represented by a tab when line
wrapping, for Richard Hendricks and others who use tabs instead of
spaces.
--wrap-descriptions length [integer, default 79]
Wrap long descriptions at this length.
--wrap-summaries length [integer, default 72]
Wrap long one-line docstrings and summary lines in multi-line
docstrings at this length.
docformatter currently provides these arguments for stakeholder requirements.
--check
Only check and report incorrectly formatted files.
--config CONFIG
Path to the file containing docformatter options.
--docstring-length min_length max_length
Only format docstrings that are [min_length, max_length] rows long.
--exclude
Exclude directories and files by names.
--force-wrap
Force descriptions to be wrapped even if it may result in a mess.
This should likely be removed after implementing the syntax option.
--in-place
Make changes to files instead of printing diffs.
--range start end
Only format docstrings that are between [start, end] rows in the file.
--recursive
Drill down directories recursively.
Arguments Needed for Future Releases¶
The following are new arguments that are needed to implement should or may convention requirements:
--wrap-one-line [boolean, default False]
Boolean to indicate whether to wrap one-line docstrings. Provides
option for requirement PEP_257_4.1.
Issue and Version Management¶
As bug reports and feature requests arise in the GitHub issue system, these will need to be prioritized. The requirement categories, coupled with the urgency of the issue reported can be used to provide the general prioritization scheme:
Priority 1: convention bug
Priority 2: style bug
Priority 3: stakeholder bug
Priority 4: convention enhancement
Priority 5: style enhancement
Priority 6: stakeholder enhancement
Priority 7: chore
Integration of a bug fix will result in a patch version bump (i.e., 1.5.0 -> 1.5.1). Integration of one or more enhancements will result in a minor version bump (i.e., 1.5.0 -> 1.6.0). One or more release candidates will be provided for each minor or major version bump. These will be indicated by appending -rcX to the version number, where the X is the release candidate number beginning with 1. Release candidates will not be uploaded to PyPi, but will be made available via GitHub Releases.