jeffswt / pydatagen

Create random data easier with pydatagen.

pydatagen

Create random data easier with pydatagen.

Now you can use Python to generate random data for OI problems, in a really Pythonified way!

Ports

You can use printf just like the way you use it on Codeforces, USACO, BZOJ and pretty much every Online Judge. It's just that simple, only the fancy formats are ignored in this version.

If you want to print something, simply write:

printf("%d %d\n", n, m);

The simple applications are really of no differences with C++.

We are currently working on the porting of scanf.

Random Generator

We provide two generators, one is called rand and the other xrand. The only difference between the two is, when the same function is called regularly, xrand functions 10x faster than rand.

For instance, the two snippets of code functions exactly the same:

for i in range(0, 10000000):
    rand(int, 64, 128)

And the optimized version as follows:

g = xrand(int, 64, 128)
for i in range(0, 10000000):
    next(g)

Detailed usage of these functions can be referred in the Python interactive help console by typing help('pydatagen.rand'), had you installed this package.

Installation

Clone the repository into an empty folder, and execute the following command:

pip install setuptools # If you have already installed this before, ignore it
python setup.py install # Install pydatagen

You can also build your Wheel package by executing the following command:

pip install setuptools # If you have already installed this before, ignore it
pip install wheel # Install bdist_wheel provider
python setup.py build bdist_wheel # Build .whl package

Alternatively you could download our official release from the releases panel, and install the compiled 'wheel' to your computer. After you have downloaded the release, you may install it in the command line:

pip install ./pydatagen-version-py3-none-any.whl # Referring to the downloaded file

Installation is as fast as creating random data!

Pros and Cons

The pros of using pydatagen:

• Easier to write than C++ random generators
• Shorten your code by 80%
• Allows more advanced ways of data generation
• Extensible modules makes adding data type support easier

However, there are a few drawbacks that cannot be avoided at present:

• Slow generation speed, around 3x slower than general Python approaches
• Way slower than NumPy and C++ implementations

If you don't care much about performance, than we believe that pydatagen is the best choice you have. It also works well with pyjudge and pyMatcher.

pydatagen's Issues

Support formatting to string in addition to print_oi

Now print_oi directly sends the data to stdout. It would be nice if one could get the formatted string instead.

kw argument `distinct` is not documented in `randlist`

by the way,

>>> random.sample([1,2,3,4,5], 3)
[3, 4, 1]

UnionFindSet shouldn't use classmethod decorator

the attributes are shared across instances in classmethods, causing unexpected bug.

for example,

>>> class Test:
...   @classmethod
...   def __init__(cls,x):
...     cls.x=x
...
>>> a=Test(1)
>>> a.x
1
>>> b=Test(2)
>>> a.x
2
>>>

stop using `queue`

queue is designed for multithreading uses and acquire a lock in every method call. in a word, it's really slow.

just use list instead. it's much (about 50x) faster.

>>> import timeit
>>> timeit.timeit('q.put(123)', 'import queue; q=queue.Queue()', number=10**6)
24.48412793993996
>>> timeit.timeit('q.append(123)', 'q=[]', number=10**6)
0.5576901608900329

Graph only succeeds with less than 20% probability

Using master

Python 3.4.3 (default, Jun 29 2015, 12:16:01) 
[GCC 5.1.1 20150618 (Red Hat 5.1.1-4)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from pydatagen import *
>>> def test():
...   try:
...     Graph(6, 12, connected=True, weighed=True, weight_range=range(10, 99))
...     return True
...   except:
...     return False
... 
>>> results = [test() for _ in range(20000)]
>>> results.count(True) / len(results)
0.18835

DRAMATICALLY enhance the speed

def possibility(rate):
    return random.random()>rate

improve `printf`

try this:

f=open('1.txt','w')
from contextlib import redirect_stdout
with redirect_stdout(f):
    print('hello world')

or

f=open('1.txt','w')
print('hello world', file=f)