Hi.

I am currently looking into integrating a custom language parser into my program and have been
searching the web for parser generators that can create Pascal code from a custom language.

So far in my search i have found the following "options":

* GOLD Parser: http://www.goldparser.org/
Was last updated in sept. 2012 and the lastest pascal implementation is not for the lastest version of the program!

* COCO/R: http://www.ssw.uni-linz.ac.at/Research/Projects/Coco/ Same as with the GOLD parser - last updated in 2014 and lastest pascal implementation is for an older version of the program.

* TP Yacc/Lex: (included in Free Pascal)
Is included with Free Pascal, but the generated code is non-OO, and seems to rely on text-files for input.

So if there is anyone out there who knows of such a generator I would be glad to know.

Kind regards,
Torsten Bonde Christiansen.

I am currently looking into integrating a custom language parser into my program and have been searching the web for parser generators that can create Pascal code from a custom language.

No parser generator that I know includes a code generator, you must write that part would write a handwritten transpiler instead, though. Easier maintenance and maximum flexibility, no rules or restrictions as imposed by every parser generator implementations. At the worst, I would pick Coco/R since it generates a recursive descent parser instead of table driven one, still possible for hand editing.

There is a german learning programm named AtoCC. It can generate Object Pascal Code:

http://www.atocc.de/cgi-bin/atocc/site.cgi?lang=en&site=main

Sorry, I had no time to test it yet.

Marc

"For target code generation you need to specify code fragments for each rule in VCC."

I think the important thing for all of these is that they still assume a lot of hard work from the person defining the custom syntax. They're not a magic bullet.

For the last 30 years or so I've been using a compiler compiler originally written in 1964, the implementation has matured as different pascal compilers have become available. I'm not suggesting that this is usable in the general case, but looked at objectively it does allow me to make gross behavioural changes fairly easily, e.g. to define novel string properties and to embed pragmata in comments.

It's been my experience that all the compiler compilers have one or more flaws that make it more work in the end to use them than to just write a parser by hand.

Örks. Didn't read that.

I think the important thing for all of these is that they still assume a lot of hard work from the person defining the custom syntax. They're not a magic bullet.

There are some tools I liked very much when dealing with language translation in the past, but they don't fit her. None of them generates Pascal Code out of the box.

ANTLR3 was my favourite, generating tree walkers with little extra work (iirc). But it generates Java code.

Beautiful. Would you mind disclosing the name of that fine tool? ;)

Marc

Marc Santhoff wrote

There are some tools I liked very much when dealing with language translation in the past, but they don't fit her. None of them generates

^e
"they don't fit here."

Fatal typo. I know Torsten is not a girls name.

Marc

Meta-II, by Val Schorre. Wikipedia has an adequate description, I first implemented it in the days of Pascal MT+ and since then have ported it to Turbo, Topspeed, Delphi and FPC. I'm not saying I've done a particularly good job, and will probably redo it for UTF16 (whatever the correct FPC type is for that) at some point- possibly adding support for Smalltalk-style keywords.

Obvious changes included saving more state to allow backtracking. Less obvious was parsing comments as part of the syntax (which allows things like XML and SNMP descriptions to be parsed), allowing string delimiters to be specified and keeping the string delimiter as a string property. Those last allow runtime behaviour of strings to vary depending on delimiter, so for example a string expression using / as delimiter can be used as a regex while a string lvalue using / as delimiter can be a backref into a pattern.

Schorre or his coworkers later defined something called Tree Meta which accumulates parsed source into a tree, and also has rules for how the tree can be manipulated and "unparsed". I don't believe that anybody's reimplemented it, but I'm occasionally tempted...

:) Thanks!

-Torsten

Previously I have worked with TP Lex/Yacc and it works fine with limitations it has. And yes, using a parser generator is not a magic wand that creates code to run without modification, but on most occations I find that the post-processing required with this approach is far less than writing my own parser.

kind regards,
Torsten Bonde Christiansen.