[Numpy-discussion] Rotate Left and Rotate Right

[josef.pktd at gmail.com]

On Wed, May 20, 2009 at 7:53 AM, Cristi Constantin <darkgl0w at yahoo.com> wrote:
> Good day, me again.
> I have this string of data :
> String = 'i want\nto go\nto the\nbeach'.
>
> I want to rotate this data to left, or to right, after is split it after
> '\n'.
> Note that it's important to use 'U' array, because i might have unicode
> characters in this string.
>
> So normal, the text is:
> i want
> to go
> to the
> beach
>
> Rotated right would be:
> btti
> eoo
> a  w
> ctga
> hhon
>  e t
>
> Rotated left would be:
> t e
> nohh
> agtc
> w  a
>  ooe
> ittb
>
> There are a few methods i guess could be used.
>
> Split like: [np.array([j for j in i],'U') for i in String.split('\n')] =>
>     [ array([u'i', u' ', u'w', u'a', u'n', u't'], dtype='<U1'),
>     array([u't', u'o', u' ', u'g', u'o'], dtype='<U1'),
>     array([u't', u'o', u' ', u't', u'h', u'e'], dtype='<U1'),
>     array([u'b', u'e', u'a', u'c', u'h'], dtype='<U1') ]
>
> Or split like: np.array( String.split('\n'), 'U' ), => array([u'i want',
> u'to go', u'to the', u'beach'], dtype='<U6')
> Both methods are impossible to use with my knowledge.
>
> Without Numpy, if i want to rotate to right: i should reverse a list of
> splited by '\n' strings, save max length for all lines, align all lines to
> max len, rotate the square with map( lambda *row: [elem for elem in row],
> *Content ), then join the resulted sub-lists, unite with '\n' and return.
>
> I need to know if there is a faster Numpy approach... maybe explode the
> string by '\n' and rotate fast, or something?
>
> Thank you very much, in advance.
>

If you have many lines of different length, then using sparse matrices
might be useful.
scipy\maxentropy\examples  might be useful to look at.
I don't know much about language processing but the following seems to
work, using 1 character arrays:

Josef


>>> ss = u'i want\nto go\nto the\nbeach'
>>> lines = ss.split('\n')
>>> max(len(line) for line in lines)
6
>>> maxl = max(len(line) for line in lines)
>>> uarr = np.zeros((len(lines),maxl),dtype='<U1')
>>> uarr
array([[u'', u'', u'', u'', u'', u''],
       [u'', u'', u'', u'', u'', u''],
       [u'', u'', u'', u'', u'', u''],
       [u'', u'', u'', u'', u'', u'']],
      dtype='<U1')
>>> for i,line in enumerate(lines):
	uarr[i,:len(line)] = [j for j in line]

	
>>> uarr
array([[u'i', u' ', u'w', u'a', u'n', u't'],
       [u't', u'o', u' ', u'g', u'o', u''],
       [u't', u'o', u' ', u't', u'h', u'e'],
       [u'b', u'e', u'a', u'c', u'h', u'']],
      dtype='<U1')
>>> uarr[::-1,::-1]
array([[u'', u'h', u'c', u'a', u'e', u'b'],
       [u'e', u'h', u't', u' ', u'o', u't'],
       [u'', u'o', u'g', u' ', u'o', u't'],
       [u't', u'n', u'a', u'w', u' ', u'i']],
      dtype='<U1')
>>> uarr[:,::-1].T    # rotate left ?
array([[u't', u'', u'e', u''],
       [u'n', u'o', u'h', u'h'],
       [u'a', u'g', u't', u'c'],
       [u'w', u' ', u' ', u'a'],
       [u' ', u'o', u'o', u'e'],
       [u'i', u't', u't', u'b']],
      dtype='<U1')
>>> uarr[::-1,:].T    # rotate right ?
array([[u'b', u't', u't', u'i'],
       [u'e', u'o', u'o', u' '],
       [u'a', u' ', u' ', u'w'],
       [u'c', u't', u'g', u'a'],
       [u'h', u'h', u'o', u'n'],
       [u'', u'e', u'', u't']],
      dtype='<U1')