issues with f$file_format lexical function

[Hein_vdHeuvel_d]

Thank you cachaza for your contribution. It is clear enough.

You are correct, IF-and-only-IF, the file is fixed length record sequential then you can calculate the number of records from eof-block * 512 + ffb / size. 

Note: Size should be picked up from LRL, not MRS, and should be rounded up to even value before dividing.

I the file is larger than 2GB (4194303 blocks) you indeed need a (cobol) helper program to do the math with more than 32 bits, but for most files you can just use DCL divides.

Too bad that RMS does not do file-length-hints for this easy file type huh!

I had an Email exchange with John... Searching for and empty string makes sure that there is an immediate hit for each record, so that's quicker and lookign at each bytes. Also, this avoids the 'no match found' message to sys$error.

Cheers,

Hein

[H.Becker]

> COMPUTE NUM-REG ROUNDED = (((( PAR-EOF9 - 1 ) * 512 ) + PAR-FFB9 ) / PAR-MRS9 )

>You are correct, IF-and-only-IF, the file is fixed length record sequential then you can calculate the number of records from eof-block * 512 + ffb / size. 

The right formula - subtracting 1 or not - depends on ffb. The one in your Cobol source is correct for ffb>0. The other one  - insert the appropriate parentheses - is correct for ffb==0 (which makes the addition somehow obsolete).

$ f:=whatever

$ cre 'f

whatever

 Exit 

$ write sys$output f$file(f,"eof")

1

$ write sys$output f$file(f,"ffb")

10

$ ! here rfm=var and the file size is 10: (eof-1)*512+ffb

$ convert/fdl=tt: 'f 'f.fix/pad=%x20

FILE

ORGANIZATION            sequential

RECORD

CARRIAGE_CONTROL none

FORMAT fixed

SIZE 512

 Exit 

$ write sys$output f$file(f+".fix","eof")

1

$ write sys$output f$file(f+".fix","ffb")

0

$ ! the file size is 512: eof*512+ffb

$