Conference turris::fortran

1176.0. "collating sequences : FOR-F-INVMATKEY" by MUNICH::WWERNER (When in doubt, do as the INTs do) Fri Feb 14 1997 10:11

Once there was a problem with collating sequences and keyge/keygt on VAX
(see FORTRAN-VAX 1870). Still a problem on Alpha ? The workaround is the
same as 1992, btw. ;-)

Versions used: OpenVMS Alpha 6.2, DEC Digital Fortran 77 V7.1-107

Wolfgang



$ TYPE R.FOR
c this is the 'reader'
        character*2 mykey
        character*2 buf
        open (unit=10,file='test1.dat',accesS='keyed',
     1        form='unformatted',organization='indexed',status='old',
     2	      key=(1:2:character:ascending),
     3        recordtype='fixed', recl=2)
        mykey='A1'
c this works
c        read (10,keyid=0,keynxt=mykey) buf

c this gives a FOR-F-INVMATKEY 
        read (10,keyid=0,keyge=mykey) buf
        type *,buf
        close (10)
        end

$ TYPE W.FOR
c indexed file is written like this
        open (unit=10,file='test.dat',accesS='keyed',
     1        form='unformatted',organization='indexed',status='new',
     2        key=(1:2:character),
     3        recordtype='fixed', recl=2)
        write (10) 'A1'
        write (10) '�1'
        write (10) 'O1'
        write (10) '�1'
        write (10) 'U1'
        write (10) '�1'
        close (10)
        end

This is the FDL File

SYSTEM
	SOURCE                  OpenVMS

FILE
	ALLOCATION              9
	BEST_TRY_CONTIGUOUS     no
	BUCKET_SIZE             2
	CLUSTER_SIZE            3
	CONTIGUOUS              no
	EXTENSION               0
	FILE_MONITORING         no
	GLOBAL_BUFFER_COUNT     0
	ORGANIZATION            indexed

RECORD
	BLOCK_SPAN              yes
	CARRIAGE_CONTROL        none
	FORMAT                  fixed
	SIZE                    8

AREA 0
	ALLOCATION              9
	BUCKET_SIZE             2
	EXTENSION               0

KEY 0
	CHANGES                 no
	DATA_KEY_COMPRESSION    no
        COLLATING_SEQUENCE      "GERMAN"
	DATA_RECORD_COMPRESSION no
	DATA_AREA               0
	DATA_FILL               100
	DUPLICATES              no
	INDEX_AREA              0
	INDEX_COMPRESSION       no
	INDEX_FILL              100
	LEVEL1_INDEX_AREA       0
	NAME                    ""
	NULL_KEY                no
	PROLOG                  3
	SEG0_LENGTH             2
	SEG0_POSITION           0
	TYPE                    collated

ANALYSIS_OF_AREA 0
	RECLAIMED_SPACE         0

ANALYSIS_OF_KEY 0
	DATA_FILL               11
	DATA_KEY_COMPRESSION    0
	DATA_RECORD_COMPRESSION 0
	DATA_RECORD_COUNT       6
	DATA_SPACE_OCCUPIED     2
	DEPTH                   1
	INDEX_COMPRESSION       0
	INDEX_FILL              2
	INDEX_SPACE_OCCUPIED    2
	LEVEL1_RECORD_COUNT     1
	MEAN_DATA_LENGTH        8
	MEAN_INDEX_LENGTH       4

Your test programs don't match your problem description.  They write and read
different files and explicitly specify normal string key.  Properly written,
they would use the same file and would open an 'OLD' file to write (and read)
the records without specifying a KEY= at all.

After all that, there is a bug in the RTL in that it has hardcoded key types
to determine ascending vs. descending and mis-interprets a COL type as
descending.  Not that this is really a big deal - just use KEYNXT - but it
should be fixed nonetheless.  We'll add it to the list.  Thanks.

				Steve

I've forgotten to add the 'CONV/FDL=TEST.FDL TEST.DAT TEST1.DAT' to 0.

Wolfgang

Ah.  Well, you also need to take the KEY= out from the OPEN in R.FOR.

				Steve