If I understand your coding right,
l is indicating the characters per field (or until the next LF),
succeeding (all) the blanks at the beginning, whereas
t indicates the characters per field (or until the next LF), preceding
(all) the blanks at the end.
So with (l*.t)#y you loose the extreme spaces.
I thought about a sequential machine doing the same and created
NB. +---+-+----+-+-+-+-+------------------------+
NB. |' '| |x;LF| |Z| | |state |
NB. +---+-+----+-+-+-+-+------------------------+
NB. |1 |1|3 |1|4|1|0|initial |
NB. +---+-+----+-+-+-+-+------------------------+
NB. |1 |0|3 |1|4|1|1|begin blanks |
NB. +---+-+----+-+-+-+-+------------------------+
NB. |2 |0|3 |1|4|2|2|other consecutive blanks|
NB. +---+-+----+-+-+-+-+------------------------+
NB. |1 |2|3 |2|4|2|3|field seperator x, or LF|
NB. +---+-+----+-+-+-+-+------------------------+
NB. |2 |2|3 |2|4|2|4|no space, not x nor LF |
NB. +---+-+----+-+-+-+-+------------------------+
m=:<:+/2 1*(<a.)[EMAIL PROTECTED]&>' ';'+',LF NB. ' ' to 0, x;LF to 1, rest
to 2
S=: _2]\"1 }.".;._2 (0 : 0)
''' '' x;LF Z ']0
1 1 3 1 4 1
1 0 3 1 4 1
2 0 3 1 4 2
1 2 3 2 4 2
2 2 3 2 4 2
)
'+' fw2dl ' + 4.31:A +- :H + -:H+4.30:A+4.32:A+4.25:A+4.25',LF
+4.31:A+- :H+-:H+4.30:A+4.32:A+4.25:A+4.25
(1;S;m);: ' + 4.31:A +- :H + -:H+4.30:A+4.32:A+4.25:A+4.25',LF
+4.31:A+- :H+-:H+4.30:A+4.32:A+4.25:A+4.25
For data=: LF,~"1 [5000 1000$' + 4.31: A +-:H +-
:H+4.30:A+4.32:A+4.25:A+4.25 '
the performance figures are:
ts' ''+'' fw2dl"1 data '
0.28420038 21950912
ts' (1;S;m);: "1 data '
0.11558694 27071040
ts' ''+'' fw2dl ,data '
0.2271174 58722560
ts' (1;S;m);: ,data '
0.090205502 16778688
so a factor 2.5
((1;S;m)&;: -: '+'&fw2dl) ,data
1
((1;S;m)&;:"1 -: '+'&fw2dl"1) data
1
R.E. Boss
-----Oorspronkelijk bericht-----
Van: Dan Bron [mailto:[EMAIL PROTECTED]
Verzonden: maandag 5 februari 2007 16:08
Aan: [EMAIL PROTECTED]
CC: Programming Forum
Onderwerp: RE: [Jprogramming] Efficiently converting fixed-width to
delimited
R.E. Boss wrote:
>As I see it, you amend d by 1(0)}d but you don't use d anymore (in fw2dl),
>is that correct?
Nope, it's not correct. You found a bug!
The line after d =. 1 (0)} d should have used d as a left hand argument.
So the complete function should have been:
fw2dl =: verb define
TAB fw2dl y
:
s =. y ~: ' '
d =. y e. LF,x
l =. d ([: ; <@(+./\ );.2) s
d =. 1 (0)} d
NB. t =. s ([: ; <@(+./\.);.1) s NB. <-- Buggy version
t =. d ([: ; <@(+./\.);.1) s
e =. l*.t
y =. e#y
)
Thanks for noticing this and pointing it out! This was in production code!
I'm actually surprised the script didn't break.
I guess next time I'll choose variable names seperated by more one key.
Brevity has its drawbacks, too.
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
