This page is primarily intended for Mercurial's developers.
How to make your code pretty the Mercurial way.
This page is intended to save new developers a few round-trips when contributing changes. It doesn't cover everything, but it does cover some of the most common mistakes people make.
2. Some code doesn't agree with the coding style!
Yes, because some code predates the coding style and has not yet been rewritten to conform. Please follow the coding style for all new code.
3. Naming conventions
For consistency and ease of reference, Mercurial uses a single style for all identifiers: all lowercase, with no underbars between words. This matches Python's core style (with the notable exception of has_key which is deprecated). For private methods and helper functions, the convention is to use a single leading underbar. Use a trailing underbar to avoid shadowing built-ins and imported modules.
Throughout the code, the following variables usually refer to the same thing:
first and second parents
a context.filectx instance
a python file(like) object
4. Whitespace and syntax
We approximately follow PEP 8 guidelines for whitespace:
- don't use tabs
- use four spaces to indent
- add a linebreak after a colon
- use whitespace around most operators
- don't use lines longer than 80 characters
- don't leave trailing whitespace
5. Language features and compatibility
- Mercurial is designed for Python 2.4 - 2.7 and onward so don't use features from 2.5 or later:
- Don't add default arguments to new functions unless you're going to use them
Avoid wasting effort by asking mpm for permission to use a class
- Think twice about using classes, functions are almost always smaller and simpler
- Think three times if you are a recovering Java programmer
Class names should still be lowercase (but exception classes are CamelCased)
- Classes should derive from 'object'
- Private methods and variables should be indicated with a leading underscore
- Destructors in Python are not reliable and should be avoided
7. Unicode and character encoding
Character encoding is a complex topic (see Encoding Strategy for an overview) but Mercurial generally follows these rules:
- Almost all string data is manipulated either as plain byte strings in the local encoding or in no encoding
- Mercurial-specific metadata like usernames is converted to UTF-8 byte strings in a restricted number of places using fromlocal/tolocal
Unicode objects are avoided wherever possible and no core APIs are designed to handle them
8. Status and error messages
- use _() to mark things for i18n
Short messages should look like this:
Note the following:
it starts with a lower-case word.
it has no trailing full-stop (.).
Some existing strings don't follow this style and are kept like that for backwards compatibility reasons. But please write all new strings in this style.
The above message should look like this in your code:
ui.status(_('adding %s\n') % filename)
The _ function is used to mark the string as translatable. Import it from the i18n module.
The string interpolation is done after the call to the _ function.
- Don't put OS-specific hacks outside of util.py and friends
- add docstrings
follow the HelpStyleGuide when adding help texts
10. Automated checking
A lot of the rules in this document and a bunch of others can be automatically checked with our 'check-code' script:
$ python contrib/check-code.py --blame `hg manifest` mercurial/i18n.py:42 (working directory): > u = u'\n\n'.join([p and t.ugettext(unicode(p, 'ascii')) or '' for p in paragraphs]) line too long
We recommend you run this script before every submission. In addition to catching style and portability issues in Python code, it will also inspect our C code and test suite.
11. See also
Contributing changes - how to send us your changes
Help style guide - pointers on writing online help text
Writing tests - how to extend the test suite