Quoting Brian Schott <[EMAIL PROTECTED]>:
m132 =: ([:-.(1:,''"_=])#.1:)}.] NB. delete trailing blanks
The algorithm is correct, but there was a transcription error
(probably). the '' should have been ' ' (a single space instead
of an empty vector):
m132 =: ([:-.(1:,' '"_=])#.1:)}.] NB. delete trailing blanks
m132 'abc def '
abc def
FYI, the standard library has a lot of useful utilities pre-defined
for you, so you needn't go to the phrases very often. In this case,
for example:
load 'strings'
dtb
#~ ([: +./\. ' '&~:)
dtb 'abc def '
abc def
Perhaps JSoftware should change this definition to take advantage of
the special code supporting the train i: = 0:. Using the framework
provided by
http://www.jsoftware.com/jwiki/DanBron/Snippets/VerbMetrics and the
definitions:
...
dtb0 =: }.~ [: -. 1 #.~ 1 , ' ' = ]
dtb1 =: #~ [: +./\. ' ' ~: ]
dtb2 =: {.~ 1 + ' '&(= i: 0:)
...
small =: 50000 H b1 k =: 'abc def abc def abc '
medium =: 500 H b1 1000 {. k
large =: 1 H b1 5e6 {. k
...
The verb dtb2 is a clear winner on large datasets:
set alg time space
------ ---- ----- --------
dtb0 2.01 2.17
small dtb1 1.00 1.06
dtb2 1.03 1.00
dtb0 6.16 72.72
medium dtb1 3.04 29.28
dtb2 1.00 1.00
dtb0 18.26 36410.50
large dtb1 9.74 14564.39
dtb2 1.00 1.00
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
