On Fri, Apr 29, 2011 at 3:13 PM, Raul Miller <[email protected]> wrote:
>> it raises index error on an empty string.
>> evalbs ''
...
>> evalbs '\a\\'
>> \a\\
Here's a fix for both issues:
NB. backslash evaluator
subst2=:(_2<\'\"\\\/\b\f\n\r\t')&(i.{(34 92 47 8 12 10 13 9{&.><a.),])"0
hexchars=. 'ABCDEF0123456789abcdef'
dfh=: 16 #. 16 | _6 + hexchars i. ]
subst6=:(8 u:[:u:@dfh 2}.]) ::]^:(6&=@#*.('\'={.)*.'uU'e.~{.@}.)&.>
splitbs=: (0;(0 10#:10*".;._2]0 :0);(a.e.hexchars,'Uu')+2*a.='\')&;:
1.1 1.1 2.1 NB. state 0 -- start
1.0 1.0 2.2 NB. state 1 -- ordinary characters
3.0 4.0 3.0 NB. state 2 -- \
1.2 1.2 2.2 NB. state 3 -- \n or \u0000
1.2 5.0 2.2 NB. state 4 -- \u
1.2 6.0 2.2 NB. state 5 -- \u0
1.2 7.0 2.2 NB. state 6 -- \u00
1.2 3.0 2.2 NB. state 7 -- \u000
)
evalbs=: [:; [:subst2 [:subst6 splitbs^:(*@#)
NB. test suite for backslash evaluator
testbs=: (7 u: evalbs@[) assert@-: 4 u: ]
NB. ordinary text should remain unchanged
'abcd' testbs 97+i.4
NB. unicode escapes should work
'\u005c\U00a5\U20a9' testbs 92 165 8361
NB. backslash escapes should work
'\"\\\/\b\f\n\r\t' testbs 34 92 47 8 12 10 13 9
NB. empty strings should work
'' testbs ''
NB. backslash processing should not get confused about state
'\a\\' testbs 92 97 92
Let me know if you see any other problems.
Thanks,
--
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
