openclonk/thirdparty/timsort/README.rst

132 lines
3.9 KiB
ReStructuredText

------
sort.h
------
Overview
--------
sort.h is an implementation a ton of sorting algorithms in C with a
user-defined type, that is defined at include time.
This means you don't have to pay the function call overhead of using
standard library routine.
You get the choice of many extra sorting routines as well, including:
* Shell sort
* Binary insertion sort
* Heap sort
* Quick sort
* Merge sort
* Bubble sort (ugh)
* Tim sort
If you don't know which one to use, you should probably use Tim sort.
Usage
-----
To use this library, you need to do three things:
* #define SORT_TYPE to be the type of the elements of the array you
want to sort.
* #define SORT_NAME to be a unique name that will be prepended to all
the routines, i.e., #define SORT_NAME mine would give you routines
named mine_heap_sort, and so forth.
* #include "sort.h". Make sure that sort.h is in your include path,
obviously.
Then, enjoy using the sorting routines.
See demo.c for example usage.
If you are going to use your own custom type, you must redefine
SORT_CMP(x, y) with your comparison function, so that it returns
a value less than zero if x < y, equal to zero if x == y, and
greater than 0 if x > y.
The default just uses the builtin <, ==, and > operators:
#define SORT_CMP(x, y) ((x) < (y) ? -1 : ((x) == (y) ? 0 : 1))
It is often just fine to just subtract the arguments as well (though
this can cause some stability problems with floating-point types):
#define SORT_CMP(x, y) ((x) - (y))
Speed of routines
-----------------
The speed of each routine is highly dependent on your computer and the
structure of your data.
If your data has a lot of, like partially sorted sequences, then Tim sort
will beat the pants off of anything else.
In general, Tim sort is probably the best sorting algorithm in this library,
even for random data.
Tim sort is not as good if memory movement is many orders of magnitude more
expensive than comparisons (like, many more than for normal int and double).
If so, then quick sort is probably your routine. On the other hand, Tim
sort does extremely well if the comparison operator is very expensive,
since it strives hard to minimize comparisons.
Here is the output of demo.c, which will give you the timings for a run of
10,000 things on my old Mac Pro (2006-era 2.66 GHz Xeons, 64-bit) on OS X 10.6:
::
Running tests
quick sort time: 740.20 us per iteration
bubble sort time: 183914.60 us per iteration
merge sort time: 954.20 us per iteration
binary insertion sort time: 20472.70 us per iteration
heap sort time: 994.50 us per iteration
shell sort time: 1170.30 us per iteration
tim sort time: 708.50 us per iteration
Author
------
Christopher Swenson (chris@caswenson.com)
References
----------
* Wikipedia
* timsort.txt (under doc/)
License
-------
All code in this repository, unless otherwise specified, is hereby
licensed under the MIT Public License:
Copyright (c) 2010 Christopher Swenson
Permission is hereby granted, free of charge, to any person
obtaining a copy of this software and associated documentation
files (the "Software"), to deal in the Software without
restriction, including without limitation the rights to use,
copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following
conditions:
The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
OTHER DEALINGS IN THE SOFTWARE.