marat1961/oz-cocor-delphi

Coco/R is a compiler generator, which takes an attributed grammar of a source language and generates a scanner and a parser for this language. The scanner works as a deterministic finite automaton. The parser uses recursive descent. LL(1) conflicts can be resolved by a multi-symbol lookahead or by semantic checks. Thus the class of accepted grammars is LL(k) for an arbitrary k.

Oz Coco/R

Coco/R is a compiler generator, which takes an attributed grammar of a source language and generates a scanner and a parser for this language. The scanner works as a deterministic finite automaton. The parser uses recursive descent. LL(1) conflicts can be resolved by a multi-symbol lookahead or by semantic checks. Thus the class of accepted grammars is LL(k) for an arbitrary k.

Why this project appeared

I've been using the turbo pascal version of Coco / R for a long time. Sometimes it was necessary to change the code. If the sources changed, I recompiled it. But after some Windows update, I found that Turbo Pascal go to a better place and stopped running. In general, I think many people's languages from Niklaus Emil Wirth cause dislike.

In general, I believe that there are few such people who have made a comparable contribution to the development of programming and computer science languages.

It is easy to find a complex and often incomprehensible solution to a problem. It is difficult to make a simple, clean and understandable solution.

When you see such a decision, it becomes clear that this is the work of the Grand Master with a capital letter. First, I ported the code from Turbo Pascal. But I discovered that in almost 20 years a lot of water has flowed under the bridge and the COCO/R code has also been improved well.

The last version I found was C ++, C #, Java. Then I decided to port the code and chose C #. In spirit, this is the language closest to Delphi, probably due to the fact that they have one chief architect.

1. The most important improvements in my opinion are the support for LL (k) grammar.
2. The utf-8 support is also very useful.
3. Great attention is paid to the good quality of the generated code.
4. Also attention was paid to the separate CocoLib unit. I developed this library when I was using the Turbo Pascal version. I usually always include it unchanged in my projects. using Coco / r.

You can see an example of use in the protobuf-delphi project.

New features

Unified class and frame structure

When using COCO/R, it is assumed that a similar class structure will be used: - Oz.Cocor.Utils - classes for working with bits and sets; - Oz.Cocor.Lib - base classes for scanner and parser; - table - data structures of the language under development; - gen - code generator for the language we're working on; - options - program settings; - main program.

It is supposed to use the same set of frames for scanner and parser for different languages. The compiler relies on a fairly fixed frame structure for scanner and parser. It is not desirable to change the order of sections, as this can lead to compilation errors in the generated code.

Syntax specifics

Namespace

Now for the current implementation it is mandatory to specify this parameter. Otherwise code will be generated with errors. Example of running a program

Cocor.exe -namespace Taste -checkEOF -trace AG taste.atg.

Namespace is used when generating a module name. This is now used in the scanner & parser frames.

Macros

Added the ability to define a simple macro, a test replacement. Three macros are added to the program by default: - scanner - parser - namespace

Prefix

When generating types it is possible to add prefix (usually two or three letters). For example, the scanner name will be generated by T + prefix + 'Scanner'. This allows you to get rid of coincidences with already existing types from Delphi and other libraries.

MACROS
  prefix = "cr" .

oz-cocor-delphi's Issues

Add syntax features to the readme.

Compile error when compiling grammar that uses IGNORECASE

Hi,

1st, great parser !

When you have IGNORECASED in your grammar, it creates 2 compile errors in the resulting scanner. I fixed them with the following but I can't tell if this is the correct fix or not

oz.cocor.dfa Line 1600

from gen.WriteLine(' if ch <> Buffer.EOF then');
to gen.WriteLine(' if ch <> #26 then');

Line 1603 from gen.WriteLine(' ch := char.ToLower(char(ch));');
to gen.WriteLine(' ch := character.ToLower(char(ch));');

Scanner.frame needs to add

uses
character;

in implementation

Show errors and warnings separately.

Update implementation using SGL collections

An error occurs when a comment is used inside a code section

96 strLit (. // todo: Process Import Statement .)
97 ";"
98 .
99
100 // Package
101
102 Package (. var
error ^ missing end of previous semantic action
103 id: string;
104 obj: PObj; .)

Change the name for the compilation listing (instead of errors.lst to Gramma.lst)

search path

....\Components\grdJv
....\Components\RxLib\units
....\Components\FastMM4

cannot comple. Dependent directories does not exists in this repository. Please add a link where are these from.

Pascal code inside attribute declaration

An error is thrown when trying to use the string '"proto2"'.
It is necessary to allow the use of strings according to the syntax of the Pascal language.

Taste: add error diagnostics

Taste: attach to example llvm

Write a letter to our comrades.

http://www.ssw.uni-linz.ac.at/Coco/Contributions/