[Datadraw-user] BNF in 42

[Bill C. <bi...@bi...>]

Hi, Richard.

I've written a ton of lex/yacc flex/bison based parsers.  Here's one
thing I've found.  Most formats are context sensitive.  You just can't
parse them the easy way, but have to build a parse tree and then
traverse it manually to extract data.  Icky, but true.  This is
massively true for VHDL, even thought the Preface of the VHDL Language
Reference says it's not.  It's even true for Verilog.  Simple formats,
like Synopsys Liberty format, are massively context dependent.

In 42, I figured I'd give up on allowing generic code to execute for
each matched rule, and just go ahead and build a parse tree for the
user.  That separates all the application specific code from the grammar
definition, and should help make the grammars more reusable and easier
to read.

I allow strings in the grammar rules to be lex-style regular
expressions.  What do you think?

The attached file is a grammar that I currently parse to build a parser
rule database.  Then, I'm planning to write a BNF interpreter and
re-parse this file using the rules defined in the file.  Once that
works, I was planning to start writing the rest of the core 42 parser in
42 directly, and then start on the interpreter for it.  When the
interpreter is running, I can start writing libraries to compile more
advanced 42 constructs into simpler code which can be directly written
out in C.  Anyway, that's the plan... too bad the holidays wont last
forever - I'll probably have to get too focused on real work to make
much progress come January.

Regards,
Bill

Regards,
Bill