mass confusion on MASK

, ,
1

The help for MASK says that for ‘c’, then one or more positions are checked for c. But is c strictly alpha? What if I say: if blah = MASK(‘99’)? Am I asking it to search for 99 in the first two positions or anywhere in blah?

2

MASK is positional. To look for a string within a variable, use SCAN.

IF  #VAR = SCAN '99'     /* Does 99 exist within #VAR?
3

No, that’s not true. MASK can be positional and not positional, too. To look for ‘99’ in a string, you can use mask, too. And you can use it more efficient than SCAN, like in:

IF #VAR = MASK(*'99') /* Does 99 exist within #VAR?
IF #VAR = MASK('99') /* Does #VAR bgin with '99'?
IF #VAR = MASK(*'99'/) /* Does #VAR end with '99'?

MASK indeed is nearly as powerful as regular expressions, but not that complicated (in my opinion).

4

[quote="Wilfried B

5

No argument, the wild card additions to MASK give it a lot more capability than SCAN.

However, in terms of performance, for the functionality that SCAN was designed for, SCAN is far more efficient than MASK.

For example;

DEFINE DATA LOCAL
1 #LOOP (P9)
1 #STRING (A20)
1 #CPU-TIME (P9)
1 #CPU-START (P9)
END-DEFINE
*
INCLUDE AATITLER
INCLUDE AASETC
*
MOVE CPU-TIME TO #CPU-START
SETA. SETTIME
FOR #LOOP = 1 TO 500000
IF #STRING = MASK (‘X’)
IGNORE
END-IF
END-FOR
COMPUTE #CPU-TIME = *CPU-TIME - #CPU-START
WRITE 5T ‘MASK TIME’ *TIMD (SETA.) 5X #CPU-TIME
*
MOVE *CPU-TIME TO #CPU-START
SETB. SETTIME
FOR #LOOP = 1 TO 500000
IF #STRING = SCAN (‘X’)
IGNORE
END-IF
END-FOR
COMPUTE #CPU-TIME = *CPU-TIME - #CPU-START
WRITE 5T ‘SCAN TIME’ *TIMD (SETB.) 5X #CPU-TIME
*
END
PAGE # 1 DATE: Feb 14, 2007
PROGRAM: SCAN01 LIBRARY: SYSTEM

MASK TIME       49            489
SCAN TIME       39            388

SCAN is about 20 % faster and uses 25 % less CPU time.

If there was an X in the first position of #STRING, the output is:

PAGE #   1                    DATE:    Feb 14, 2007
PROGRAM: SCAN03               LIBRARY: SYSTEM

MASK TIME       36            355
SCAN TIME       35            345

The times above are on my PC; will run mainframe times later today (snowing, hailing and raining outside, so will stay in and play on the computer).

steve

6

Okay, it’s later.

Mainframe numbers are very interesting.

First, here are the numbers with no X

MASK TIME 19 112
SCAN TIME 9 61

The SCAN is twice as fast as the MASK.

With an X in the first position, the numbers are

MASK TIME 9 61
SCAN TIME 9 61

So, if the functionality you are looking for requires “searching” for a text string anywhere in an alpha variable, use SCAN, not MASK.

steve

7

Getting back to the original question:

IF #STRING = MASK(‘AB’)

will look for AB in positions 1 and 2

IF #STRING = MASK(??‘AB’)

will look for AB in positions 3 and 4

? . _ can all be used to indicate "skip a single position

  • % are the “wildcard” (any number of characters) characters

On the PC (mainframe results later) MASK (??‘XX’) is slightly faster (2%) than SCAN, however CPU time is 10% faster than SCAN.

steve

8

Your test on Solaris:

MASK TIME       56            562
SCAN TIME       47            468

But you can make it much more faster if you use an I4-field as a loop counter… :wink:

MASK TIME       29            282
SCAN TIME       19            195

Anyway: SCAN is slightly faster