Bug in Harbour Descend()

[Enrico Maria Giordano]

This is a sample of the problem:

Code: Select all  Expand view  RUN

REQUEST HB_CODEPAGE_ITWIN


FUNCTION MAIN()

    LOCAL aData[ 7 ]

    LOCAL i

    HB_SETCODEPAGE( "ITWIN" )

    aData[ 1 ] = { "TEST1", 3107.77 }
    aData[ 2 ] = { "TEST2", 852.07 }
    aData[ 3 ] = { "TEST3", 191.00 }
    aData[ 4 ] = { "TEST4", 148.68 }
    aData[ 5 ] = { "TEST5", 44.73 }
    aData[ 6 ] = { "TEST6", 15.24 }
    aData[ 7 ] = { "TEST7", 255.65 }

    ASORT( aData, , , { | aItem1, aItem2 | Descend( Str( aItem1[ 2 ], 10, 2 ) ) + aItem1[ 1 ] < Descend( Str( aItem2[ 2 ], 10, 2 ) ) + aItem2[ 1 ] } )

    FOR i = 1 TO LEN( aData )
        ? aData[ i, 1 ], aData[ i, 2 ]
    NEXT

    INKEY( 0 )

    RETURN NIL

Result:

Code: Select all  Expand view  RUN

TEST1       3107.77
TEST2        852.07
TEST5         44.73
TEST6         15.24
TEST7        255.65
TEST4        148.68
TEST3        191.00

Any workaround?

EMG

[hmpaquito]

IMHO Test is erronous

Sorting so:

Code: Select all  Expand view  RUN

ASORT( aData, , , { | aItem1, aItem2 | Descend( Str( aItem1[ 2 ], 10, 2 )  + aItem1[ 1 ] ) < Descend( Str( aItem2[ 2 ], 10, 2 )  + aItem2[ 1 ]  )} )

Regards

[Enrico Maria Giordano]

No, as I want to sort for the second item descending and the first item ascending. Anyway, it doesn't work either.

EMG

[Enrico Maria Giordano]

The problem seems to be the codepage. Without setting it, the sort order is correct. But the codepage should not affect the behaviour of Descend() function, should it?

EMG

[nageswaragunupudi]

I am not commenting on the behaviour of the Descend() function, but I would have approached it in a different way:

Code: Select all  Expand view  RUN

ASort( aData,nil,nil, { |x,y| If( x[ 2 ] == y[ 2 ], x[ 1 ] > y[ 1 ],  x[ 2 ] > y[ 2 ] ) } )
 

[Enrico Maria Giordano]

Thank you. It is not feasible as it would require to change all the ASort() calls. I prefer to remove the codepage.

EMG

[hmpaquito]

Similar discussions are:

https://groups.google.com/g/harbour-use ... 8IIuDBQAJ+
https://groups.google.com/g/harbour-use ... gkZEOLy3sJ
https://groups.google.com/g/harbour-use ... S1EeUNAgAJ
viewtopic.php?f=6&t=29052&p=163632&hilit=descend&sid=17d6d2aeb2ab1b539390e4fd93f78e55#p163632
http://www.pctoledo.com.br/forum/viewto ... 21&start=0

Solution and explication:

https://groups.google.com/g/harbour-use ... Ofx4M4ECsJ

[Enrico Maria Giordano]

It looks like the bug has never been fixed.

EMG

[AntoninoP]

have you tryied using StrZero instead of Str?

[Enrico Maria Giordano]

Yes, the order changes but it's not correct either.

EMG

[hmpaquito]

The purpose of Descend () is to be Clipper compatible. Therefore it is not an error, but an unexpected behavior for those who use codepages
The existence of hb_Descend (), compatible with the codepage system, would be highly desirable.

[AntoninoP]

I see that if you remove HB_SETCODEPAGE from your example the array looks correctly sorted...
I investigated a little and I found that there is a flag in the code pages "BinarySort" that indicate if the comparison can be simply binary or more complex, it is true for "en" and false for "itwin"

Try this:

Code: Select all  Expand view  RUN

    ? chr(199)<chr(205) //--> T
    HB_SETCODEPAGE( "ITWIN" )
    ? chr(199)<chr(205) //--> F
 

the problem is not the descend... very sad

[Enrico Maria Giordano]

How do you deal with this bug? I need to use codepage with Harbour, otherwise the accented chars are not correctly read from the file system (ie. filenames, Directory(), File(), Memoread(), etc.). But doing so, Descend() function is not working anymore. I can't even build index with Descend() in the key. Any help, please.

[karinha]

Master Enrico, see if it helps:

https://linguagemclipper.com.br/dicas/arrays

Regards, saludos.

[Enrico Maria Giordano]

Thank you, but I can't found anything about Descend() function in that web page. I need to use Descend() in Harbour just like I use it in xHarbour and used in Clipper. Please look at my first message on this thread for a sample of the problem.