Re: SCAN vs %scan

On 09-Jan-2012 12:16 , sjl wrote:

It is a one-off program for creating address information during a
system conversion.

As it reads records from the legacy address book file, the program
SCANs for for values in the name field and replace them with other
hard-coded values before writing out the address book information to
the file which will be imported into the new system.

For example:

C 'Abc ' SCAN PNAME P
C IF P > 0
C MOVEA 'ABC ' PL(P)
C ENDIF

The array named PL redefines the field PNAME.

Perhaps I can better improve the program's performance by replacing
the MOVEA with %replace:
Where I have added code in this routine, I have used %replace.

So basically a UDF against the column to effect replacement of some 'tokens' with modified variations; e.g. as with the upper-cased value as shown.?

FWiW: Is the SCAN going to, or even need to, handle replacing the token when found at the end-of-string; or for a ' Abc' at the beginning of a string? That is, the space as token delimiter may limit replacements [in]correctly at the beginning and end of strings.

Because there are apparently "hundreds" of tokens that can be replaced within a string, I would probably just use %replace; i.e. the scan is implicitly a part of that builtin.

I wonder however was there any consideration made for the order of operations or to prevent accidental effects from "rescan" of the same values; i.e. where one replaced value can become a value that will again be replaced, having formed another matching token? For the effect of "no rescan", the scanning for the specific tokens may be required up-front, for which the offset-array may be more desirable.

The example does not show, though implies, the replaced value will always match in length to the scanned value. I would think a loop implementing the effective scan might be a better implementation to limit the amount of times the full string is reprocessed; i.e. scanned from begin-to-end for either an explicit %scan or the implicit scan when using %replace. Checking to see if the current byte in the string matches the first byte of any scan-strings, then checking only for the number of bytes of each /potentially matching/ scan-string; for a match, overwrite the data from that position for the length, and for no-rescan continue scanning from the end of the replaced value.

Regards, Chuck