Jack,
Kalle's example (with correction) works for me. Complete example (assuming you want your own terminal):
$ PIPE WRITE SYS$OUTPUT -
"tcpip show device/full ",F$TRNLNM("TT")-"_" |-
ANALYZE/SYSTEM |-
SEARCH sys$pipe "Port:" | -
(READ sys$pipe p ; -
p=F$ELEMENT(3,":",p) ; -
DEFINE/JOB/NOLOG port &p)
$ SHOW LOG/JOB port
For a more general mechanism for scripting variable input into programs, complex sequences of commands, either place multiple "WRITE SYS$OUTPUT" commands within brackets in the first pipe stage, or execute another command procedure to generate the output lines. (I tend to recurse into myself with p1 as a tag to avoid multiple files). Example for this problem:
$ IF F$EXTRACT(0,5,p1).EQS."STAGE" THEN GOTO 'p1'
$ self=F$PARSE(";",F$ENVIRONMENT("PROCEDURE"))
$ PIPE @'self' STAGE1 |-
@'self' STAGE2 |-
@'self' STAGE3 |-
@'self' STAGE4
$ EXIT
$
$ STAGE1:
$ WRITE SYS$OUTPUT "tcpip show device/full "+F$GETDVI(F$TRNLNM("TT"),"DEVNAM")-"_"
$ EXIT
$
$ STAGE2:
$ DEFINE/USER SYS$INPUT SYS$PIPE:
$ ANALYZE/SYSTEM
$ EXIT
$
$ STAGE3:
$ SEARCH SYS$PIPE: "Port:"
$ EXIT
$
$ STAGE4:
$ READ SYS$PIPE p
$ p=F$ELEMENT(3,":",p)
$ DEFINE/JOB/NOLOG port 'p'
$ EXIT
Notes:
For this case, there are some peculiarities in device and logical names which require F$GETDVI to be used in this context to determine the terminal name, but not from the command line. The leading "_" also needs to be stripped for the SDA extension to recognise the device.
In general, only the complex cases need the self recursion. One liners can be coded directly into the PIPE command.
Testing can be done from the command line by invoking the stages explicitly:
$ PIPE @proc STAGE1 | @proc STAGE2
etc...
It also sometimes makes sense to parameterise the input pipe:
$ PIPE ...
@'self' STAGE3 SYS$PIPE: "Port:" | -
etc...
$ STAGE3:
$ SEARCH 'p2' "''p3'"
This means the stages can be used to process files, as well as piped input. Thus:
$ PIPE @proc STAGE1 | @proc STAGE2 > check.tmp
$ @proc STAGE3 check.tmp "Port:"
A crucible of informative mistakes
