Almost gave up on that one - but got it solved somehow and still don't know
why it works...
I started to parse the instructions into 14 iteration verbs i0-i13. For
this I used a mapping m for translating instructions into J verbs - mapping
inp to : is merely used for splitting; the jx verb then creates the
corresponding J expression for the instructions which are eventually
written to i, the J code to produce the iterations. I also wrote i to disc
for inspection in a spreadsheet:
m=: 'inp
';':';'-';'_';'add';'+';'mul';'*';'div';'<.@%';'mod';'|~';'eql';'='
d=: >: i.9
jx=: {{ ('[ ', >@(1&{), '=.', [: ; 1 0 2&{) <;._1 ' '&, y }}
NB. i holds the J code for iteration verbs i0, i1,... ,i13
i=: ,/"2 ([: jx@> [: |.@}. <;._2);._1 rplc&m fread 'xxiv.txt'
i=: i,"1 '[ ''w x y z''=. |:y,.~0,.~d,.0 [ k=.d+10*x }}"0'
# ". i=: i,~"1 (,&'=. {{ k,.z')@('i'&,)@":"0 i.#i
14
'xxiv.ijs' fwrite~ (,~ ,&LF)~/ i NB. Write i to disc for inspection
3681
The maximum and minimum model number are then computed with the adverb mn -
which takes >. and <. as a verb resp.- constructed from the J code stored
in noun j. mn filters in 2 ways:
1. The verb dm calculates the distinct maximum/minimum. This was not enough
- ran out of memory.
2. Then it appeared that z can only get smaller when it is divided by 26
during the iteration and then it gets the same values as 2 iterations
before. So in that case throw away all values larger than the maximum 2
iterations ago, which is stored in q - p is the maximum of the previous
iteration. Valid records are stored in r.
This at least solved my puzzle input:
NB. j holds the J code for the model number adverb mn
j=: 'mn=: 1 : 0',LF
j=: j, 'dm=. ([: ~. {:"1) ,.~ {:"1 u//. {."1',LF
j=: j, '''q p''=. 0, >./ {:"1 r=. i0/ 0 0',LF
j=: j, ; {{
'''q p''=. p, >./ {:"1 r=. (#~ q>:{:"1)^:(0<[: I.&q {:"1)
dm,/i',(":y),'/"1 r',LF
}} &.> }.i.#i
j=: j, '{., (#~ 0={:"1) r',LF ,')'
{{ load y [ j fwrite y }} 'xxiv.ijs'
>.mn NB. (*)
99394899891971
<.mn NB. (**)
92171126131911
I was puzzled that I had to write >.mn - first tried to use >.mn 0 - but
that's probably due to y missing in mn's definition.
What also didn't work was executing j with ". - is there a way of doing
that without using fwrite and load?
Resources used were quite small:
timespacex '>.mn'
0.0939154 8786816
Thanks. Stefan.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
