Well...
That's actually a pretty big flaw in my implementation -- once there's
a letter pair which is not in a digraph, my rewrite completely fails.
It's certainly fixable, but with significant added complexity.
For example, this almost works:
lookup=: {&((8#2),8$0 3)
digrap=: # {. {. , 2|[:]F:.(lookup@(+4*])) (+ 2 * 2 | +/\)
dedouble=: #!.'X'~ 1 j. [: digrap #{.}.=}:
To make it actually work, I need to work around an error thrown from
F:. which I do not understand:
digrap=: #{.{.,2|[:]F:.(lookup@(+4*])) ::0: (+2*2|+/\)
This can probably be simplified...
Thanks,
--
Raul
On Thu, Feb 4, 2021 at 8:35 PM Ric Sherlock <tikk...@gmail.com> wrote:
>
> Yes as Hauke has previously identified, we need some more test cases to
> fully test whether the spec is correctly implemented or not.
> I'm thinking that the example outputs you've shown in your email aren't
> correct, but I agree that your approach and Raul's generate different
> solutions for that input and that your one (and Ben's and the current
> solution on Rosetta code) generate the "correct" output as I
> understand it. Here
> is an updated set of test cases:
>
> 'data1 soln1'=: 'THEEQUICKBROWFFOX';'THEXEQUICKBROWFXFOX'
> 'data2 soln2'=: 'THEEQUICKBROWFOOX';'THEXEQUICKBROWFOOX'
> 'data3 soln3'=: 'TTHEEQUICKBROWFFOX';'TXTHEXEQUICKBROWFXFOX'
> 'data4 soln4'=: 'TTHEEQUICKBROWFOOOX';'TXTHEXEQUICKBROWFOOXOX'
> 'data5 soln5'=: '';''
>
> showDigraphs=: dquote@(',' joinstring _2 <\ ])
>
> dd_bg=: {{ ((+ idx&<)#y) {. y ({. ,'X', }.)~ idx =. >:+: 1 i.~ (=/"1) _2
> ]\y }}^:_
> dd_md=: dtb@:((] ({.~ , 'X' , }.~) 1 (1 + 2 * i.)~ _2 =/\ ]) {.~ >:@#)
> ::($:@:,&' ')
> digra=: * 2 | i.@# + +/\
> dd_rm=: #!.'X'~ 1 j. [: digra # {. }. = }:
> dd_pj=: ;@:(_2&(,`([,'X',])@.=/each@<\))^:_
>
> Raul's approach looks as though with some more refinement of how to
> calculate which sets of repeat letters need to be split, it should be
> possible. At the moment it produces the incorrect result for data4.
>
> ([: showDigraphs dd_rm)&> data1;data2;data3;data4;data5
> "TH,EX,EQ,UI,CK,BR,OW,FX,FO,X"
> "TH,EX,EQ,UI,CK,BR,OW,FO,OX"
> "TX,TH,EX,EQ,UI,CK,BR,OW,FX,FO,X"
> "TX,TH,EX,EQ,UI,CK,BR,OW,FO,OO,X"
> ""
> Iteratively applying Pascal's approaches do result in all digraphs being
> split, but some pairs of letters are split unnecessarily.
>
>
> On Fri, Feb 5, 2021 at 5:04 AM 'Michael Day' via Programming <
> programm...@jsoftware.com> wrote:
>
> > Just before I venture into updating to the correct beta d...
> >
> > Raul's is elegant, single-pass and pretty spare. Here's a verbose,
> > inelegant,
> > multi-pass one, but it seems to work, except perhaps for the last
> > example,
> > below.
> >
> > I stuck with _2 which means that with odd length strings an error is thrown
> > on the last pair which is actually not a pair, so needs to be catered for.
> >
> > anyway, (it should appear on one line):
> > dedouble =: dtb@:((] ({.~ , 'X' , }.~) 1 (1 + 2 * i.)~ _2 =/\ ]) {.~
> > >:@#) ::($:@:,&' ')
> >
> > dtb is delete trailing blanks, in strings, I think.
> >
> > dedouble^:_'THEEQUICKBROWFFOOX'
> > THEXEQUICKBROWFXFOOX
> > dedouble^:_'THEEQUICKBROWFOOX'
> > THEXEQUICKBROWFOOX
> >
> > This next example is a bit unlikely to occur, but is this what's required?
> > dedouble 'THEEQUICKBROWFOOOX'
> > THEXEQUICKBROWFOOOX
> >
> > Raul's version doesn't agree with mine here! Is this result prederred?
> > rmdedouble 'THEEQUICKBROWFOOOX'
> > THEXEQUICKBROWFOOOX
> >
> > Cheers,
> >
> > Mike
> >
> >
> > On 04/02/2021 14:46, Raul Miller wrote:
> > > Ah, I see it now -- I should have looked closer at your digraphs.
> > >
> > > Here's a fixed version:
> > >
> > > digra=: * 2 | i.@# + +/\
> > > dedouble=: #!.'X'~ 1 j. [: digra #{.}.=}:
> > >
> > > (I could not think of a good name for the "compress out splits of
> > > non-digraphs" mechanism.)
> > >
> > > The moral of the story here is that numeric calculations can often
> > > replace simple recursive processes. (Because numbers can be defined
> > > recursively.)
> > >
> > > I hope this helps,
> > >
> >
> >
> > --
> > This email has been checked for viruses by Avast antivirus software.
> > https://www.avast.com/antivirus
> >
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> >
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
