| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
| |
The ordering of accented A's and a's was not the same between them -
the accent has been made the same in an attempt to improve the
situation.
|
|
|
|
| |
Missing call to swapcase. Copy/paste error.
|
| |
|
| |
|
| |
|
|
|
|
| |
Consolidate into parametrized tests.
|
|
|
|
|
|
|
| |
All tests have been put into a class so that only one skipif clause
is needed (no tests are run on Python 3+).
Refactor test that ensures an error is raised to use pytest.raises.
|
|
|
|
|
|
| |
This is intended to replace the purpose slow_splitters.py, by checking
several different types of input against all the regular expressions
used by natsort to split strings into numbers/non-numbers.
|
|
|
|
|
|
|
|
|
| |
Much unneeded duplication has been removed. Most of the work in this
file had been going into validating the regular expressions - this
will be added elsewhere in a future commit.
slow_splitters.py has been removed. Docs have been updated to remove
the reference to slow_splitters.py.
|
|
|
|
|
| |
Consolidated tests into a smaller number of tests using
parametrization. Also improved warning testing.
|
|
|
|
|
|
|
|
| |
All tests have been consolidated into a single parametrized test. Much
nicer.
Additionally, some updates to the fake_fastnumbers were added as they
were noticed during refactoring.
|
|
|
|
|
| |
Better use of fixtures and parameterization, less unnecessary
hypothesis testing.
|
|
|
|
| |
Remove many unnecessary tests and details.
|
|
|
|
| |
Reduced code to two parameterized functions.
|
|
|
|
|
| |
This makes the default value in argument lists clearer, as well
as making unit tests more clear to read.
|
|
|
|
| |
All tests have been consolidated into a single parameterized test.
|
|
|
|
| |
Eliminate some code duplication with fixtures, clarify variable names.
|
|
|
|
|
| |
This refactor uses parametrize wherever possible, and also places
expected output into an "expected" variable for clarity.
|
|
|
|
|
| |
All of the functions that are to be used outside of utils.py
have had their leading underscore removed.
|
|
|
|
|
|
|
|
|
| |
The class of warnings that have been silenced are:
- shadowing names in outer scope
- incorrect docstring
- invalid type (where inspection was bad)
- unresolved references (in Python2/3 compat code)
|
|
|
|
|
| |
The refactoring uses parameterizations to reduce copy/paste. It also
uses more modern pytest functionality to check exception details.
|
|
|
|
|
|
| |
- Utils functions are now accessed as part of the utils namespace.
- Use of parametrize where appropriate.
- Remove use of slow splitters utility functions.
|
|
|
|
|
|
| |
These have been moved to conftest.py, or now use the pytest-mocker
third-party plugin. Some tests have been re-written to account for the
change.
|
|
|
|
|
|
|
| |
Previously, many module-level regular expressions were created, then
accessed through a dictionary. The regular expressions are now
encapsulated in a class, and the user access the dictionary through
a function call which can better handle cleaning up the key accessor.
|
|
|
|
| |
Only the modules with odd import mechanics have been left out.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The reasons for doing this were as follows:
1. The fields could be added programmatically and thus be able to be
made more clearer by placing into well-named variables.
2. Better follow style guide by having the instance be all lower
case and not the class itself.
Additionally, the _DUMB field was removed and made an internal
module-level value, so that it is certainly not exposed to the user
and the code need not access the private attribute of the ns class.
|
| |
|
|
|
|
|
|
| |
The new implementation has only one import statement, and
blocks PyCharm from warning against fastnumbers not being installed
(since it is optional this is preferred).
|
|
|
|
| |
Its only purpose was to raise the given error, which is not necessary.
|
|
|
|
| |
The small differences between Python 3 and Python 2 are now clearer.
|
| |
|
|
|
|
|
|
|
|
|
| |
Many of the utility functions had terse docstrings. All functions
now have full docstrings that should help developers who want to
contribute better understand what is going on.
Some function names or arguments have been changed if it improved
self-documentation.
|
|
|
|
|
|
|
|
| |
As part of the blackening, the huge tuple of hex literals that represent
unicode numerals has been placed into its own file, and the _version.py
contents has been moved to __init__.py.
PEP8 checks are updated to conform to black's style.
|
| |
|
| |
|
|\
| |
| |
| |
| |
| | |
It was important to me to get CI running on Python 3.7 before
I made a fully Python 3.7 compatible release as requested
in issue #62.
|
| |
| |
| |
| |
| |
| | |
This is because fastnumbers has a python 3.7 bug, so
there is no point using it for the tests if it fails.
A future commit will revert this change.
|
| |
| |
| |
| |
| |
| | |
It seems like once one introduces a matrix, implicit
matrices no longer work. So I now have to spell out the
with and without extras stuff.
|
|/
|
|
|
| |
Apparently, the ubuntu trusty distribution cannot run python
3.7, but xenial will. For 3.7 we have to specify xenial.
|
|
|
|
| |
This holds the previously created "How it works" section.
|
|
|
|
| |
This reverts commit b25949f157fa163a378701f5eca8a8cf1b02820a.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
This gives a high-level description of how natsort works,
and also some caveats so that users have a better idea of
what natsort is doing under the hood.
This address and closes issue #60.
|
| |
|
| |
|
| |
|
| |
|