Hi folks,

We are trying to migrate a complete appliacation from one platform (VME)
to a completely different platform, running a completely different OS
(Linux), while at the same time trying to make it look as similar as
possible, if not identical, to the end users. All this is done using tools
written in house to do as much of the conversion as possible automatically.

Getting something to convert VME Cobol to Micro Focus cobol is not exactly
difficult, (INITIALISE to INITIALIZE for example) and converting the IDMSX
database to Oracle, while 'interesting' is again relatively straightforwards
if somewhat tedious.

However we also have to convert something of the order of 2400 SCL
programs. SCL is VME's job control language and, for those that might know
IBM's JCL, has a similar function. SCL, however, is a fully fledged
compiled language in it's own right, yet can still be used from the
command line. It is an integer only subset of a language caled S3, which
is itself a superset of Algol-68.

What we need is help in parsing, and in particular help in parsing
conditions. The guy who is writing this part of the converter is using
perl 5 on windows and is asking for help with something called a descent
parser???

His email to me is attached inline below.

The sample is entirely made up as for confidentiality reasons we cannot
access the clients code from outside their own servers. Although this
example only contains IF conditions they could also be UNLESS or
WHENEVER, and also UNTIL on loops.

Is anyone able to help us with guidance on the parsing, or could offer us
a better alternative?

Many thanks,

Dave


From: Robert Hood
Sent: 14 December 2016 12:40
To: Dave Stratford
Subject: Perl Issue


Hello can anybody help with a best perl solution

I believe we would have to use Parse::RecDescent for this (or maybe
another if it is
simpler)

We need to change data in the format marked (1) below to the format marked
(2)

===== Start Of Input Sample =====

IF ( SAMENDED EQ "HELLO" AND SDATE EQ "TODAY" ) OR ( SAMENDED EQ "GOODBYE"
AND SDATE
EQ "TODAY" )
THEN
SENDMESSAGE ( "HELLO" )
ELSF ( SAMENDED EQ "AFTERNOON" AND SDATE EQ "TOMORROW" ) OR ( SAMENDED EQ
"GOODBYE"
AND SDATE EQ "TOMORROW" )
THEN
SENDMESSAGE ( "AFTERNOON" )
ELSE
SENDMESSAGE ( "GOODNIGHT" )
FI

===== End of input sample =====

===== Start Of Output Sample =====

@COBOL*01  EVALUATE TRUE
@COBOL*01  WHEN ( O2L::EQ(SAMENDED, z"HELLO") = 1 AND O2L::EQ(SDATE,
z"TODAY") = 1 )
@COBOL*01    OR ( O2L::EQ(SAMENDED, z"GOODBYE") = 1 AND O2L::EQ(SDATE,
z"TODAY") = 1 )
@COBOL*01    CALL "SENDMESSAGE" USING "HELLO"
@COBOL*01  WHEN ( O2L::EQ(SAMENDED, z"AFTERNOON") = 1 AND O2L::EQ(SDATE,
z"TOMORROW") = 1 )
@COBOL*01    OR ( O2L::EQ(SAMENDED, z"GOODBYE") = 1 AND O2L::EQ(SDATE,
z"TOMORROW")
= 1 )
@COBOL*01    CALL "SENDMESSAGE" USING "AFTERNOON"
@COBOL*01  WHEN OTHER
@COBOL*01    CALL "SENDMESSAGE" USING "GOODNIGHT"
@COBOL*01  END-EVALUATE.

===== End Of Output Sample =====

The biggest problem with Parse::RecDescent is a good example that covers
this
creating the grammars

Robert Hood
Migration Consultant Professional Services > Transoft > Advanced

-- 
Dave Stratford - ZFCB
http://daves.orpheusweb.co.uk/

In comp.lang.perl.misc, Dave Stratford  <daves@orpheusmail.co.uk> wrote:
> However we also have to convert something of the order of 2400 SCL
> programs. SCL is VME's job control language and, for those that might know
> IBM's JCL, has a similar function. SCL, however, is a fully fledged
> compiled language in it's own right, yet can still be used from the
> command line. It is an integer only subset of a language caled S3, which
> is itself a superset of Algol-68.

Coming from a background where you did this stuff in yacc and lex,
I was tempted to use the Parser::Input::Yacc module when I wanted to
create a shell language in Perl. I quickly found that to be buggy.
Parse::RecDescent worked great, though.

> What we need is help in parsing, and in particular help in parsing
> conditions. The guy who is writing this part of the converter is using
> perl 5 on windows and is asking for help with something called a descent
> parser???

Parse::RecDescent has a copyright dating to 1997, and has been used in a
huge number of projects. Without difficulty I found tutorials dating
back to 2001 on the web. I think your developer needs to learn how to
use web search engines on his own.

> Is anyone able to help us with guidance on the parsing, or could offer us
> a better alternative?

Parse::RecDescent is very likely the best Perl tool for the job. But you
need to understand parsing regular grammars. 

Elijah
------
wikipedia has a page on recursive descent parsers

In article <eli$1612141943@qz.little-neck.ny.us>,
   Eli the Bearded <*@eli.users.panix.com> wrote:
> In comp.lang.perl.misc, Dave Stratford  <daves@orpheusmail.co.uk> wrote:
> > However we also have to convert something of the order of 2400 SCL
> > programs. SCL is VME's job control language and, for those that might know
> > IBM's JCL, has a similar function. SCL, however, is a fully fledged
> > compiled language in it's own right, yet can still be used from the
> > command line. It is an integer only subset of a language caled S3, which
> > is itself a superset of Algol-68.

> Coming from a background where you did this stuff in yacc and lex,
> I was tempted to use the Parser::Input::Yacc module when I wanted to
> create a shell language in Perl. I quickly found that to be buggy.
> Parse::RecDescent worked great, though.

> > What we need is help in parsing, and in particular help in parsing
> > conditions. The guy who is writing this part of the converter is using
> > perl 5 on windows and is asking for help with something called a descent
> > parser???

> Parse::RecDescent has a copyright dating to 1997, and has been used in a
> huge number of projects. Without difficulty I found tutorials dating
> back to 2001 on the web. I think your developer needs to learn how to
> use web search engines on his own.

> > Is anyone able to help us with guidance on the parsing, or could offer us
> > a better alternative?

> Parse::RecDescent is very likely the best Perl tool for the job. But you
> need to understand parsing regular grammars. 

I think this is where he's specifically looking for help.

Dave

> Elijah
> ------
> wikipedia has a page on recursive descent parsers

-- 
Dave Stratford - ZFCB
http://daves.orpheusweb.co.uk/

Eli the Bearded <*@eli.users.panix.com> writes:
<snip>

> Parse::RecDescent is very likely the best Perl tool for the job. But you
> need to understand parsing regular grammars.

I think that's a typo or some other misunderstanding.  Regular grammars
are those that define regular languages -- the same languages matched by
(un-extended) regular expressions.  So in some rather fussy sense the
statement is correct because RecDescent uses REs to match tokens.

The grammars that can parsed by RecDescent will be some subset of the
context free grammars, possibly those that go by the name of LL(k)
grammars.  The documentation may say exactly what grammars can be
parsed.

<snip>
-- 
Ben.