Previous was relational Distinct N or M.
It could be relational Union of N or M less N.
In J, you can selectively update new records,
which preserves order
NEW [`(i.&:(0&{"1)~)`] } MASTER
+--------+-+-+-+
|LiteralA|1|0|0|
+--------+-+-+-+
|LiteralB|2|4|3|
+--------+-+-+-+
|LiteralC|0|0|0|
+--------+-+-+-+
|LiteralD|0|3|5|
+--------+-+-+-+
|LiteralE|0|0|0|
+--------+-+-+-+
--- Joey K Tuttle <[EMAIL PROTECTED]> wrote:
> Making the assumption that unique keys are the objective, and that
> the non-zero numbers are normally not single digits (and are all, in
> fact, numbers) then my preference is to simply combine and eliminate
> duplicates e.g. -
>
> MASTER =: (,.<"1 'Literal',"1 ,.'ABCDE'),"1 ] 0;0;0
> NEW =: _4]\'LiteralA';1;0;0 ;'LiteralB';2;4;3;'LiteralD';0;3;5
> CONDENSE =: 13 : '(~: 0{"1 y)#y'
> CONDENSE NEW, MASTER
> +--------+-+-+-+
> |LiteralA|1|0|0|
> +--------+-+-+-+
> |LiteralB|2|4|3|
> +--------+-+-+-+
> |LiteralD|0|3|5|
> +--------+-+-+-+
> |LiteralC|0|0|0|
> +--------+-+-+-+
> |LiteralE|0|0|0|
> +--------+-+-+-+
>
> Of course, if you want the alphabetic key sort, then -
>
> /:~ CONDENSE NEW, MASTER
> +--------+-+-+-+
> |LiteralA|1|0|0|
> +--------+-+-+-+
> |LiteralB|2|4|3|
> +--------+-+-+-+
> |LiteralC|0|0|0|
> +--------+-+-+-+
> |LiteralD|0|3|5|
> +--------+-+-+-+
> |LiteralE|0|0|0|
> +--------+-+-+-+
>
> This does mean that if you want to return a file for Excel or some
> such you need to convert the numbers back to literal representations
> (so Excel can convert them again...) e.g. -
>
> untable ": &.> /:~ CONDENSE NEW, MASTER
> LiteralA 1 0 0
> LiteralB 2 4 3
> LiteralC 0 0 0
> LiteralD 0 3 5
> LiteralE 0 0 0
>
> untable =: 3 : 0
> TAB untable y
> :
> ; ((,(i.1{$y),. 1{$y),1+1{$y){"1 y,"1 x ;NL
> )
>
> TAB and NL have usual meanings (for DOS NL =: LF,CR)
>
> But perhaps energy would be better spent preening the keys. In my
> experience, such things often acquire "junk" like trailing, or even
> embedded, blanks. In any case, ~: is a good tool for the job.
>
>
> At 17:26 -0500 2008/04/12, Hahn, Harvey wrote:
> >Our library automation system's statistical reports contain only
> >non-zero rows of data. However, this makes them challenging to use
> >effectively because comparisons among reports (for example, in Excel)
> >requires data manipulations by hand to "normalize" two reports that
> >differ in which rows contained non-zero data. Because I'd like to
> >"clean up" this situation to help out our staff, I have a question about
> >which approach (algorithm) I should take in manipulating data (tables)
> >in J. What I'm thinking of doing is to compare each table file (upon
> >creation) against a "master" table file and then "normalizing" the new
> >data file so that, in the future, it can be compared with any other
> >table file. Below is a "toy example" of the two files, boxed, and the
> >desired end result. The first column of the two tables is the row name
> >to match upon (the "key", if you will). (The "master" table is merely a
> >table containing zero values for all possible row name values; these
> >zero values would have existed in the original, but suppressed, data
> >that was output in the "new" table.)
> >
> >"NEW" TABLE INSIDE J: "MASTER" FILE INSIDE J:
> >
> >+----------+---+---+---+ +----------+---+---+---+
> >| LiteralA | 1 | 0 | 0 | | LiteralA | 0 | 0 | 0 |
> >+----------+---+---+---+ +----------+---+---+---+
> >| LiteralB | 2 | 4 | 3 | | LiteralB | 0 | 0 | 0 |
> >+----------+---+---+---+ +----------+---+---+---+
> >| LiteralD | 0 | 3 | 5 | | LiteralC | 0 | 0 | 0 |
> >+----------+---+---+---+ +----------+---+---+---+
> > | LiteralD | 0 | 0 | 0 |
> > +----------+---+---+---+
> > | LiteralE | 0 | 0 | 0 |
> > +----------+---+---+---+
> >
> >DESIRED END RESULT:
> >
> >+----------+---+---+---+
> >| LiteralA | 1 | 0 | 0 |
> >+----------+---+---+---+
> >| LiteralB | 2 | 4 | 3 |
> >+----------+---+---+---+
> >| LiteralC | 0 | 0 | 0 |
> >+----------+---+---+---+
> >| LiteralD | 0 | 3 | 5 |
> >+----------+---+---+---+
> >| LiteralE | 0 | 0 | 0 |
> >+----------+---+---+---+
> >
> >My question: should I (1) attempt to copy/insert rows from the "master"
> >table into the "new" table or (2) overlay/replace the rows from the
> >"new" table in the "master" table (without, of course, destroying the
> >original "template", if you will, on disk)? Is there an algorithmic
> >preference in J (especially with regard to the matching of row names),
> >or is there not any significant difference between the two approaches?
> >Or does the choice depend on the size ratio between the "new" and
> >"master" tables?
> >
> >Thanks in advance for your opinions and insights!
> >
> >Harvey
> >
> >----------------------------------------------------------------------
> >For information about J forums see http://www.jsoftware.com/forums.htm
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
