# Reverse Read **URL:** \ **Category:** Forum\ **Tags:** Adabas-Natural, Natural, Natural-Code-Samples\ **Created:** [November 26, 2007, 12:07pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746 "2007-11-26T12:07:18Z")\ **Posts on this page:** 6\ **Page:** 1
**Author:** ![tcs](https://avatars.discourse-cdn.com/v4/letter/t/76d3ee/32.png) [@tcs](https://techcommunity.softwareag.com/u/tcs)\ **Post date:** [November 26, 2007, 12:07pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746/1 "2007-11-26T12:07:18Z")
Hi, IP := 2797 SPRINGHILL DR ALBANY, OR, 23456 #TEMP := 2797 SPRINGHILL DR ALBANY OP1 := 2797 SPRINGHILL DR OP2 := ALBANY In the samples shown above i want the outputs has shown in in OP1 and OP2. First step i am using an examine to find ‘,’ and moving that to the #Temp Variable. From the #Temp variable i want to do an reverse read to find the position of the first space and from that position move that to OP2. thanks, TCS 😃 ---
**Author:** ![Steve\_Robinson](https://sea1.discourse-cdn.com/techcommunity/user_avatar/techcommunity.softwareag.com/steve_robinson/32/107_2.png) [@Steve\_Robinson](https://techcommunity.softwareag.com/u/Steve_Robinson)\ **Post date:** [November 26, 2007, 5:43pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746/2 "2007-11-26T17:43:15Z")
There are several ways to do this; EXAMINE BACKWARD #TEMP would work or MOVE #TEMP (PM=I) TO #TEMP followed by an EXAMINE or SEPARATE #TEMP My guess is that the SEPARATE might be the most efficient, but I have not done a comparison as yet; perhaps later today. One more possibility, do a SEPARATE on IP and retain the delimiters, then look for the first comma steve ---
**Author:** ![Steve\_Robinson](https://sea1.discourse-cdn.com/techcommunity/user_avatar/techcommunity.softwareag.com/steve_robinson/32/107_2.png) [@Steve\_Robinson](https://techcommunity.softwareag.com/u/Steve_Robinson)\ **Post date:** [November 26, 2007, 6:51pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746/3 "2007-11-26T18:51:12Z")
Okay, here is a PC timing comparison of two approaches DEFINE DATA LOCAL 1 IP (A50) INIT \<‘2797 SPRINGHILL DR ALBANY, OR, 23456’\> 1 #PIECES (A15/1:20) 1 OP1 (A50) 1 OP2 (A50) 1 #INDEX (I2) 1 #COMMA (I2) 1 #BLANK (I2) 1 #OP2LENGTH (I2) 1 #LOOP (I4) 1 #CPU-START (I4) 1 #CPU-ELAPSED (I4) END-DEFINE \* MOVE _CPU-TIME TO #CPU-START SETA. SETTIME FOR #LOOP = 1 TO 10000 SEPARATE IP INTO #PIECES (_) WITH RETAINED DELIMITERS EXAMINE #PIECES (\*) FOR ‘,’ GIVING INDEX #INDEX MOVE #PIECES (#INDEX - 1) TO OP2 COMPRESS #PIECES (1:#INDEX - 2) INTO OP1 END-FOR COMPUTE #CPU-ELAPSED = \*CPU-TIME - #CPU-START WRITE ‘separate ip’ \*TIMD (SETA.) #CPU-ELAPSED * * * MOVE \*CPU-TIME TO #CPU-START SETB. SETTIME FOR #LOOP = 1 TO 10000 EXAMINE IP FOR ‘,’ GIVING POSITION #COMMA EXAMINE DIRECTION BACKWARD SUBSTRING (IP,1,#COMMA) FOR ’ ’ GIVING POSITION #BLANK MOVE SUBSTRING (IP,1,#BLANK) TO OP1 ADD 1 TO #BLANK COMPUTE #OP2LENGTH = #COMMA - #BLANK MOVE SUBSTRING (IP,#BLANK,#OP2LENGTH) TO OP2 END-FOR COMPUTE #CPU-ELAPSED = \*CPU-TIME - #CPU-START WRITE ‘examine ip’ \*TIMD (SETB.) #CPU-ELAPSED END Page 1 07-11-26 08:16:44 separate ip 7 66 examine ip 2 19 On the mainframe I think the second approach will look even better since the SUBSTRING facility there was recently rewritten and is substantially more efficient than on the PC. steve ---
**Author:** ![Weihnachtsbar](https://sea1.discourse-cdn.com/techcommunity/user_avatar/techcommunity.softwareag.com/weihnachtsbar/32/146_2.png) [@Weihnachtsbar](https://techcommunity.softwareag.com/u/Weihnachtsbar)\ **Post date:** [November 26, 2007, 7:02pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746/4 "2007-11-26T19:02:31Z")
Steve Robinson wrote: > [@](#): > > On the mainframe I think the second approach will look even better But you will get: ```auto NAT0132 Operand not defined or not of alphanumeric/binary format. ``` during catalog. Even with the current NAT42x ```auto EXAMINE DIRECTION BACKWARD ``` is not supported. ---
**Author:** ![Steve\_Robinson](https://sea1.discourse-cdn.com/techcommunity/user_avatar/techcommunity.softwareag.com/steve_robinson/32/107_2.png) [@Steve\_Robinson](https://techcommunity.softwareag.com/u/Steve_Robinson)\ **Post date:** [November 26, 2007, 7:49pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746/5 "2007-11-26T19:49:22Z")
My bad. Difficult to keep track of which options are on which platform. Here is a rewrite of the second approach: MOVE \*CPU-TIME TO #CPU-START SETC. SETTIME FOR #LOOP = 1 TO 10000 EXAMINE IP FOR ‘,’ GIVING POSITION #COMMA MOVE SUBSTRING (IP,1,#COMMA) (PM=I) TO #WORK EXAMINE #WORK FOR ’ ’ GIVING POSITION #BLANK COMPUTE #BLANK = #BLANK - 2 MOVE SUBSTRING (#WORK,2,#BLANK) (PM=I) TO OP1 MOVE SUBSTRING (IP,#BLANK) TO OP2 END-FOR COMPUTE #CPU-ELAPSED = \*CPU-TIME - #CPU-START WRITE ‘examine no backward ip’ \*TIMD (SETC.) #CPU-ELAPSED Here are the three times on the PC ``` PAGE # 1 DATE: Nov 26, 2007 PROGRAM: STRIP03 LIBRARY: INSIDE ``` separate ip 7 67 examine ip 2 19 examine no backward ip 2 20 So on the PC, the third solution is slightly slower than the second. On the mainframe, I ran the first and third (second, as noted, will not compile) separate ip 2 11 examine no backward ip 1 6 So the ratio is about two to one in favor of the last solution. steve ---
**Author:** ![system](https://global.discourse-cdn.com/techcommunity/original/3X/c/d/cd19967bb605d28a72d77eb83a1cae22f3df2eb8.svg) [@system](https://techcommunity.softwareag.com/u/system)\ **Post date:** [April 2, 2021, 2:48pm UTC](https://techcommunity.softwareag.com/t/reverse-read/75746/6 "2021-04-02T14:48:56Z")