Sorting indexed arrays

[norbert doerre]

PB9 online Help describes multidimensional array sorting the following way:

TYPE SalesType
OrderNum AS LONG
PartNumber(1 TO 20) AS STRING * 20
END TYPE
...
DIM Sales AS SalesType
...
DIM Temp(1 TO 20) AS STRING * 20 AT VARPTR(Sales.Partnumber(1))
ARRAY SORT Temp()
ERASE Temp()

The problem is always the need of using a type outside a function.
So, if you need often to calculate a smallest/biggest value under variant conditions then you need a bunch of types. Functions with Min/Max calculations are as a consequence not simply portable.

In PB there exists a one dimensional function Min/Max, which unfortunately cannot be used with arrays. It is quite useless for common math purposes. So, why not simply implement a Min/Max function like in C++ with at least one row containing an index() array and at least one row containing a value() array with hard codable relevance to the index() array?

Currently, I have to use functions with built in bubble sorting or similiar. These are too much time consuming in some cases.

[Michael Mattias]

Send those new feature suggestions ... eg

  Z =   SUM|MIN|MAX|MEAN|AVERAGE (numericarray ([Start element])[FOR count])

.. to [email protected]

[Michael Mattias]

BTW, you don't need to sort arrays to find any of the aggregate functions listed.

[norbert doerre]

Sent it to support slightly changed:

I'd introduce a suggestion for handling indexed numeric arrays.

I changed Michael's thoughts from
Z = SUM|MIN|MAX|MEAN|AVERAGE (numericarray, ([Start element])[FOR count])

to

Z(index) = SUM|MIN|MAX|MEAN|AVERAGE (indexed_numericarray, index, ([Start element])[FOR count])
index = the resulting index of the given back indexed_numericarray
indexed_numericarray = element(index)

indexed_numericarray Z(index) means that the array values have an index value attached similiar to a hidden item value in combo boxes.
The index values must always remain unchangeable and "hard coded" to their specific array values. So it is possible
to identify the values by their indices even after the values have been resorted.
The elements of the array must therefore have the syntax
element(index, value).
Sample:
Z(index) = MIN(indexed_numericarray(), index, 12, For 44)

[John Gleason]

Yes, no sort is needed for min and max etc. aggregates. The below finds min and max of 3 fields in a 64MB file in a couple seconds.

#COMPILE EXE
#DIM ALL

TYPE partType
   partNum     AS QUAD
   partName    AS STRING * 20
   partCode    AS LONG
END TYPE

FUNCTION PBMAIN () AS LONG
    LOCAL xs AS STRING, y AS LONG, qMax, qMin AS QUAD
    LOCAL qMaxS, qMinS AS STRING * 20, qMaxL, qMinL AS LONG
    xs = SPACE$(64000032)
    OPEN "C:\boomBinFil03.dat" FOR BINARY SHARED AS #1 'any file > 64MB for demo purposes. A video file is a good stand-in.
    GET #1,, xs
    DIM x(2000000) AS partType AT STRPTR(xs)

    CALL quikMaxMin(x(), qMax, qMin, qMaxS, qMinS, qMaxL, qMinL, 4) 'find the max and min of chosen field, or all 3--CASE 4.
    REPLACE ANY CHR$(0,9,10,13) WITH "OBLC" IN qMaxS 'for display reasons only
    REPLACE ANY CHR$(0,9,10,13) WITH "OBLC" IN qMinS
    ? STR$(qMax) & $CRLF & STR$(qMin) & $CRLF & $CRLF & qMaxS & $CRLF & qMinS & $CRLF & $CRLF & _
      STR$(qMaxL) & $CRLF & STR$(qMinL)

END FUNCTION

FUNCTION quikMaxMin(x() AS partType, qMax AS QUAD, qMin AS QUAD, _
                                     qMaxS AS STRING * 20, qMinS AS STRING * 20, _
                                     qMaxL AS LONG, qMinL AS LONG, _
                                     choice AS LONG) AS LONG

   LOCAL ii AS LONG

   SELECT CASE choice
      CASE 1
         qMax = &hffffffffffffffff
         qMin = &h7fffffffffffffff

         FOR ii = 0 TO UBOUND(x())
            IF x(ii).partNum > qMax THEN qMax = x(ii).partNum
            IF x(ii).partNum < qMin THEN qMin = x(ii).partNum
         NEXT

      CASE 2
         qMaxS = ""
         qMinS = STRING$(20, 255)

         FOR ii = 0 TO UBOUND(x())
            IF x(ii).partName > qMaxS THEN qMaxS = x(ii).partName
            IF x(ii).partName < qMinS THEN qMinS = x(ii).partName
         NEXT

      CASE 3
         qMaxL =  &hffffffff
         qMinL =  &h7fffffff

         FOR ii = 0 TO UBOUND(x())
            IF x(ii).partCode > qMaxL THEN qMaxL = x(ii).partCode
            IF x(ii).partCode < qMinL THEN qMinL = x(ii).partCode
         NEXT

      CASE 4                    'all fields max and min
         qMax = &hffffffffffffffff
         qMin = &h7fffffffffffffff
         qMaxS = ""
         qMinS = STRING$(20, 255)
         qMaxL =  &hffffffff
         qMinL =  &h7fffffff

         FOR ii = 0 TO UBOUND(x())
            IF x(ii).partNum  > qMax  THEN qMax  = x(ii).partNum
            IF x(ii).partNum  < qMin  THEN qMin  = x(ii).partNum
            IF x(ii).partName > qMaxS THEN qMaxS = x(ii).partName
            IF x(ii).partName < qMinS THEN qMinS = x(ii).partName
            IF x(ii).partCode > qMaxL THEN qMaxL = x(ii).partCode
            IF x(ii).partCode < qMinL THEN qMinL = x(ii).partCode
         NEXT
   END SELECT

END FUNCTION

[Rodney Hicks]

Seems like those items (SUM|MIN|MAX|AVERAGE|MEAN|MEDIAN|STNDEV|VARIANCE|etc) would be more suited to the ARRAY SCAN statement.

[Gösta H. Lovgren-2]

" AVERAGE|MEAN"

I've never understood the difference between "Average" and "Mean". Is ther any difference? I've just figured they were the same and 'In" people used "Mean" and real men used "average" (having no need to impress anyone).

==========================================
"Most people would sooner die than think;
in fact, they do so."
Bertrand Russell (1872-1970)
==========================================

[David Roberts]

'Average' and 'Mean' are synonymous and 'Mean' is short for 'Arithmetic Mean' as opposed to Harmonic Mean and Geometric Mean.

[Rodney Hicks]

In my accounting days, AVERAGE was the average, MEAN was the closest of the figures to the average, MEDIAN was the one in the middle. The only place I ever encountered the situation regarding MEAN was in the course itself. AVERAGE and MEDIAN I used outside the course in practice.
The courses were taken before the days of the personal computer, which may have a little bearing on the slight differences in what we're accustomed to now.

And Gösta,
Mediocre men use MEDIAN, mean men use MEAN, average men use AVERAGE, and real men tell someone else to do it!!

[John Gleason]

Seems like those items (SUM|MIN|MAX|AVERAGE|MEAN|MEDIAN|STNDEV|VARIANCE|etc) would be more suited to the ARRAY SCAN statement.

How about ARRAY STATS(arr([index]), SUM|MIN|MAX|ETC)

[Michael Mattias]

Like I said, send those new feature suggestions to the address above.

That way the PBPTB are looking at ALL the ideas and can pick and choose the best/most requested/easiest to use.

Me, I really don't need any more intrinsics which may not be used in the function position; that is I would much prefer

   X =      functionname(arguments)

instead of

   functionname (arguments)  [U]TO X [/U]

MCM

[Michael Mattias]

FWIW, I don't see any value to supporting an "index" value attached to/associated with an array element.... just create another array of same size to hold related values OR change the array to a UDT array with an 'IndexValue' member.

[John Gleason]

FWIW, I don't see any value to supporting an "index" value attached to/associated with an array element....

I second that codemotion.

[Bern Ertl]

Norbert, you might be able to create your own library (or class) to encapsulate the new functions (or data type) that you need. I had a similar problem recently.

[Michael Mattias]

>indexed_numericarray

Just so we can be on the same page, what exactly do you mean by an "indexed array?"

That term has no meaning to me.

You might be speaking of an "index array", likely meaning "an array containing ordered subscripts of another array."

You might be using the word "index" as a pseudo-synonym for "subscript."

But the meaning of "indexed array" eludes me.

MCM

[Rodney Hicks]

Like I said, send those new feature suggestions to the address above.

I sent mine in over a year ago. However the required calculations are so simple that I can see why something like that would be on the back burner.

[Gösta H. Lovgren-2]

In my accounting days, AVERAGE was the average, MEAN was the closest of the figures to the average,

Okay Rodney I get it now. In an average of whole numbers for example, the Average could be 5.5 while the Mean would be 5 or 6 (if ether was used to get the Average.) Or perhaps a better way to say is the Mean is actual number nearest to the Average.

And Gösta,
Mediocre men use MEDIAN, mean men use MEAN, average men use AVERAGE, and real men tell someone else to do it!!

And yet again, you clarifed a deep and puzzling mystery for me. Thanks. My gratification runs deeper than the widest river.

======================================================
"How can I lose to such an idiot?"
A shout from chessmaster Aaron Nimzovich (1886-1935)
======================================================

[Paul Dixon]

Gösta,
"Mean" is the correct term but average means the same thing. It's the sum of all your data divided by the number of data items.

"Median" is the value nearest the middle of a sample.

"Mode" is the most often occurring value.

e.g. in the group 1,2,2,3,4,5,5,5,6

The mean is (1+2+2+3+4+5+5+5+6)/9 = 33 / 9 = 3.666..
The average is also 3.666..

The median is 4 as it's the middle value of the sample when the values are put in numerical order.

The Mode is 5 since the number 5 occurrs more than any other number in the sample.

Paul.

[Michael Mattias]

However the required calculations are so simple that I can see why something like that would be on the back burner.

Well maybe, just maybe...

... the 'burner position' is based on a very rational number - the number of users who actually invest the five minutes it takes to submit a new feature suggestion.

I know if I have five users request a new feature for The Provider Payment Partner (tm) System, it gets a better 'burner position' than does the feature requested by one user.

Besides, if you can't take the time to send in a suggestion, how badly do you really want it?

MCM

[Gösta H. Lovgren-2]

Gösta,
"Mean" is the correct term but average means the same thing. It's the sum of all your data divided by the number of data items.

e.g. in the group 1,2,2,3,4,5,5,5,6

The mean is (1+2+2+3+4+5+5+5+6)/9 = 33 / 9 = 3.666..
The average is also 3.666..

Paul.

Thanks, Paul. I guess my question is why have two different words if they "mean" exactly the same thing? Jeez, I was feeling all happy when Rodney differentiated between the two (his mean would be 4 and maybe 3), but now I'm back in the river of consternation. {deep sigh}

======================================
"The things that will destroy us are:
politics without principle;
pleasure without conscience;
wealth without work;
knowledge without character;
business without morality;
science without humanity;
and worship without sacrifice."
Mahatma Gandhi
======================================